EFFICIENT VIDEO TRANSMISSION OVER CORRELATED NAKAGAMI FADING CHANNELS F O R IS-95 C D M A S Y S T E M S by NORMAN CHAN B. Eng. (Electrical Engineering), McMaster University, Canada, 1994 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR T H E D E G R E E OF MASTER OF APPLIED SCIENCE in THE FACULTY OF GRADUATE STUDIES DEPARTMENT OF E L E C T R I C A L ENGINEERING We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA April 1999 © Norman Chan, 1999 In presenting degree this thesis in at the University of partial fulfilment British Columbia, of the requirements for that permission copying granted department this or thesis by for scholarly his publication of this thesis or her purposes may representatives. be It for financial gain shall not is hzLl~C7RjC/)/_ The University of British Columbia Vancouver, Canada Date DE-6 (2/88) ApU/ ^ H.^ IrNGltifrFR by understood for extensive the head that be allowed without permission. Department of advanced I agree that the Library shall make it freely available for reference and study. 1 further agree of an I of my copying or my written Abstract This thesis deals with the problem of efficient transmission of video signals over generalized fading channels in Direct Sequence-Spread Spectrum (DS-SS) Code D i v i s i o n M u l t i p l e Access ( C D M A ) systems. The video codec is based upon the I T U H.263 compression algorithm which targets at providing low bit-rate video telephony services suitable for wireless transmission. In order to reduce the overall impact of errors due to mobile channels on the video sequence, a modified version of the original H.263 codec is proposed incorporating a selective Forward Error Correction ( F E C ) coding scheme combined with a forced I N T R A frame update mechanism. This modified version of H.263 codec provides improvement in both average video and frame-to-frame performance. We further consider a coherent D S - C D M A system based upon the IS-95 standard for the forward link (base-to-mobile) i n both single-cell and multiple-cell environment. We provide performance evaluation results by both analysis, employing the Gaussian approximation, and computer simulations, using Monte Carlo error counting techniques. The proposed C D M A system uses concatenated Walsh/maximal-length coding scheme for spectrum spreading. The resulting spread codes maintain orthogonality while reducing inhomogenous cross-correlations among Walsh sequences. The frequency-selective fading channel is modeled by a tapped delay line model with channel coefficients of each path following an independent Nakagami-m distribution. We have implemented in software a correlated Nakagami fading simulator based upon the principle of superposition of complex partial waves with a (possibly strong) component resulting from the direct waves. The time correlation is generated by assigning each partial wave a Doppler shift as a function of time. This approach is an exact replica of the actual physical situation as it reproduces the wave propagation process, as opposed to the Doppler spectrum approximation approach used in other simulators. The received signal is demodulated coherently using a R A K E ii receiver w i t h variable r e s o l v i n g fingers, where multipath components are m a x i m a l - r a t i o combined'. F o r our analysis, we assume perfect knowledge of the channel, w h i c h c o u l d be accomplished either by the usage of pilot tone or some type of channel-parameter estimation circuits. However, for the computer simulation, such perfect knowledge of the channel is not necessary. In terms of performance evaluation results, we first present the improved performance of the modified H.263 codec as a function of Peak Signal-to-Noise Ratio ( P S N R ) transmitted i n additive white G a u s s i a n noise ( A W G N ) environment. T h e n , the a n a l y t i c a l and computer simulated results for the bit error rate ( B E R ) performance of C D M A forward link in Nakagami fading channels for both single-cell and multiple-cell environment are presented. Further, we present the P S N R performance results for the video transmission featuring the modified H.263 coding scheme over the proposed C D M A systems. Finally, a variety of performance evaluation results, both in single-cell and multiple-cell environment, are presented for different number of resolving paths, signal propagation characteristics, cell user capacity, as well as for the presence of channel estimation errors. In all cases, heuristic explanations and interpretations of the trend of the obtained results are also given. Table of Contents H Abstract List of Figures viii Chapter 1 INTRODUCTION 1 Chapter 2 BACKGROUND 5 2.1 Introduction • 2.2 I T U - T H.263 L o w Bit-Rate Video Compression.... --5 '.. 5 2.2.1 Encoded Video Bitstream Structure 8 2.2.2 Bitstream Syntax 8 2.3 M o b i l e Radio Propagation: Large-Scale Fading '. 11 2.4 M o b i l e Radio Propagation: Small-Scale Fading 12 2.4.1 Small-Scale Fading: Parameters and Characterization 12 2.4.2 Small-Scale Fading: Distribution Models 14 : 2.5 Spread Spectrum Modulation Techniques 17 2.6 The R A K E Receiver 19 2.7 IS-95 Code Division Multiple Access System 20 2.8 Chapter 3 2.7.1 C D M A Forward L i n k '. 2.7.2 Direct Sequence Spreading Codes Conclusions ...= 21 23 • 25 M O D I F I E D H.263 V I D E O C O D E C 27 3.1 Introduction 27 3.2 Effects of Errors on H.263 Video 28 3.3 Selective F E C Coding 28 iv 3.4 Forced I N T R A Frame Update 30 3.5 Numerical Results for the A W G N Channel 30 3.6 Conclusions Chapter 4 ...32 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS 34 4.1 Introduction 34 4.2 Direct Sequence-Spread Spectrum (DS-SS) C D M A System M o d e l 36 4.3 M o b i l e Channel M o d e l .37 4.4 Correlated Nakagami Fading Simulator 39 4.5 Receiver Structure 43 4.6 C D M A Forward L i n k B E R Performance Analysis 44 4.7 Forward L i n k B E R Analysis with Gaussian Approximation 48 4.8 Computer Simulation System Description 53 4.9 B E R Performance Evaluation Results for C D M A Forward L i n k 54 4.10 Video Transmission Performance in C D M A Forward L i n k 58 4.11 Conclusions 67 Chapter 5 FORWARD LINK VIDEO TRANSMISSION IN MULTIPLE-CELL CDMA SYSTEMS 69 5.1 Introduction '. 69 5.2 Multiple-Cell Configuration M o d e l 70 5.3 Channel Modeling 71 5.4 Multiple-Cell D S - S S C D M A System M o d e l 72 5.5 5.6 Multiple-Cell B CD A Forward L iAnalysis n k B E R Performance Multiple-Cell EM R Performance with GaussianAnalysis Approximation 73 74 vi 5.7 Computer Simulation Model Description 80 5.8 BER Performance Evaluation Results for Multiple-Cell Systems 80 5.9 Video Transmission Performance in Multiple-Cell Systems 82 5.10 Conclusions Chapter 6 90 CONCLUSIONS AND FUTURE RESEARCH 92 6.1 Introduction 6.2 Contributions 92 6.2.1 A Modified H.263 Video Codec 92 6.2.2 Correlated Nakagami Fading Simulator 92 6.2.3 Analysis of CDMA Forward Link 93 6.2.4 End-to-end Video Transmission Performance Evaluation 93 6.3 r. Suggestions for Future Research Glossary 92 94 95 Bibliography 101 Appendix A. Derivations of Multiple-Access Interference Variance for Gaussian Approximation 106 Appendix B. Evaluation of the Mean of Attenuation Factor 109 List of Figures Fig. 2.1 H.263 Video Encoder B l o c k Diagram 6 Fig. 2.2 Picture Format and Resolutions for Q C I F in H.263 7 Fig. 2.3 H.263 Coding Hierarchical Structure Fig. 2.4 Simplified H.263 Video Bitstream Syntax 10 Fig. 2.5 Signal Spectra Before and After Spreading 17 Fig. 2.6 A n M-finger R A K E Receiver 20 Fig. 2.7 Forward C D M A Channel 21 Fig. 3; 1 Sample Frames of Miss America Video Sequence (Frame #0 and 90) 31 Fig. 3.2 Video Performance of Modified H.263 Codec 32 Fig. 3.3 Frame Performance of the Modified H.263 Codec 33 Fig. 4.1 Simulated D S - S S C D M A Forward L i n k Video Transceiver System M o d e l 35 Fig. 4.2 Simulated D S - S S C D M A System M o d e l 37 Fig. 4.3 Tapped Delay Line Model of Frequency-Selective Fading Channel 39 '. 9 Fig. 4.4 Example of a Phasor Summation of Partial Waves 41 Fig. 4.5 Comparison of Nakagami Fading Simulator and P D F for m = 1 43 Fig. 4.6 Comparison of Nakagami Fading Simulator and P D F at for m = 3 43 Fig. 4.7 Coherent R A K E Receiver M o d e l 45 Fig. 4.8 Single-Cell C D M A Forward L i n k B E R Performance 56 Fig. 4.9 Simulation Results Comparison of Concatenated and Non-Concatenated Schemes Fig. 4.10 Fig. 4.11 : Video Performance of Modified H.263 Codec in Single-Cell C D M A System P S N R Performance with Various Number of R A K E Fingers vii 58 61 62 Fig. 4.12 P S N R Performance with Different Fading Figures 63 Fig. 4.13 P S N R Performance with Different Logarithmic M I P Exponents 64 Fig. 4.14 P S N R Performance with Different M o b i l e Velocities F i g . 4.15 P S N R Performance with Non-Identical Multipath Fading for Each Path 67 F i g . 4.16 P S N R Performance with Imperfect Channel Estimation 68 F i g . 5-1 Multiple-Cell Configuration M o d e l F i g . 5.2 Cell Geometry M o d e l 80 Fig. 5.3 Analysis vs. Simulation for Multi-cell System 83 Fig. 5.4 P S N R Performance with Non-Identical m Values for Each Multipath Fig. 5.5 P S N R Performance with Non-Identical m Value Set for Channels from. ...66 .....72 „..86 Different Tiers 87 Fig. 5.6 P S N R Performance with Different Surrounding-Cell User L o a d 89 Fig. 5.7 P S N R Performance with Non-Identical M I P for Channels from Different Tiers..90 Fig. 5.8 P S N R Performance with Imperfect Channel Estimation 91 Fig. 5.9 P S N R Performance with Non-Identical Number of Multipaths for Different Tiers 92 viii Acknowledgments This thesis is dedicated to my family, whom I am forever indebted to, for their love and support. From the bottom of my heart, I thank my supervisor, Dr. P. Takis Mathiopoulos, who has always been so supportive and helpful, for his guidance, encouragement, and financial support throughout my thesis work. I really appreciate his friendly personality and his confidence in me, which make everything so much easier during the course of work. I would also like to thank everyone in the Communications Group for their help and friendship, as they have tremendously enriched my campus life at U B C . Finally, I would like to express my gratitude to the financial support provided by the University Graduate Fellowship and the N S E R C scholarship. ix Chapter! INTRODUCTION In recent years, Code Division Multiple Access ( C D M A ) systems have become a viable alternative to both Frequency D i v i s i o n Multiple Access ( F D M A ) and Time D i v i s i o n M u l t i p l e Access ( T D M A ) schemes for mobile cellular telecommunication systems. The use of spread spectrum techniques in wireless communications applications has been a very active area topic for research and development [1], [2], [3]. The C D M A " M o b i l e Station-Base Station Compatibility Standard for D u a l M o d e Wideband Spread Spectrum Cellular System," was issued asIS-95 in December 1992 [9]. It provides a common framework for wireless service providers to develop new compatible C D M A networks. Since its release, C D M A has become a popular choice for new cellular systems around the world. Direct Sequence-Spread Spectrum (DS-SS) technique used in the IS95 systems allows sharing of the same frequency band for all users within each cell and among multiple cells, thus greatly simplifies frequency planning w i t h i n a market. In addition, the inherent wide bandwidth nature of spread spectrum waveforms provides effective fading mitigation and interference combating ability. Due to these distinct advantages of spread spectrum, and the fact that majority of the third generation cellular system proposal adopt spread spectrum techniques [4], this technology seems to be a very promising area worth investigating. A s wireless communications become more popular, information exchange is no longer confined to basic voice transmission. In the last few years, the demand for wireless multimedia services such as the transmission of text data, voice, still images, and video has grown rapidly [10]. A m o n g those services, wireless digital video transmission has probably the most profound impact on the development of future wireless telecommunication applications [11]. A t the same time, it also brings along with it many challenges for systems design. To transmit video through 1 Chapter 1 INTRODUCTION 2 limited bandwidth of radio channels, the first problem is to compress the huge amount of video data to a manageable size that the wireless networks can handle. A m o n g many video compression schemes, I T U - T H.263 low bit-rate video coding standard [12], which targets for videophone and videoconferencing applications, provides a very promising solution. Its low-bit coding scheme aims at compressing video data at rates of no more than 64 kbit/s, low enough for transmission through Public Switched Telephone Networks ( P S T N ) as well as wireless networks. However, there exists a trade-off between compression performance and error sensitivity. The higher the compression rate of the video encoder, the more sensitive channel errors affect the decoded video. Given the possibly severe error-prone nature of the wireless environment, special error control scheme has to be incorporated i n the video codec for satisfactory performance in unreliable mobile channels. In [13]-[21] and [49], there have been a variety of approaches investigated for video transmission over wireless networks. However, only a few of them considered an end-to-end multiple access radio link [13], [18], [19], [20], [21]. A m o n g them, only [13] and [49] uses C D M A channels for video transmission simulations. Most of these papers have dealt with single cell environment but have not taken into account the effects of multiple-cell interference. A l s o , the m o b i l e channel models they employed usually assume R a y l e i g h fading characteristics; however, in the problem of obtaining the distribution of the signal strength for each fading multipath, Nakagami-m distribution gives a more general solution [24] compared to Rayleigh and R i c i a n distributions which provide only special case solutions. E m p i r i c a l data from [22] also suggests that path fading statistics are more adequately described by Nakagami-m fading. In general, Nakagami distribution provides a more general and versatile way to model wireless channels since it can model a greater and wider variety of fading environments. Chapter! INTRODUCTION "' 3 Motivated by the above, in this thesis, we are addressing the problem of efficient transmission o f video data through IS-95 based C D M A systems over correlated N a k a g a m i fading channels. In particular, the contributions of this thesis can be summarized as follows: • A modified version of the H.263 video codec is proposed for improved transmission performance in error-prone mobile environment. • A correlated Nakagami fading simulator is implemented in software for more general and versatile fading channel modeling. • B E R analysis for the D S - S S C D M A forward link in both single-cell and multiple-cell environment are performed. Numerical results are evaluated and compared with those obtained from the Monte Carlo simulations. • End-to-end video transmission through a IS-95 based C D M A system over correlated Nakagami fading channels is simulated. Transmission performance results of the modified H.263-encoded video data in both single-cell and multiple-cell environment are evaluated under a wide variety of system conditions. The following is the organization of this thesis. In Chapter 2, background material for this thesis is outlined. I T U - T H.263 video coding scheme is summarized and its coding syntax related to the thesis is elaborated. Then, mobile channel characteristics including large-scale fading and different kinds of small-scale fading are described with emphasis in Nakagami-m fading and its relationships with conventional Rayleigh and Rician fading models. Next, the basic concept of spread spectrum and its inherent advantages in C D M A systems are described. Lastly, the IS-95 C D M A forward link and its major components are presented. In Chapter 3, introduction of the proposed modified H.263 video codec is presented. This chapter includes.a detailed description of the structure of the proposed codec as well as various Chapter 1'INTRODUCTION evaluation results under additive white Gaussian noise ( A W G N ) channel condition. In Chapter 4, a D S - S S C D M A system operating in a single-cell environment is investigated. We first present both analytical and computer simulated C D M A systems, including the channel model used for the Nakagami fading. Then, the analysis of the bit error rate ( B E R ) performance of the forward l i n k is presented. N u m e r i c a l results o f the B E R performance evaluated from both the analysis and the computer simulations are compared. Lastly, we present and discuss the simulation results of the end-to-end video transmission performance utilizing the proposed H.263 codec over a IS-95 based C D M A forward link system. In Chapter 5, the forward link of a D S - S S C D M A system operating in a multiple-cell environment is investigated. We describe the models for the cellular configuration, the multiplecell radio channels as w e l l as the C D M A forward l i n k system. Then, the B E R performance analysis is presented and its numerical results are compared with those obtained from M o n t e Carlo simulations. Numerical results of the video transmission performance over the multiple-cell C D M A forward link in correlated Nakagami fading channels are evaluated through computer simulations. Finally, in Chapter 6, we present our conclusions and suggest potential topics for future studies. 4 Chapter 2 B A C K G R O U N D 2.1 Introduction The purpose of this chapter is to provide for the background knowledge o f the compression scheme, channel modeling, spread, spectrum systems and standard relevant to this thesis. In Section 2.2, we describe the I T U H.263 low bit-rate video compression scheme by covering its encoding mechanism, video format, hierarchical structure, and bitstream syntax. It is followed by the discussion of mobile radio propagation channels in Section 2.3 and 2.4, where large-scale and small-scale fading models are presented. In Section 2.5, an overview of spread spectrum modulation techniques is provided. We then explain the diversity mechanism of R A K E receiver i n Section 2.6. We conclude the chapter by describing the IS-95 C D M A system for the forward link and discussing the characteristics of the spreading codes the standard employs. 2.2 ITU-T H.263 Low Bit-Rate Video Compression In the past few years, with the emerge of multimedia services, there has been great demand for digital video communications. Thus, the I T U - T (formerly C C I T T ) H.263 standard [12] for low bit-rate video compression scheme was proposed for videophone and videoconferencing applications. The targeted bit-rate, for less than 64 kbit/s, is relatively low compared to other standardized video compression schemes such as M P E G - 2 [54], of which bit-rates are i n the range of M b i t s / s . It is due to this low bit-rate compression that makes H.263 video encoding scheme suitable for wireless mobile applications [21], in which bandwidth constraint imposes a major design problem. In order to achieve the high compression ratio, H.263 video coder incorporates a combined effort of two dimensional (8 x 8) Discrete Cosine Transform ( D C T ) and m o t i o n - 5 Chapter 2 BACKGROUND 6 compensation prediction to compress moving images in spatial and temporal domains, respectively. The block diagram of the H.263 video encoder is shown in F i g . 2.1 [12]. The C o d i n g INTER/INTRA decision flag "Transmitted or not" flag Coding Control Quantizer Indication Video In DCT 0 — ^ Quantizer Index for VLC Quantizer Inverse Quantizer Inverse DCT P , Picture Memory Motion Vector J Fig. 2.1 H.263 Video Encoder B l o c k Diagram [12], Control module determines the I N T E R / I N T R A decision flag. It is used to indicate i f the encoder is in inter or intra mode. The encoder is said to be in intra mode i f it operates D C T directly on the input image without the use of motion compensation. The resulting frame is called an I N T R A or I-frame. Otherwise, the encoder is said to be in inter mode and the resulting frame is called a Prediction or P-frame. Coding Control also determines the value of "Transmitted or not" flag. The flag is set to 1 i f coded data is transmitted, or 0 i f none is transmitted in which case data from previous frame is reused. The Quantizer Indication from Coding Control provides quantization step size information. Chapter 2 BACKGROUND 7 Each video image is divided into blocks of 8 x 8 pixels before being sent to the encoder input as Video In [12]. When the encoder is in inter mode, prediction generated by the previous frame is subtracted from each block. The difference between the block being encoded and the prediction is sent for transform coding by the D C T block. On the other hand, when the encoder is in intra mode, the block is directly sent for D C T coding. In either case, the resulting transform coefficients are uniformly quantized by the Quantizer block. The resulting Quantization Index are then encoded using Variable Length C o d i n g ( V L C ) such as Huffman codes. The quantized transform coefficients are also sent to Inverse Quantizer and Inverse D C T blocks. They are used to generate prediction frame which is stored in Picture Memory block for subsequent I N T R A frame encoding and to generate the Motion Vector ( M V ) from motion-compensated prediction. F i g . 2.2 shows the picture format and resolutions for Quarter C o m m o n Intermediate Format (QCIF) of video sequence used in the H.263 compression scheme [54]. QCTF is a univer- Luminance Y Chrominance Cb Chrominance Cr 88 90 pixels I] I 90 pixels 180 pixels F i g . 2.2 Picture Format and Resolutions for QCTF in H.263 [54]. sal video format designed to accommodate different kinds of existing television formats such as P A L , S E C A M and N T S C . It consists of one luminance component Y and two chrominance components C b and C . Since the basic block size of H.263 is 8 x 8 pixels, the pixels outside the r dotted lines in F i g . 2.2 is cropped out, and the resulting area is called the Significant Pel A r e a Chapter 2 BACKGROUND 8 (SPA). Thus, the Y component has a resolution of 176 pixels x 144 lines, whereas C b have a resolution of 88 pixels x 72 lines. The YC C b r and C r both color coordinate adopted by H.263 standard was developed as part of I Y U - R BT.601 [25] during the establishment of a worldwide digital video component standard. This yields a compatible digital approach between the two different systems namely, the 525-line N T S C and 625-line P A L / S E C A M . 2.2.1 Encoded Video Bitstream Structure The H.263 standard defines a consistent structure so that the decoder may decode the received bitstream without any ambiguity. The H.263 syntax for the coded video bitstream has a hierarchical representation (Fig. 2.3) with four data layers, namely the Picture Layer, the Group of B l o c k (GOB) Layer, the Macro B l o c k ( M B ) Layer and the B l o c k ( 8 x 8 pixels) Layer [12]. Each layer is composed of data and the corresponding header information. In Fig. 2.3, the Picture Layer divides a Q C I F picture frame into 9 Group of Blocks denoted as G O B 0 to G O B 8 [12]. In the Group of Blocks Layer, each of the G O B s in a picture frame is s u b - d i v i d e d into rows of 11 M a c r o b l o c k s . E a c h M B i n turns consists o f 1 l u m i n a n c e (Y) component and 2 chrominance (C b and C ) r components in the M a c r o b l o c k L a y e r . The Y component is made up of 4 basic blocks ( 8 x 8 pixels), where the C b and C components are each r made up of 1 basic block, i.e., half the spatial resolutions of luminance. Finally, the B l o c k Layer consists of basic blocks which are 8 x 8 pixels in dimension. 2.2.2 Bitstream Syntax F i g . 2.4 shows the simplified H.263 video bitstream syntax of each hierarchical layer of the coding [21]. It consists of the Picture Layer, the Group of B l o c k s Layer, the M a c r o b l o c k Layer Chapter 2 BACKGROUND 9 176 p i x e l s G O B O G O B 1 G O B 2 176 p i x e l s G O B 8 M B O MB10 | M B I | " 16 l i n e s Group of Blocks Layer Picture Layer 8 pixels , ' Luminance Component Yl 8 pixels 64 57 Block Layer Fig. 2.3 Y2 Y4 Y3 Chrominance'' s Components Cl s s C2 Macroblock Layer H.263 Coding Hierarchical Structure [54]. and the B l o c k Layer. The Picture Layer begins with the Picture Start Code (PSC), followed by the Picture Header and the data for the G O B layer [12]. It terminates w i t h an optional E n d - O f Sequence code ( E O S ) and stuffing bits ( S T U F ) . The Picture header consists of Temporal Reference (TR), Picture Type information ( P T Y P E ) , Picture Quantizer information ( P Q U A N T ) , Continuous Presence Multipoint indicator ( C P M ) , Picture Sub-Bitstream Indicator (PSBI), Extra Insertion information (PET), and spare bits ( S P A R E ) . Temporal Reference for B i - d i r e c t i o n a l frames ( T R B ) and quantization information for bi-directional picture ( D B Q U A N T ) are also used in the Picture layer i f optional PB-frames mode is used. A G O B Layer consists of a G O B header followed by data for the Macroblock layer [12]. For the G O B that starts at the beginning of Macroblock row number 0, the G O B header is not Chapter 2 BACKGROUND 10 Picture Layer Picture Header pse »(^GOB Layer^- EOS STUF Group of Blocks (GOB) Layer GOB Header GBSC MB Layer ^ - Macroblock (MB) Layer MB Header ^) Fixed Length Coding L_—j—Block Layer ^ Block Layer Variable Length Coding TCOEF INTRADC Fig. 2.4 ^ C D Simplified H.263 Video Bitstream Syntax [21]. transmitted. G O B header starts with Group of B l o c k Start Code ( G B S C ) , followed by Group N u m b e r ( G N ) , G O B Frame I D ( G F I D ) , G O B S u b - B i t s t f e a m Indicator ( G S B I ) , and G O B quantizer information ( G Q U A N T ) . Each G O B is sub-divided into Macroblocks in the M a c r o b l o c k layer [12]. M a c r o b l o c k header includes Coded Macroblock Indication ( C O D ) , Coded B l o c k Pattern ( C B P ) , Macroblock type and C o d e d B l o c k Pattern for C h r o m i n a n c e ( M C B P C ) , C o d e d B l o c k Pattern for B i directional blocks ( C B P B ) , Coded B l o c k Pattern for Luminance ( C B P Y ) , M o t i o n Vector Data ( M V D ) , M o t i o n Vector Data for optional advanced prediction mode ( M V D _ ) , and M o t i o n 2 4 Vector Data for Bi-directional macroblocks ( M V D B ) . Macroblock mode for Bi-directional blocks Chapter 2 BACKGROUND 11 ( M O D B ) indicates whether C B P B and/or M V D B are transmitted for a Macroblock. Quantizer for the Macroblock is specified in D Q U A N T . Lastly, for the B l o c k Layer, a block is composed of 64 (8 x 8) DCT-coded residual data. D C coefficient for I N T R A blocks ( I N T R A D C ) is present for every prediction block that is intracoded. This code is followed by variable length coding representation of Transform Coefficients (TCOEF). 2.3 Mobile Radio Propagation: Large-Scale Fading In this section, we describe the large-scale fading model for the mobile radio channels. Largescale propagation models predicts the average received signal strength for an arbitrary transmitterreceiver separation distance in usually several hundreds or thousands of meters [5]. Average signal strength attenuation in large-scale fading is mainly due to two factors, namely propagation loss and shadowing. B o t h theoretical and measurement-based propagation models indicate that average received signal power decreases logarithmically with distance, for both outdoor or indoor radio channels [5]. The attenuation is proportional to d , where d is the distance between two points and y is the path loss exponent. In addition to path loss in a mobile radio link, there also exists the phenomenon of shadowing. Shadowing comes from the diffraction effect on signals due to terrains and obstructive objects that are much larger than the wavelength. It is usually modeled as a random variable with a log-normal distribution [7], [53], [55]. W i t h the overall effect of path loss and shadowing, the signal power P R proportional to the transmitted signal P T described as follows [7]: received is Chapter 2 BACKGROUND 12 10 (2.1) PR~PT V J where \\ is a Gaussian random variable with zero mean and standard deviation a in d B , d is the distance and y is the path loss exponent. Lee [7] suggests that standard deviation rj of 8 dB and path loss exponent y of 4 be used for urban macrocellular environment. 2.4 Mobile Radio Propagation: Small-Scale Fading In this section, we describe the small-scale fading model for the mobile radio channels. S m a l l scale fading model characterizes the rapid fluctuation of the received signal strength over very short travel distance (a few wavelengths) or short time duration (on the order of seconds) [5]. It is caused by interference between two or more versions of the transmitted signal arriving at the receiver at slightly different times due to multipath reflections by local scatterers [55]. These multipath waves combine at the receiver antenna to give a resultant signal which can vary widely in amplitude and phase, depending on the distribution of the intensity and relative propagation time of the waves and the bandwidth of the transmitted signal. 2.4.1 Small-Scale Fading: Parameters and Characterization To describe the time dispersive nature of a channel, multipath spread T m is used in the measure- ment. It denotes the difference in time delay of arrivals from different signal paths i n a mobile radio environment due to multipath reflections [5]. For an impulse transmitted at the cell site, by the time this impulse is received at the mobile unit it is no longer an impulse but rather a pulse with a spread width of T . m Related closely to the multipath spread is the coherence bandwidth Chapter 2 BACKGROUND B. co 13 It represents the range of frequencies over which two frequency components have a strong potential for amplitude correlation [5]. Coherence bandwidth is defined as the reciprocal of multipath spread T . m Another parameter related to the time spreading of a multipath signal is the multipath intensity profile (MIP). It is the average power at the output of the channel as a function of path delays [26]. Different distributions of M I P such as exponential [38], Gaussian [39] and uniform [40] have been used in the literature. To describe the time varying nature of the channel in small-scale fading, Doppler spread B d is being used. It denotes the amount of frequency spectrum broadening caused by the time rate of change of the mobile radio channel [26]. D o p p l e r spread B d is a function of the relative velocity, of the mobile, and the angle a between the direction of receiver and the direction of arrival of the scattered waves. The dual of Doppler spread in time domain is called the coherence time T . co It designates the time duration over which the channel impulse response is essentially invariant [5]. It is inversely proportional to the Doppler spread B. d W h e n an information-bearing signal is transmitted through the fading channel, i f the coherence bandwidth is small in comparison to the bandwidth of the transmitted signal, the channel is said to be frequency-selective [5]. O n the other hand, i f the coherence bandwidth is large i n c o m p a r i s o n to the bandwidth of the transmitted s i g n a l , the channel is said to be frequency-nonselective or simply flat. For relatively high Doppler spread or short coherence time compared to a symbol duration (e.g. T co is less than 10% of T), the channel is categorized as fast fading [5]. O n the other hand, for relatively low Doppler spread (e.g. B d less than 10% of the original spectrum) or long coherence time compared to a symbol duration, the channel is categorized as slow fading. Chapter2 BACKGROUND 2.4.2 14 Small-Scale Fading: Distribution Models E m p i r i c a l results as well as physical reasoning show that the total intensity of each resolvable path is formed by the superposition of radio waves that arrive at the receiver almost simultaneously [22], [24]. Thus, the problem of obtaining the distribution of the signal strength of each multipath coincides with the random phasor problem [29]. Although Rayleigh and Rician distributions are often used in literature for modeling small-scale fading [13], [14], [18]-[21], [27], [42], [43], they only provide the special case solutions of the random phasor problem. In contrast, the Nakagami-m distribution provides a more general solution [35], [36], [37]. Moreover, it is shown in [22] that Nakagami-m distribution provides a better fit for the empirical fading statistics. D u e to the above advantages, in this thesis we have adopted the Nakagami-m distribution to model for the small-scale fading. In the following sub-sections, we w i l l first describe the conventional Rayleigh and Rician fading models. Then Nakagami model w i l l be introduced and its relationship with Rayleigh and Rician distributions w i l l be explained. 2.4.2.1 Rayleigh Fading Model When there is an absence of line-of-sight wave between the transmitter and receiver, addition of pure scattered waves causes the resultant wave amplitude exhibiting a Rayleigh-distributed behavior. It is w e l l k n o w n that the envelope of a complex Gaussian noise signal f o l l o w s a Rayleigh distribution. The probability density function (PDF) of the Rayleigh distribution is [26] (2.2) a where G a is the root mean squared.(RMS) value of the received voltage signal, and o% is the Chapter! BACKGROUND 15 time-average power of the received signal. The P D F of the phase of a R a y l e i g h fading signal follows a uniform distribution around [0, 2n). 2.4.2.2 Rician Fading Model In Rician fading, random multipath components arriving at different angles are superimposed on a stationary dominant signal. A s the dominant signal becomes weaker, the composite signal resembles a noise signal which has an envelope that is Rayleigh. The Rician distribution is given as [26] r +A 2 p(r) where = —.e 2 * '7 te\ A>0,r>0 2o 0 (2.3) is the average power of the fading signal, A is the peak amplitude of the dominant signal, and is the modified Bessel function of the first kind and zero-order. The Rician distri- bution is often described in terms of a parameter K which is defined by the deterministic signal power and the variance of the multipath signal. It is given by [36] K = —2 - a (2-4) 7 The parameter K is known as the Rician factor and completely specifies the Rician distribution. A s A —> oo, K —> » , and as the dominant path decreases in amplitude, the Rician distribution degenerates to a Rayleigh distribution. The phase of a Rician fading signal has a P D F of [58] e~ K p(Q) VKcos0exp(-Ksin 0) = —. + ^ n 2 2 ^ 2JK r ~ „, / >- • [ 2 - e r f c ( V K c o s © ) ] . c . (2.5) Chapter 2 BACKGROUND 2.4.2.3 16 Nakagami-m Distribution The P D F of Nakagami-m distribution was deduced by Nakagami from his large scale experiments on rapid fading in high frequency long distance propagation and has the form of [35] / \ p{r) = _ m 2m- 1 2m r -{m/Q)r T(m)Q. , —e ,~ s-\ . (2.6) -oo In (2.6), T ( m ) denotes the G a m m a function, defined as T(m) = t ~ e~ dt for m > 0 . The m Jo x l parameter Q is the second moment of r, or the mean power, defined as Q. = E[r ], where 2 E[] is the average function. The parameter m is the fading figure characterizing the severity of the fading, and is equal to the inverse of the normalized variance of r E [r ] 2 m = E[(r -E ^1 2 — 2 T [r ]) ] 2 >- . 2 as [35] (2.7) 2 A s for special cases, m = 1 corresponds to purely diffusive scattering or R a y l e i g h fading, m —> oo corresponds to the non-fading condition [23]. The Rician distribution can also be closely approximated by the Nakagami-m distribution. The approximation can be obtained through functional relationship of the parameters of the two distributions K and m [36] ' 1- K T+ X K A s shown by (2.8), the relationship between m and K is not strictly linear. However, an accurate linear approximation can be obtained for K > 2 as [36] m = sK m where s ~ 0.5 and m 0 +m Q (2.9) = 0.7622 for K ~ {2 - 100} . F r o m (2.9), it is observed that m is a Chapter 2 BACKGROUND 17 similar parameter as K in terms of physical interpretation, i.e. it is the amount of specular power in the received signal with respect to scattered power. 2.5 Spread Spectrum Modulation Techniques In this section, we explain the basic concept of spread spectrum modulation techniques. Spread spectrum techniques transform a signal with bandwidth B into a noise-like signal of much larger bandwidth B , ss which usually is of several orders of magnitude greater than the original signal bandwidth [3], as illustrated in F i g . 2.5. The amount o f spreading is measured as the ratio Power Spectral Density , ss Fig. 2.5 . N = B /B ss Signal Spectra Before and After Spreading [3]. called the processing gain. A s a result, the power of the radiated spread spectrum signal is spread over a much broader bandwidth, while its power spectral density is correspondingly reduced by the same proportion. Each spread signal is pseudorandom and has noise-like properties when compared with the digital information data [5]. The spreading waveform is controlled by a pseudo-noise (PN) Chapter 2 BACKGROUND 18 sequence, which is a binary sequence that appears random but can be reproduced in a deterministic manner by the intended receiver. Spread spectrum signals are demodulated at the receiver through cross-correlation with a locally-generated version of the pseudorandom codes. Crosscorrelation with the correct P N sequence despread the desired spread spectrum signal and restores the modulated message in the same narrowband as the original data, whereas cross-correlating the signal from other users result in small amount of wideband noise at the receiver output. . There are many distinct advantages of spread spectrum waveforms. The most important one is its inherent interference mitigation capability [5]. Since each user is assigned with a unique P N sequence w h i c h has very little correlation with others, the receiver can distinguish the intended user from the rest based on their spreading codes, even though they all share the same frequency spectrum and bandwidth. Moreover, since narrowband interference affects only a small portion of the spread spectrum signal, their effects are negligible for spread spectrum signal. Furthermore, a direct sequence C D M A cellular system can apply a universal one-cell frequency reuse pattern [1]. Hence, there is no need for complicated frequency planning or frequency guardband assignment as in T D M A or F D M A schemes. A s a result, it greatly simplifies radio resource management and minimizes wastage. Wide bandwidth characteristics of spread spectrum signal also provides effective mitigation of multipath fading by introducing multipath diversity [1]. Since its bandwidth is usually much wider than the channel coherent bandwidth, the channel is categorized as frequency•selective [5]. A t any given time, only a small portion of the wideband signal w i l l undergo fading, whereas the rest of it remains high in signal-to-noise ratio facilitating correct signal detection. Besides this resistant nature to multipath fading, spread spectrum systems can take advantage of the frequency-selective characteristics of the signal by resolving the multipath components using Chapter 2 BACKGROUND . a R A K E receiver for improved performance [ 6 ] , the structure of which w i l l be presented in the next section. 2.6 The R A K E Receiver In this section, we describe the structure of a R A K E receiver used in C D M A systems. In a multipath frequency-selective fading channel, conventional modulation techniques require equalizer to cancel intersymbol interference between adjacent symbols [5]. For spread spectrum modulation used in C D M A systems, however, the situation is quite different. For direct sequence spread spectrum signals, the spreading codes are designed to have very low correlation between successive chips. Thus, propagation delay spread in the radio channel merely provides uncorrec t e d multiple versions of the transmitted signal at the receiver, on condition that these multipath components have time separations of more than one chip duration. In a C D M A system, it is not only that no equalization is required, but the receiver can actually gain performance from the multipath characteristics of the channel by c o m b i n i n g those multipath signal components constructively. S u c h a C D M A receiver w h i c h uses multipath diversity is referred to as the R A K E receiver, and was first proposed by Price and Green [ 6 ] . A R A K E receiver consists of a bank of correlators called fingers, each of which correlates to a particular multipath component of the desired signal. The correlator outputs are weighted according to their relative signal strength to achieve maximal ratio combining. For a C D M A system, the R A K E receiver acts essentially as a diversity receiver, where the diversity is provided by the fact that the multipath components are practically uncorrelated from one another when their relative propagation delays exceed one chip duration. If the output from one correlator is corrupted by fading, the others may not be, and the ^ Chapter 2 BACKGROUND 20 corrupted signal may be discounted through the weighting process. The adverse effect of fading can thus be greatly reduced by diversity provided by combination of all finger statistics before receiver, output decision is made. F i g . 2.6 illustrates a R A K E receiver structure employing M fingers [5]. In the receiver, each finger of the R A K E correlates to a portion of the received signal r(t) which is delayed by at least one chip in time from the other fingers. The output of the M correlators are denoted as Zj, Z , 2 Z. M They are weighted by c i j , oc , oc 2 M correspondingly, whose coefficients are directly proportional to the instantaneous power from each correlator output. The weighted output of the M correlators are then summed up as Z ' and integrated to produce the decision variable Z before being sent to the decision device for symbol detection. Correlator 1 r(t) baseband CDMA multipath signal 1 •( *( a. f (-)dt Correlator 2 Jo +1 £0 -1 Symbol Detection Decision Device Correlator M Fig. 2.6 aM A n M-finger R A K E Receiver [5]. 2.7 IS-95 Code Division Multiple Access System The communication system model employed in this thesis is based on the Interim Standard 95 (IS-95) " M o b i l e Station-Base Station Compatibility Standard for Dual M o d e Wideband Spread Spectrum Cellular System" originally issued by the U . S . Electronics Industries Association ( E I A ) Chapter 2 BACKGROUND 21 [9]. The following sub-sections describe the relevant topics of the standard for this thesis. 2.7.1 CDMA Forward Link Fig. 2.7 shows the forward (down-link) traffic channel modulation of a IS-95 C D M A system [9]. I-Channel P N Sequence Walsh Code U s e r data Data Scrambling Convolutional Encoder r=l/2 K=9 Power ControlBit Baseband Filter Block Interleaver Baseband Filter Long Code for nth user Long Code Generator Q-Channel P N Sequence Fig. 2.7 Forward C D M A Channel [9]. The information transmitted is c o n v o l u t i o n a l l y encoded to provide the c a p a b i l i t y of error detection and correction at the receiver. The code used has a constraint length of nine, K = 9 , and a code rate of one-half, r = 1 / 2 . The output of the convolutional coder is then b l o c k interleaved using a 24 x 16 block interleaver over a 384-symbol interval to reduce bursty errors resulting from mobile radio channels. The data symbols are then scrambled by a user-specific long code with a period of 2 4 2 - 1 chips. Data scrambling serves the purpose of user identifica- tion and security [55]. To minimize the average B E R for each user, IS-95 strives to force each user to provide the same power level at the base station [56]. Power control commands are sent to each subscriber unit through the Power C o n t r o l B i t . F o l l o w i n g data scrambling is the orthogonal covering. E a c h data symbol is orthogonally spread by one of the 64 orthogonal Walsh codes, Chapter 2 BACKGROUND 22 which are binary sequences completely orthogonal to each other [5]. Signals from different users within a cell are distinguished by their Walsh codes. After the orthogonal covering, data symbols go through quadrature spreading where a pair of short binary P N sequences (the basic codes) with a period of 2 1 5 -1 chips is used in the in-phase and quadrature branches. A l l base stations share a common quadrature pair of P N codes, but each are assigned a unique time offsets value [56]. Signals from different cells are distinguished by the time offsets from the basic codes. Following quadrature spreading, each of the two branches is filtered separately with a finite impulse response (FIR) filter prior to carrier modulation. The forward link consists of a pilot channel, a synchronization channel, up to seven paging channels, and up to sixty-three forward traffic channels [55]. A m o n g them, the pilot channel is one of the most important aspects of the forward link signal design. Each cell-site transmits a pilot tone which is used as a coherent carrier reference for demodulation by all mobile receivers [56]. The pilot channel signal is transmitted at a relatively higher power level than the other channels (e.g. synchronization channel, paging channels and traffic channels) so that extremely accurate tracking can be achieved. Specifically, it is simply a constant-level signal that is modulo-2 added with the all-zeros Walsh code and sent over the air after quadrature spreading. The mobile station synchronizes with the nearest base station by detecting the pilot tone with the strongest signal level. It then determines the identity of the base station by the time offset from the basic codes, a unique value for every cell in the system. After synchronization, the pilot signal is used as a coherent carrier phase reference for demodulation of the other signals from this base station. Remaining synchronization details and other system information is sent to the mobile station through the synchronization channel. Once the synchronization channel has been received, the mobile station uses one of the paging channels to receive other system information and paging 23 Chapter 2 BACKGROUND commands. 2.7.2 Direct Sequence Spreading Codes The ideal spreading code would be an infinite sequence of equally likely random binary digits [57]. Unfortunately, in order to despread the signal, we need to have the same version of the spreading code at both the transmitter and receiver. It implies the need for infinite storage at both ends, which is clearly impossible. Alternatively, periodic pseudo-noise (PN) codes are used in spread spectrum systems. These are codes that can be easily generated by means of shift registers and have noise-like behavior. IS-95 C D M A systems use maximal-length codes (m-sequences) for part of the spectrum spreading (see F i g . 2.7). For a multiple-access system, it is important to minimize interference among signals from different users. Therefore, in IS-95 C D M A spectrum spreading scheme, a class of orthogonal codes called the Walsh codes is also used to provide nearly perfect isolation between the multiple-user signals transmitted by a base station [56]. We w i l l discuss each of these codes in the following sub-sections. Their unique properties, either desired or undesired for a C D M A system, w i l l be explored. We w i l l also describe how the IS-95 standard employs the method of code concatenation to combine the best of both codes. 2.7.2.1 Maximal-Length Sequences The P N sequences used in IS-95 are m-sequences generated by 15 linear shift registers based on the characteristic polynomials given in [9] with a period of 32767 chips: pJ) = x P (x) = x x Q 1 5 15 +x 1 3 +x +x +x +x +1 +x n +x 9 n 8 +x 7 10 5 +x +x +x +x +1 6 5 A (2.10) 3 Periodic autocorrelation function is an important characteristic of a periodic P N sequence, and it Chapter 2 BACKGROUND 24 is denned as [26] n ^(2b -l)(2b Hj)= i i + j -l), 0<j<n-l (2.11) i= 1 where n is the period and b is the ith binary bit o f the sequence. It w o u l d be ideal to have i sequences that have periodic autocorrelation values of (j)(0) = n and - 0 for 1 < j < n .— 1 . D u r i n g code despreading, it w o u l d provide m a x i m u m correlation value for matched code detection while rejecting any other offset sequences by the zero correlation value. In case of msequences, the periodic autocorrelation function is [26] (2.12) A s the period n gets larger, the off-peak values become insignificant compared to the peak value. Thus, m-sequences are almost ideal in terms of their autocorrelation function. However, msequences are not practical to be used alone for C D M A because the periodic cross-correlation function between any pair of the m-sequences of the same period can have relatively large peaks [34]. This is clearly undesirable in C D M A systems since it w i l l increase interference which in turn reduce capacity. Although it is possible to select a small subset of m-sequences that have relatively smaller cross-correlation peak values, the number of such sequences in the set is usually too small for C D M A applications [26]. 2.7.2.2 W a l s h Sequences Walsh codes provide orthogonal covering for IS-95 C D M A system [5]. These Walsh functions comprise o f binary sequences generated from a dimension-64 Hadamard matrix. E a c h W a l s h sequence is strictly orthogonal to one another. D u r i n g spreading, each information symbol is Chapter 2 BACKGROUND 25 replaced by one of the 64-bit long Walsh codes which is uniquely assigned to each user within a cell. In the absence of multipath, the Walsh codes provide perfect orthogonal channelization for users on the forward link within the cell. However, the orthogonality breaks down in a multipath mobile environment where multipath delays introduce inhomogenous auto-correlation and crosscorrelation characteristics [33]. If Walsh sequences are used alone, the value of auto-correlation and cross-correlation of orthogonal sequences can be very high for certain time delays. Moreover, not all of the Walsh sequences has wideband spectral characteristics as desired in spread spectrum [5]. These undesirable effects can be mitigated by code concatenation with a P N sequence as is done in IS-95 C D M A systems. 2.7.2.3 Concatenated W a l s h / P N Sequence IS-95 C D M A systems use code concatenation scheme with Walsh functions and m-sequences, which combines the desirable properties from both codes [9]. Each of the 64-chip long Walsh sequence is modulo-2 added with a chip from the 32767-chips long m-sequence. Since the length of these two sequences are relatively prime to each other, it allows every c h i p from the msequence to occur at the beginning of some data bit during the concatenation. The resulting concatenated Walsh/ P N codes provide orthogonality between multiple users within the same cell in a single path propagation environment, while reducing the inhomogenous behavior of Walsh cross-correlation due to non-zero time delays in multipath environment [32]. Moreover, concatenation scheme reduces interference between mobiles that use the same Walsh code in different cells. Furthermore, it provides the desired wideband spectral characteristics for direct sequence spread spectrum signals in C D M A systems. We w i l l provide numerical results to Chapter 2 BACKGROUND 26 verify the advantage of using concatenated codes over non-concatenated codes in Chapter 4. 2.8 Conclusions In this chapter, we presented some of the essential background material for the better understanding of this thesis. We described the H.263 video coding scheme. Then, different wireless channel models were e x p l a i n e d . Furthermore, spreading spectrum techniques and R A K E receiver mechanism were illustrated. Lastly, the IS-95 standard and its spreading codes were discussed. Chapter 3 M O D I F I E D H.263 V I D E O C O D E C 3.1 Introduction One inherent problem with any communications system is that information may be altered or lost during transmission due to channel noise, interference and distortion. The effect of such information loss can be devastating for the transport of compressed video because any damage to the compressed bitstream may lead to objectionable v i s u a l distortion at the decoder. Several techniques have been suggested to limit the damage of channel errors for H.263 video compression [13]-[21], [45]-[50]. In general, they fall into three categories namely, Forward Error Correction ( F E C ) [14], [21], [45], [50], Automatic Repeat reQuest ( A R Q ) [16], [17] and a combination of both [13], [18], [19], [20], [46], [49]. For schemes utilizing A R Q , retransmission of corrupted frames introduces additional delay. This makes the use of A R Q for voice oriented applications very limited because a round-trip delay of less than 300 ms is required [17]. Therefore, for realtime applications, F E C approach is more suitable, especially i f round-trip delay is considerably long. In this chapter, we propose a modified H.263 codec which is based upon a combination of F E C coding scheme and periodic frame refresh procedure. The proposed modification is compatible with the H.263 standard to improve video quality for low bit-rate video transmission over error-prone wireless channels. In Section 3.2, we w i l l first investigate the impact o f video bitstream errors on H.263 video quality. Then, in Section 3.3 and 3.4, we w i l l describe the details of the proposed scheme. Lastly, in Section 3.5, numerical results for the video codec performance in a A W G N channel is presented. 27 Chapter 3 MODIFIED H.263 VIDEO CODEC 3.2 Effects of Errors on H.263 Video Generally, the severity of bit errors to video quality depends on the spatial and temporal location of the error/Since H.263 coding scheme uses prediction frame statistical redundancy removal coding techniques, temporal and spatial error propagation problems are inherent. Huffman coding causes spatial error propagation problem, whereas motion compensation causes temporal error propagation problem. Based on some recently reported results [21], [45], [50], the effects of transmission errors on H.263 video can be summarized as follows: • Errors in video headers can cause major damage, especially for headers in higher hierarchical layer of H.263 coding such as the Picture Layer and the Group of Blocks L a y er. • Errors propagate in spatial domain due to improper decoding of variable length codes; The Picture and Group of B l o c k Layer headers stop error propagation in spatial domain by providing start codes. • Errors propagate among Prediction frames in the temporal domain; I N T R A frames stop the propagation by coding the picture frame independently. To m a x i m i z e video quality, while avoiding time delay for real-time applications and m i n i m i z i n g channel coding redundancy and complexity, we propose a selective F E C coding scheme combined with an I N T R A frame forced update mechanism. 3.3 Selective FEC Coding The family of B C H codes are powerful linear block codes for which efficient decoding algorithms exists [57]. This family of codes contains codes of many rates and a wide range of error correcting capability. We adopt B C H codes because of their strong error-correcting power and relatively low 28 Chapter 3 MODIFIED H.263 VIDEO CODEC complexity. A n appropriate block interleaver is needed to transform the bursty error statistics of the mobile channel into Gaussian-like statistics required by the shorter B C H codes. A s mentioned in Section 2.7, such block interleaver is included in the IS-95 forward link standard. Since IS-95 C D M A system uses half-rate convolutional coding on all transmitted data, it provides the base layer F E C protection. Considering the impact of errors on the H.263 video bitstream and in order to m i n i m i z e the redundancy introduced by the c o d i n g , we choose to introduce two extra F E C protection to selected bits of the upper two hierarchical layers of the H.263 bitstream syntax. For the lower two hierarchical layers, because of the lack of start code for synchronization and of the variable nature of the sequence lengths, the positions of their bits are unknown. Thus, in order to protect any selected bits from these two layers, extra information bits have to be introduced to indicate the positions and lengths of those chosen bits. However, these extra information bits, which are as important as the chosen bits we would like to protect, do not have any extra F E C coverage. Therefore, the effort to selectively protect bits in the lower two hierarchical layers may not be worthwhile because it introduces equally error-sensitive information bits during the process. Assuming that a Q C I F sequence (176 x 144 luminance resolution, 4:1:1 chrominance subsampling) is transmitted, F E C I provides error protection for the important header bits in Picture layer of both I N T R A and Prediction frame which include P S C , G N , T R , P T Y P E , P Q U A N T , C P M , P E L They add up to a total of 50 bits and a (71, 50, 3) B C H codeword is assigned. 21 bits of redundancy are introduced for correcting up to three errors. F E C 2 provides error protection for header bits in G O B layer w h i c h includes G B S C , G N , G F I D , G Q U A N T . These 29 bits are protected by a (41, 29, 2) B C H codeword. 12 bits of redundancy are introduced for correcting up to two errors. 29 Chapter 3 MODIFIED H.263 VIDEO CODEC 30 3.4 Forced INTRA Frame Update The major compression achieved by low bit rate encoders such as H.263 is mostly due to the removal of temporal redundancy through motion compensation. Therefore, it is necessary for the encoder to emulate the decoder at the transmitter. If the information available to this decoder and the decoder used at the receiver are not the same, the quality of the reconstructed picture-frame can degrade considerably. Moreover, temporal mitigation of these reconstruction errors can affect the quality of the subsequently reconstructed picture frames. If the reconstructed signal is degraded due to some channel errors, subsequent reconstruction of error-free transmissions may also be incorrect. Therefore, it is apparent that to ensure high quality of transmission, effects of errors must not propagate too far beyond the temporal interval of the channel, errors. One solution to this is to apply forced update of the prediction frames b y I N T R A frames periodically. Considering the trade-off between bit usage and overall video quality, we have chosen to apply I N T R A frame update in every 10 prediction frames. 3.5 Numerical Results for the AWGN Channel The implementation of the modified H.263 video codec was based on the Test M o d e l Near-term ( T M N ) version 2.0 software platform developed'by Telenor R & D , N o r w a y . We consider the transmission of 100 Q C I F video frames of the well known sequence Miss America (Fig. 3.1) over A W G N channels at a nominal rate of 64 kbits/s. The overhead introduced by F E C coding is about 1.1%. Results with average bit error rates of 1 0 - 3 to 1 0 - 4 are considered. The average Peak Signal-to-Noise Ratio ( P S N R ) of weighted luminance and chrominance components is used as a measure of objective quality, and is given by Chapter 3 MODIFIED H.263 VIDEO CODEC Fig. 3.1 Sample Frames of Miss America Video Sequence (Frame #0 and 90) PSNR = (PSNR) Y + 0.3(PSNR) + 0.3(PSNR) Ch Cr (3.1) where P S N R of each component is calculated as M (PS*K), = I O . o g l £ - | * i=l v « (3.2) ' ' In (3.2), A / is the number of video frames, of and rf are the amplitudes of the original and reconstructed luminance (Y) or chrominance (C , C ) component values, respectively. The b r conventional method of calculating P S N R is to take luminance components into account only [19], [20]. For monochrome video, it is probably a sufficient measure to calculate the objective video quality. However, for present and future video communication applications, full color video is more likely to be transmitted over the networks. Since both brightness and colors of the video frames affect human's subjective judgement of the video quality, it is thus logical to take a weighted average of both luminance and chrominance components o f the video frames into account when calculating the P S N R for a better measurement of the overall video quality. Notice that every data point is generated by taking the ensemble average of 20 simulation runs using different random seeds. A s shown in F i g . 3.2, the performance of the modified H.263 video codec is superior to Chapter 3 MODIFIED H.263 VIDEO CODEC 32 P S N R vs Pe @ 64 kb/s Modified H.263 Codec X H.263 Codec —e — •\ 9\ \\ \ \ \^ \ : v \ •• \-\ \\ •: \ v s \ : \ V x Sc. ^ " 0 0.5 -j - J I I I 1 L 1 1.5 2 2.5 3 3.5 ^ 4 P e Fig. 3.2 4.5 5 X10" 3 Video Performance of Modified H.263 Codec that of the original version. Improvement in P S N R ranges from 3 dB to 10 dB depending upon the values of error probability P . The gain in performance becomes smaller for lower P e e because the loss in video quality due to F E C redundancy is greater than the gain from protected bits being hit by errors. In the other end of higher P e values it is observed that the gain in performance from the proposed scheme becomes smaller again. It is because in severe error condition, the video performance degrades to a point where F E C can provide very little help to recover the damage. In F i g . 3.3, we show the frame-to-frame performance o f both c o d i n g schemes at P e = 5 x 10 - 3 compared to error-free condition. A g a i n , it can be observed that the modified Chapter 3 MODIFIED H.263 VIDEO CODEC 33 coding scheme offers better performance in most of the frames, thus resulting in better overall P S N R value. The step-like shape of the modified H.263 codec curve (solid line) is because of the forced I N T R A frame refresh mechanism occurs in every ten frames. 90 Modified Codec Original Codec 80 No Error 20 10 10 20 30 40 50 60 70 80 90 100 Frame Number F i g . 3.3 Frame Performance of the Modified H.263 Codec 3.6 Conclusions In this chapter, we discussed the impact of errors on unprotected H.263 bitstream. Then, we proposed a modified encoding scheme which uses selective F E C protection for the higher hierarchical layer bits and forced I N T R A frame refresh mechanism to improve performance in error- Chapter 3 MODIFIED H.263 VIDEO CODEC prone wireless environment. We introduced a weighted luminance and chrominance P S N R measurement to better quantify the objective video quality. The chapter was concluded with numerical results which showed the improved P S N R performance of the modified codec over the original one. 34 Chapter 4 F O R W A R D L I N K V I D E O T R A N S M I S S I O N I N SINGLE-CELL CDMA SYSTEMS 4.1 Introduction The purpose of this chapter is to investigate the P S N R performance of video transmission over a single-cell D S - S S C D M A forward link under correlated Nakagami fading channel conditions. We first evaluate the B E R performance of a C D M A forward link operating in a N a k a g a m i fading environment by employing Gaussian approximation method. The numerical results are then compared with those obtained from the Monte Carlo simulations. For the mobile channel simulations, a correlated Nakagami fading channel simulator is implemented in software. The modified H.263 codec proposed in the previous chapter is integrated with the IS-95 based C D M A system for video transmission performance evaluation. Modified H.263 Video Encoder Coherent RAKE Receiver Fig. 4.1 Block Deinterleaver CDMA Transmitter Viterbi Decoder Nakagami Fading + AWGN + Interferences Modified H.263 Video Decoder | Reconstructed | Video Simulated D S - S S C D M A Forward L i n k Video Transceiver System M o d e l 35 Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS Fig. 4.1 illustrates the block diagram of the simulated video transceiver system model for the f o r w a r d l i n k o f a D S - S S C D M A s y s t e m t h r o u g h N a k a g a m i f a d i n g c h a n n e l s . T h e uncompressed video sequence in Q C I F format is first encoded by the modified H.263 encoder. Then, the compressed video data are sent to the C D M A transmitter located at the base station. The C D M A transmitter consists of c o n v o l u t i o n a l coder, b l o c k interleaver and direct sequence spectrum spreading sub-blocks. Details of the direct sequence spread spectrum C D M A model are given in Section 4.2. The signal is modulated using binary phase shift keying ( B P S K ) . Next, the transmitted signal goes through the C D M A mobile channel with Nakagami fading characteristics, and is corrupted by A W G N , self-noise interference and multiple-access interference. Details of the channel model and the Nakagami fading software simulator w i l l be described in Section 4.3 and 4.4, respectively. Following the channel, a coherent R A K E receiver located at the handset is used to resolve the multipath signal through the aid of a pilot tone. The coherent R A K E receiver model w i l l be discussed in Section 4.5. The recovered symbols are then deinterleaved followed by hard-decision Viterbi decoding. Finally, the decoded bits are decompressed by the modified H.263 video decoder to obtain the reconstructed video sequence. In Section 4.6, we first establish the mathematical expressions for the single-cell C D M A forward link system model. To obtain the B E R performance for the system, we use both an analytical approach and the M o n t e C a r l o simulation method. In Section 4.7, the Gaussian approximation is employed in the analysis to obtain the B E R performance for the forward link. In Section 4.8, we explain the M o n t e Carlo method and describe the computer simulation model used for the overall system performance evaluation. The numerical results obtained from analytical and computer simulation methods are then compared in Section 4.9. In Section 4.10, we integrate the modified H.263 video codec with the proposed C D M A system and present the numerical results of video transmission performance 36 Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS 37 obtained from computer simulations under different system conditions. Finally, the chapter is completed by its conclusions which are presented in Section 4.11. 4.2 Direct Sequence-Spread Spectrum (DS-SS) CDMA System Model Transmitter Compressed Video Data Convolutional Code r=l/2K=9 Correlated Nakagami Fading Mobile Channel Simulator Block Interleaver Walsh PN Code Sequence 0 - 0 Coherent RAKE Receiver Block Deinterleaver BPSK Modulator Viterbi Decoder s(t) Corrupted Video Data . AWGN + Interference Fig. 4.2 Modulo-2 addition Simulated D S - S S C D M A System M o d e l Fig. 4.2 illustrates the block diagram of the D S - S S C D M A system model which was considered in this thesis. A s per IS-95 standard, the encoded video data is half-rate convolutionally encoded with constraint length of K = 9 . The coding process is described by generator vectors G 0 and G\ which are 753 (octal) and 561 (octal), respectively. After convolutional coding, symbols are interleaved by a block interleaver of size 24 x 16 to randomize bursty errors resulting from the multipath fading channels. The symbols are then spread by a concatenated Walsh/ P N sequence. Each symbol is first replaced by one of the sixty-four orthogonal Walsh codes of length 64 chips, Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS - which is uniquely assigned to each user. Then, modulo-2 addition (concatenation) is performed on each chip with a P N sequence of period 2 15 - 1 chips. The resulting concatenated codes retain orthogonality among users, while reducing cross-correlation surges among non-concatenated Walsh codes. We assume the forward link to be chip-synchronized so that the signals can take advantage of orthogonal covering by the Walsh sequences. The data are finally B P S K modulated before transmitted through the channel. For the mobile C D M A channel, we model the fading statistics as Nakagami-m distributed. In addition to fading and A W G N , the transmitted signal is also corrupted by self-noise interference and multiple-access interference. Self-noise contributes interference which originates from the sidelobes of the autocorrelation function of the spread spectrum code assigned to the reference user. Multiple-access interference is caused by the cross-correlation of the spread spectrum codes among reference user and other simultaneous system users in a multipath signal environment. A t the receiver end, we employ a coherent R A K E receiver to resolve multipaths for improved B E R performance. A number of correlators are assigned to capture the different multipath components. It is assumed that the base station continuously transmits a pilot signal which is used by the mobile receiver to acquire synchronization as well as to make estimation of the channel impulse response. The output of each finger is maximally-ratio combined before being sent to the decision device. The received data are then deinterl'eaved and fed into the hard decision Viterbi decoder to recover transmitted video data bits. 4.3 Mobile Channel Model Since the bandwidth of the spread spectrum signal is usually much wider than the coherence bandwidth of the channel, the channel is considered to be frequency-selective, as explained in 38 Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS 39 Section 2.4.1. A s such, the channel can then be modeled by a tapped delay line with statistically independent time-variant tap weights {c (t)} as shown in F i g . 4.3 [26]. n s(t) \/w[ \/w i/w ----*\\/w\ L- 1 k =0 AWGN + Interferences F i g . 4.3 Tapped Delay Line M o d e l of Frequency-Selective Fading Channel [26]. In the figure, {c (t); n = 0, 1, L- 1} are independent complex-valued stationary random n processes. Their magnitudes \c {t)\ = P „ ( 0 are assumed to be Nakagami-m distributed with P D F n given i n (2.6). The phases Zc (t) = {Q (t)} are uniformly distributed over [0,271) and are n independent of (P„(0l n [23], [29]. The time delay blocks of 1 / W represent the resolution of the multipath delay profile where W is the bandwidth occupied by the real bandpass signal. n(t) is the complex-valued zero-mean white Gaussian noise process with two-sided power spectral density of r\ /2. The number of multipaths L , which may be a random number, is bounded by 0 +1 (4.1) where \_x J is the floor function, T is the maximum multipath spread of the channel and T is c Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS the chip duration. T m is assumed to be less than the bit interval T such that intersymbol interfer- ence (ISI) may be neglected. This is a reasonable assumption for bit-rates of 200kHz or less [23], due to the fact that the typical delay spread T m for urban areas is typically less than 5\ls [30]. For instance, considering a system with data bit-rate 1/7/ = 64 kbits/s and processing gain N = 6 4 , the resulting chip duration, T c multipaths L max = T/N, is around 25 \ls. Therefore, the maximum number of = 5 as given by (4.1). In cases of scattering processes which generate pure diffusive wave-fields, the fading figure m ~ 1 and the N a k a g a m i distribution is identical to the R a y l e i g h distribution. In the presence of direct component, the Nakagami-m distribution approximates the Rician distribution with m > 1. The approximate relationship between the Rician K factor and Nakagami m fading figure has been described in (2.8). 4.4 Correlated Nakagami Fading Simulator To simulate the time-selective correlated fading channel of mobile radio environment, a software simulator was implemented based on the physical model for the radio wave propagation process in different scattering environments [51]. This approach deviates from other more conventional simulators for t i m e - v a r y i n g m o b i l e radio channels, w h i c h use digital filters to filter white Gaussian noise to model Doppler spectra [41]. These conventional methods produce Doppler . spectrum w h i c h only approximates the actual one because the digital filter used i n the noise generator must be implementable. However, in the real world, the measured D o p p l e r spectra usually have irregularities that differ themselves from the perfectly smooth U - T u b shape of the spectrum generated by the transfer function [24]. The'approach followed here employs a Doppler spectrum generated by a direct summation of the various partial waves, which is an exact replica 40 Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS of the actual physical situation. This superposition.of partial waves constitutes the Random Phasor Interference problem in which the total phasor can be interpreted as a vector sum in the complex plane as shown in F i g . 4.4 [51]. When applying the model to the physical situation of radio wave propagation, the total solid: t = 0 dotted: t = At bold: total phasor at t = 0 F i g . 4.4 Example of a Phasor Summation of Partial Waves phasor (bold-line vector in F i g . 4.4) represents the total wave field of a mobile signal. The total phasor is the vector sum of numerous random phasors (solid-line vectors). In analogy, the total wave field is the summation of a continuum of partial waves caused by scatterings. To take the Doppler effect into account, consider a short time period At for which the mobile velocity v causes a change of receiver position by y • At. A s s u m i n g relatively large distances between scatterers and the mobile, we can approximate the phase shifts of the partial 2K waves (dotted-line phasors) caused by the mobile movement as -r-vArcosoc •, where v = |v| is A, J the magnitude of v , X is the wavelength and oc- is the incident angle of the j t h partial waves ; relative to the mobile's velocity vector y . In terms of a mathematical expression, the time-dependent total wave field can thus be 41 Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE- CELL CDMA SYSTEMS 42 expressed as [37] E (t) z = e -expj/^(p? + ^ v ? c o s O y j j + e exp|/^cpQ + ^ v ? c o s a ^ | 7 0 0 (4.2) where e • and cp^ are theyth partial wave amplitude and original phase, and K is the total number of scattered partial waves. In the above equation, the first summation term represents the superposition of partial waves, whereas the second term represents a (possibly strong) component resulting from direct or specularly reflected waves. Clearly, the Doppler shifting effect is taken into account by the phase shift factor - r - v r c o s a • in the phase component terms. A J A s far as short-term fading is concerned, over a short time interval A r , the quantities e •, cp° and CLj can be assumed to be constant, because the distances between scatterers and receiver are large compared to the mobile's motion in At. Thus, their values are fixed over the mobile run in fading simulation. For a N a k a g a m i fading channel, the amplitude r = \E (t)\ follows the N a k a g a m i - m z distribution described in (2.6). Nakagami has shown that the following relations exist between chosen Nakagami parameters and physical partial wave parameters [35]: Q. = N -ej + el (N-7l m = +e ) 2 (4.3a) 2 0 - J (N-ej + e) Q -e Q which can be solved to give [51] e 0 = JQJI — 1/m (4.3b) K • ej = Q ( l - 7 l - 1/m). Thus, for a certain fading characteristics quantified by the average power Q. and the fading Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS figure ra, the partial wave parameters, N , e 0 and {ey, 43 j = 1,2,..., K } can be determined. The authors in [51] suggest around 100 partial waves to be sufficient to reproduce the actual physical situation. The following graphs show the comparison between the simulated signal and the actual Nakagami distribution with different values of fading figures m. The left side of F i g . 4.5 and F i g . 4.6 show the histograms of the Nakagami fading simulator output using l x l O 6 data samples with 3.5 Normalized Fading Amplitude Fig. 4.5 Nakagami Random Variable r Comparison of Nakagami Fading Simulator and P D F for m = 1 2.5 Normalized Fading Amplitude Fig. 4.6 Nakagami Random Variable r Comparison of Nakagami Fading Simulator and P D F at for m = 3 Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS fading figure m = 1 and 3, respectively. In both cases, the average power Q. is set to be unity, total number of partial waves K = 100, and velocity v = 27.78 m/s, which is equivalent to 100 km/hr. The right side of the figures are the P D F plots of Nakagami-ra distribution of corresponding fading figures for comparison. A s shown in both figures, the fading simulator produces samples with excellent accuracy with respect to the theoretical distributions. To measure the accuracy of the simulator outputs quantitatively, we compare the Cumulative Probability Function (CDF) of the Nakagami-ra distribution with the number of sample counts in the histogram. A s an example, from the simulator output histogram of Fig. 4.5, we obtain a minor deviation of 0.91% from the theoretical C D F value at r = 0.5 . Thus, the implemented fading simulator produces accurate fading outputs following Nakagami-m distribution. 4 . 5 Receiver Structure To take advantage of the wideband characteristics of spread spectrum signals, a coherent R A K E receiver is employed in the system to provide multipath diversity. To simplify the mathematical analysis, we w i l l be assuming that perfect knowledge of the channel phase and gain can be obtained [23], [29]. However, for computer simulation, perfect knowledge of channel amplitude is not required. Coherent R A K E combining method weights the resolved multipaths in proportion to their instantaneous received signal envelopes and adds the components constructively. F i g . 4.7 shows the receiver structure for reference user (k = 0 ) , where the number of fingers, L , is a variable parameter less than or equal to the total number of multipaths L, as r defined in (4.1). L is made to be a variable so that the effect of diversity may be observed. The r matched filter is matched to the reference user's C D M A spreading code and is assumed to have achieved time synchronization with the initial path of the reference signal. The tap weights 44 Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS F i g . 4.7 (P-; i = 0, Lr 45 Coherent R A K E Receiver M o d e l 1} and phases {()•; / = 0, Lr 1} are a s s u m e d to be the p e r f e c t estimates of the channel parameters. In practice, the estimation and coherent demodulation may be done jointly through pilot tone calibration techniques [27], [28]. The sampling times of the receiver are t n = nT + (L r l)T , c where n is an integer index. T h e first term is from the matched filter sampling time, and the second term is from the combining of the (L r 1) paths following the initial path. The sampling output U is then used in the decision device for demodulation decision. 4.6 CDMA Forward Link BER Performance Analysis In this section, we w i l l derive mathematical expressions for the signals in a single-cell C D M A forward link system. In such a system, the transmitted signal for the kth user is a phase-coded carrier expressed as s \t) (k = j2Pa \t)b \t)cos((a t (k ik c + $) (4.4) Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS 46 where a "'{t) is the spreading code sequence of the M i user, K oo a \t) = X afp {t-jT ), (k a afe{-\,\} c (4.5) (k) and b (t) is the data waveform, oo b \t) = £ {k b p (t-jT), bf e{-l,\}. {k) . } b (4.6) j =•-•*> In (4.4), P = E /T is the average transmitted power assumed to be equal for all users, where E b b is the signal energy per bit and T is the bit duration. 0 ) is the common carrier radian frequency, c and ()> is the initial phase of the modulator uniformly distributed around [0, 2%). In ( 4 . 5 ) and (4.6), T is the chip duration, p {t) and p {t) are rectangular pulses of unit height and duration c a b of T and T, respectively. c For synchronous operation with K simultaneous users, the composite signal at the input of the channel of the C D M A forward link is K-\ s(t) = £ j2Pa \t)b \t)cos(® t [k + <))). {k c (4.7) k=0 For a single-cell, multi-user model, the received signal after the channel is K-\L-\ r(t) = j2P^ ^^ a \t-x )-b \t-x )cos((i) t {k r (k l [ c + ^ ) + n(t) (4.8) l k = 0l = 0 where K is the total number of users in the cell, and L is the number of multipaths of random value upper bounded by (4.1). (p = § + Q - (d %[ is the phase of the Ith path, where 0 is the channel ; l c ; phase shift and % is the multipath time delay for the /th path. E a c h path is assumed to fade l independently with fading coefficient (3^ ' of which the amplitude P follows a Nakagami-ra ; Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS 47 distribution and the phase 6 follows an uniform distribution over [0, 2K). Finally, n(t) is the ; additive white Gaussian noise process. E a c h signal path w h i c h matches w i t h the nth R A K E finger gives a desired signal («) component denoted as S . I n addition, at the nth finger of the R A K E receiver, there is multiple- access interference denoted as I^ ai generated from the cross-correlation between the reference user's code and other users' codes. Moreover, there is self-noise interference denoted.as produced by cross-correlation with multipath delay copies of its own signal. Finally, A W G N contributes noise interference term denoted as to the desired signal. In [29], the coherent R A K E receiver output is given for an asynchronous (chip-misaligned) C D M A system. To adapt the receiver output for a synchronous system as i n the case of forward link, we replace the independent fading amplitudes and signal phases for each user's multipath signal by common values which are shared by all users in the same cell. Thus, the response of the coherent R A K E receiver with L fingers at each sampling time t L-\(T + nT ) r n U= " r C can be expressed as c j KOM ('-"^os(oy + (p)</f X = = nT + (L -\)T (1) B 0 nT < (4.9) -*(«) , r O ) , {n) , An T n= 0 where the four output response components are given by: S M = J'-C^-iM K -— 1 L i\ k=1 1 U W - 1 — 1 1=0 2 (4.9a) Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS r7 2 ) L I if" = ( 0 ) 48 -l P P/-{^ooK/]+C^ooK/]}-'cos(<p B f l / ) (4.9c) /= 0 l*n T + nT c *ni = j n(t)$ a \t-nT )cos((>i t c In the above equations, whereas b^ + q> )dt. (0 n c (4.9d) n is the information bit of the reference user to be detected is the preceding bit. x = x -i nl l n is the time delay difference and cp the phase difference between the Ith and the nth multipath, respectively. R Q(i) k n/ = <p -(p l n and R o(x) k is are the continuous-time partial cross-correlation functions between the kth and the reference user and are defined by [31] as i ko( ) R % = \a (t-x)a (t)dt k (4.10) 0 o T hoW = ja (t-x)a (t)dt. k (4.11) 0 T We need to transform the continuous-time partial cross-correlation functions into discrete form to evaluate for discrete binary sequences. By choosing / to fulfill 0 < IT < x < ( / + 1 ) T < 7 \ then these two cross-correlation functions can be written as [31] C C koW R = C (l-N)T +[C (l+l-N)-C (l-N))-(x-lT ) k0 ^ A O ( X ) = C (l)T c k0 where C k0 defined as c k0 k0 c + [C (l+l)-C (l)]-(T-lT ) k0 k0 c is the discrete aperiodic cross-correlation function for the sequences {a^} (4.12) (4.13) and ( a j ° ) } Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS 49 N-X-l £ 0<l<N-\ flj«aj'] ; 7= 0 C V) k0 =\ N £ (4-14) 1 + l l-/V</<0 aWjaC'') 7=0 U\>N. 0 For a chip-synchronized system which is considered in this thesis, the residual cross-correlation values represented by the second term of (4.12) and (4.13) are zero. Thus, we can simplify the partial cross-correlation functions as follows: **oCO = C (l-N)T k0 hoW (4.15) C = C (l)T . k0 ' c (4.16) 4.7 Forward Link BER Analysis with the Gaussian Approximation The R A K E receiver output U i n (4.9) is the summation o f the output o f each finger, w h i c h consists of the desired signal component, multiple-access interference, self-noise interference and A W G N components. In order to simplify the B E R performance analysis of C D M A systems, the Gaussian approximation is often applied to the interference terms [5], [23], [29], [33]. It is found to be quite accurate even for small values o f K (e.g. less than 10) when B E R > 10~ , (see for 3 example [42]). Therefore, in this section, we w i l l apply Gaussian approximation on the multipleaccess interference I^ ai and the self-noise interference I^f in (4.9b) and (4.9c) for B E R perfor- mance analysis of the C D M A forward link. 2 The variance of multiple-access interference oMAl n for an asynchronized C D M A system at the nth finger of the receiver, conditioned on | 3 , can be expressed as [29] n Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS 50 • I I iP„}£[{P/}] K-\L-\ *MAl,n = f 2 k= 2 l/ =0 E[{b^R {x ) k0 nl *=1 + b R (x )} ] { k) 0 2 k0 nl • £[cos (cp )] 2 n / (4.17) /=0 The detailed derivation of (4.17) is presented in Appendix A . r {N) is the average cross-correlaK tion parameter over a l l the possible I J = {^(yX"- l ) / 2 } combinations of sequence pairs among K spreading sequences with length of N [31]. For the purpose of the C D M A B E R performance evaluation, we w i l l be using random binary spreading sequences for both the analysis and Monte Carlo simulations. It is equivalent to the assumption that code sequences of period much longer than the processing gain is being used. The resulting value of r (N) for random binary K spreading sequences is found to be 2N in [29] which is also confirmed by our simulation results. 2 (k) (k) The data bits b_{ and b Q in (4.17) are considered to be identically, independent distributed (i.i.d.) random variables with equal probability of taking values +1 or - 1 . For the C D M A forward link system that we are considering, chip synchronization among signals can be assumed. Thus, to obtain the conditional variance of multiple-access interference for a synchronous system, we follow a similar approach as in [23] and [29] by adding a multiplication factor of 3 / 2 to (4.17). This correction factor is used to compensate for the increased interference statistics from overlapping chips among the reference user and other users due to chip synchronization. The compensation is further justified by the results obtained by F o n g et al. in [33]. B y substituting the value of r (N) and adding the chip-synchronized compensation factor K 2 to (4.17), the conditional variance of the multiple-access interference o fied to M A [ „ can now be simpli- Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS 2 51 K-\L-\ PT 2 (4.18) MAI,n *= i / = o where Q. = E[{ (3;} ] is the average power for the signal of the Ith path. l 2 Self-noise interference a s i n can be derived using the similar approach as in [23]. Self- noise (SI) may be considered as an additional multiple-access user where instead of L paths, there would only be L - 1 paths at the input to the receiver because one path contributes to the desired signal component U . Thus, self-noise interference can be approximated by s 2 PT Z o L-l r n , 2 V n (4.19) From (4.9d), the variance of A W G N conditioned on (3 is [29] n 2 a NI,n (4.20) ~ Therefore, the response of the reference receiver U to the received signal at any sampling instant can now be modeled as a conditional complex Gaussian random variable with conditional mean of the desired signal component of the total received signal! (4.21) n= o and conditional variance equals to the sum of all interference terms Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS L- 1 R a. = 52 (°MAI,n X + a SI + a i n NI, ) n n =0 K-\L-\ ET b L - l (4.22) L -1 r jfc= l / = 0 /= 1 | N N '"•o | n = 0 An exponential multipath intensity profile (MIP) is assumed for the average power Q. at T the output of the channel as a function of path delay. The assumption is justified by the actual measurements made by Turin in an urban environment [38]. Therefore, we have Q, = Q e - , (4.23) 8>0 / 5 n where £l is the first path average signal strength and 6 is the rate of average power decay. 0 Substituting (4.23) into (4.22) yields 2 _ (E Tn \ b Q \(K-l)q(L,b) g(L,6)-\ T\ 1 0 ^ (4.24) •L- 1 where q(L, 8) = £ e~ = (1 - e ) / ( 1 - <r ) . /= o The received signal-to-noise ratio (SNR) at the output of the receiver is l5 L 5 5 (4.25) SNR = 2 By denoting the random component of the SNR as L„- 1 s and the deterministic component as =h E<J>.> : n= 0 (4.26) Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL T = i i • ((K-l)q(L, 5) | g (L, 5)-l /v 53 CDMA SYSTEMS ^ + /v ( 4 > 2 7 > the received S N R can be written in compact form as TS. We observe that when there is an absence o f multipath (i.e. L = 1), then q{L, 8) = 1. Thus, the self-noise term represented by ^—1 i n (4.27) becomes zero. The self-noise interference vanishes in a single path environment because it originates from the cross-correlation of the reference user in the existence of different multipath signals. For coherent demodulation in the presence of A W G N , the probability of error conditioned . on the instantaneous signal-to-noise ratio (SNR) can be expressed as [23] P (S) = Q(JTS) (4.28) e where Q(x) = —L= f°e~ x2/2 dx. To obtain the average error probability for random variable 5, the average of P (S) over the P D F of S has to be obtained as follows: e oo P = jP (S)p(S)dS. e (4.29) e o The solution for the average error probability P for the general case of non-integer Nakagami e fading has been evaluated in [23] as - T\ A 5 2 s ( i I N where ,L r \ ,L -\ N2 . r y -' / y 5 ( -8i)2^ e e Vi = 0 J \i =0 J q(L„b) 2 q(L , 25) r Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS 54 (4.30b) i = 0 and TQ S Yq(L , 2 5 ) 2m s 2mq(L , 8)' r (4.30c) r In the above equations, T(-) is the Gamma function defined as T(x) = e~'t ~ dt, x x>0 x (a) (b) d and k 2 F j ( - ) is the h y p e r g e o m e t r i c f u n c t i o n d e f i n e d as F\(a, 2 b, c, d) = ^ k k with (c) k\ k =0 = 1. The a n a l y t i c a l B E R performance of the C D M A k (a) k = a(a+ l)---(a + k- 1), (a) 0 forward link is evaluated through numerical computations of (4.30)-(4.30c). The results w i l l then be compared to those obtained from the Monte Carlo simulation method. 4.8 Computer Simulation System Description Besides evaluating the C D M A forward link B E R performance through analysis, we also compare the results with those obtained from Monte Carlo computer simulation techniques based on the response of the coherent R A K E receiver given in (4.9). D u r i n g the simulations, the M i user (k) (k = 0, K - 1) is assigned a r a n d o m signature sequence {a (/); i = 1, N - 1} o f processing gain N. For each simulation run, the preceding information bit b[\^, and the information bit to be detected , are assigned to the kth user, b^ and b^ take values from the alphabet { ± 1 } with equal probabilities and are i.i.d. The fading signal amplitude of the nth multipath P , is a random variable with Nakagami-m distribution, and is generated by means of n computer simulation using the Nakagami fading simulator as described in Section 4.4. The /th multipath delay x is randomly generated in discrete units of the chip duration T , 0 < % < T. In l c t addition, the phase of the nth path of the carrier (p is a random variable with a uniformly distrin Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS 55 bution over [0, 2TC). Lastly, A W G N term n(r) is generated by a Gaussian random process with zero mean and variance given by (4.20). A t the nth finger of the coherent R A K E receiver, the desired signal component along with other interference components such as multiple-access interference 7^2 •, self-noise interference 7 ^ and noise interference term 7^"\ are generated according to (4.9a)-(4.9d). These components are summed up to form the output of each finger. The total output from L number of r fingers of the reference user's R A K E receiver forms the decision output [ / s h o w n in Fig. 4.7. The detected bit is then compared with the transmitted bit to check i f an error is made. The process is repeated to receive an adequate number of errors (e.g. greater than 500) in order to achieve a 95% confidence interval of +5 % of the average B E R . 4.9 B E R Performance Evaluation Results for C D M A F o r w a r d L i n k In this section, we present the numerical results for the B E R performance of the proposed singlecell C D M A forward link system. Fig. 4.8 compares the B E R performance of the C D M A forward link obtained from both analytical method and Monte Carlo simulation techniques. For comparison purposes, similar to the analysis presented in Section 4.7, random spreading sequences are used in Monte Carlo simulations. The transmitted bit energy-to-noise density ratio was selected to have a relatively large value of E /r\ b 0 = 30 d B so that A W G N effect is r e l a t i v e l y l o w as compared to the user interferences. Therefore, we can concentrate on the B E R performance results of the user-interference-limited C D M A systems in terms of the number of users K, while ignoring the insignificant interference from A W G N [33]. The total number of multipaths is set to L = 5 as previously explained in Section4.4, and the cases of using R A K E fingers L r = 2 and 3 are considered. The processing gain is assumed to be A = 64 as per IS-95 standard. We use the 7 Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS 56 fading figure m = 1 for the Nakagami fading channel which is equivalent to Rayleigh fading commonly encountered in urban environments [5]. 10" Gaussian Approximation Analysis L=5,Lr=2 Monte Carlo Simulation L=5,Lr=2 Gaussian Approximation Analysis L=5,Lr=3 Monte Carlo Simulation L=5,Lr=3 30 35 40 Number of Users K F i g . 4.8 60 Single-Cell CDMA Forward Link B E R Performance As shown in the figure, the B E R performance results obtained from the analysis using the Gaussian approximation are very optimistic with respect to the simulation results for both cases of L r = 2 and 3. The B E R performance differences can be explained by the fact that the Gaussian approximation used in the analysis is based on the central limit theorem, which assumes the random variables to be i.i.d [26]. During the analysis in Section 4.7, we apply the Gaussian approximation on the multiple-access and self-noise interference terms. However, on a forward Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS link the interference signals of other users in the same cell arrive at the reference user with the same channel characteristics such as amplitude fluctuations, phase distortions and transmission time delays. Therefore, the Gaussian approximation does not quite accurately predict the B E R performance due to the lack of statistical independency among multiple-access interference signals. However, as we w i l l see in the following chapter, the Gaussian approximation provides much more accurate B E R performance predictions for multiple-cell systems. From the same figure, we also observe that in the case of L r = 2, there is a closer match in B E R performance results obtained from the analytical and simulation methods when compared to the case of L r = 3 . Moreover, the result discrepancies between the two methods become smaller as the number o f users increases for both eases of L . These observations can be r explained by the fact that, in general, the Gaussian approximation provides more accurate results at higher bit error rates [43]. Next, in Fig. 4.9 we compare the B E R performance of the proposed C D M A system utilizing concatenated codes and one that uses non-concatenated codes. The purpose is to illustrate the effectiveness of orthogonal spreading in a multipath and multiuser environment. Concatenated coding scheme uses spreading codes generated from modulo-2 addition between the P N sequence and the W a l s h codes. O n the other hand, non-concatenated scheme uses spreading codes generated from the maximal-length, sequence directly, without performing modulo-2 addition with the Walsh codes. Two cases with different number of multipaths and R A K E fingers are considered here. In b o t h cases, m u l t i p a t h s i g n a l o f i d e n t i c a l and n o r m a l i z e d path strengths ( i . e . f2 ; = Q = 1) are generated with the same fading figure of TO = 1. Spreading gain is N = 6 4 , the period of the P N sequence is 2 1 5 - 1, and E /r\ b Q = 10 d B . The first pair of curves (solid lines) represents the B E R performance of non-concatenated -57 Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS 58 • 10 Non-concatenated Code L=3 Lr=2 Concatenated Code L=3 Lr=2 Non-concatenated Code L=1 Lr=1 Concatenated Code L=1 Lr=1 ,-K 10" 10" Fig. 4.9 10 15 20 25 30 35 40 Number of Users K 45 50 55 60 Simulation Results Comparison of Concatenated and Non-Concatenated Schemes codes versus concatenated codes of a C D M A system with L = 3 and R A K E receiver of L r = 2. A s shown, the concatenated coding scheme provides better performance, and the improvement becomes greater as the number of users increases. This improvement happens because of the orthogonal spreading effect provided by the Walsh codes among users. A s the number of users increases, multiple-access interference remains suppressed relative to the non-concatenated coding scheme. In the second set of curves (dotted lines), we assume L = 1 (i.e., no multipath) instead, and a single correlator receiver is used. We can see a dramatic improvement of the concatenated c o d i n g scheme over the non-concatenated one. T h i s is because i n the absence o f Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS multipaths, perfect orthogonality between multiple users' spreading codes is maintained, thus resulting in zero multiple-access interference and self-noise. In fact, the B E R performance of the concatenated coding scheme remains constant over the number of users in a single path environment, as the only source of interference is from the A W G N . O n the other hand, P N sequences used in the non-concatenated case do not possess such properties and thus, when the number of users increases, multiple-access interference increases accordingly, which results in degradation of B E R performance. 4.10 Video Transmission Performance in CDMA Forward Link In this section, we present the numerical results from computer simulation of the proposed video transmission system over a single-cell C D M A forward link operating i n correlated Nakagami fading channel environment. We have incorporated the m o d i f i e d H . 2 6 3 v i d e o codec into previously described C D M A system and have evaluated the video performance as a function of P S N R for different number of resolving paths and signal propagation characteristics, as well as for the presence of channel estimation errors. For the simulations, we use the well-known video sequence Miss America of 100 frames coded by the modified H.263 encoder at a fixed bit-rate of 64 kbits/s. The video sequence represents a typical videophone/video-conferencing situation. It shows the upper half of a woman talking in front of a plain, steady background with natural facial, head and shoulder movements during the conversation. The 64kbit/s video data is half-rate convolutionally coded and then spread by the concatenated codes of length 64, resulting in a chip rate of 8.192 M C h i p / s . The P N sequence used is a maximal-length sequence of generator polynomial 100003 in octal representation [57], and has a period of 2 1 5 - 1 chips. The shift registers are initialized by one 1 and all 59 Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS zeros. The 64 Walsh sequences are randomly assigned to each user, except for the all-zero Walsh code which is reserved for the pilot channel. The frequency-selective channel is simulated by a tapped delay line model as described in Section 4.3, whereas each tap weight is independently generated by the correlated N a k a g a m i fading simulator as described in Section 4.4. The multipath intensity profile is assumed to be logarithmic with the exponent value made variable to allow for more general case evaluation. Signal phases are random variables uniformly distributed in the interval [0, 2n). We assume that the base station continuously transmits a pilot signal which is used by the mobile receiver to make an estimate of the channel impulse response for maximal-ratio combining of different multipath components as well as acquiring synchronization for coherent demodulation. P S N R of each data point is obtained by averaging 20 simulation runs using different random seeds. Unless stated otherwise, the default values of the system parameters used to obtain the following numerical results are as follows: • Total number of multipaths L = 3 . This value is chosen to be less than the maximum number of multipaths L max = 5 as discussed in Section 4.3 for the assumed maxi- mum multipath spread of 5 \is. • Number of multipath resolving R A K E fingers L • Nakagami fading figure m = 1 for simulating mobile channels in typical urban envi- r = 2. ronment. • M o b i l e velocity is 100 km/hr and carrier frequency is 2 G H z , which results in a maximum Doppler frequency B d of about 180 H z . Thus, the default channel has B T ~ 2.8 x 10~ representing fast fading characteristics. 3 d 60 Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS • 61 Multipath intensity profile (MIP) is logarithmic with default exponent o = 0 , which is equivalent to uniform MIP. In F i g . 4.10, we present the video performance of the modified codec operating in a single-cell C D M A system. We observe a similar P S N R improvement of the modified version over the original version as in Fig. 3.2. In this case, there is l - 4 d B P S N R performance improvement over the range of user number under consideration. 10 F i g . 4.10 15 20 25 30 35 Number of Users K 40 45 50 55 60 Video Performance of Modified H.263 Codec in Single-Cell C D M A System In Fig. 4.11, the video transmission performance is evaluated against different number of resolvable paths by the R A K E receiver.As it can be observed, the P S N R performance improves as Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS 10 F i g . 4.11 L r 15 20 25 30 35 Number of Users K 40 45 50 55 62 60 P S N R Performance with Various Number of R A K E Fingers increases from 1 to 3. This coincides with the theoretical o p t i m a l value o f L r = L for coherent demodulation because a coherent L -finger R A K E with perfect estimates of the channel r tap weights is equivalent to a maximal ratio combiner with L th-order diversity [26]. In fact, the r improvements are quite substantial with each increase in the number of resolving paths. For the case of L r = 1 (single correlator receiver), the recovered video data is so corrupted that both F E C offered by the modified H.263 video codec and half-rate convolutional coder of IS-95 are unable to provide enough error recovery, resulting in relatively poor P S N R performance. F o r L = 2 , the quality of video is acceptable at P S N R = 40dB for up to about 25 users. In the case of Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS L r 63 = L = 3 , there is virtually no degradation of the video sequence quality for up to 60 users due to the fact that the channel coding is able to correct almost all of the errors. It is worth noticed that three-path diversity is indeed used in practical C D M A forward link systems [55]. F i g . 4.12 shows the P S N R performance of the video transmission under different channel conditions with Nakagami fading figure m = 1, 2 and 3. The fading figure m is related to the 30 • 35 40 Number of Users K Fig. 4.12 60 P S N R Performance with Different Fading Figures severity of fading. Increasing value of m corresponds to less amount of fading. A s the value of m increases, we observe that there is significant improvement in the video quality in terms of P S N R because of the diminishing severity of the fading. The gaps in P S N R among different m-value Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS cases become smaller when the number of users K gets larger. This is because the increasing multiple-access interference has become the dominant factor in the degradation of the video transmission over the effects of fading of the channels. 70 I —e— —-0— 60< I 5=0 5 = 0.2 5 = 0.4 •- 50 •- 20 10 0 10 15 20 25 30 35 40 45 I I 50 55 60 Number of Users K Fig. 4.13 P S N R Performance with Different Logarithmic M I P Exponents F i g . 4.13 illustrates the effects of different multipath intensity profiles (MIP) of the signal on the P S N R p e r f o r m a n c e where we c o n s i d e r cases o f l o g a r i t h m i c e x p o n e n t v a l u e s 5 = 0, 0.2, 0.4. Without loss of generality, we assume that the multipath signals have normalized power in their first path, i.e. Q 0 = 1. For 5 = 0 , it is equivalent to a constant M I P , i.e. all the multipaths have the same average power. 5 of 0.2 and 0.4 correspond to multipath signal power 64 Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS set of {OdB, -0.86dB, -1.74dB} and {OdB, -1.74dB, -3.47dB}, in the order of arrivals, respectively. The decreasing power in the multipath signals has two major effects when multipath signals are partially resolved (L < L) as in these cases. Firstly, there is less interference arriving r at the receiver. This is especially true for the self-noise due to the power reduction of delayed versions of the reference user's signal. A t the same time, the sum of the desired signal power decreases as well. A s shown in the figure, the average P S N R performance shows improvement for increased values of 8. This happens because the reduced multiple-access and self-noise interference have a higher impact on the system performance than the reduced desired signal power does. In F i g . 4.14, we present results showing the impact of D o p p l e r spread on the system performance. We consider cases of mobile velocity at 100 km/hr, 50 km/hr and 5 km/hr to r e p r e s e n t d i f f e r e n t s c e n a r i o o f h a n d s e t u s a g e . T h e f i r s t t w o c a s e s c o r r e s p o n d to BT d = 2.8 x 10~ and 1.4 x 10~ w h i c h c h a r a c t e r i z e fast f a d i n g , whereas the last case 3 corresponds to B T d 3 = 1.6 x 10~ which characterizes slow fading. Average duration of fades is 5 inversely proportional to B T d product. Thus, fast fading cases have shorter average fade duration than slow fading. A s observed from the curves, higher mobile velocity gives better average P S N R performance. The reason is that the error control used in the system, either B C H or convolutional coding, is designed to combat random errors. Thus, for a fixed degree of interleaving, higher velocity corresponds to shorter fade durations and thus shorter error bursts, which in turns facilitate error correction by the channel coding. The opposite applies to situations of slower speed, of which longer error bursts make channel coding ineffective, thus the resulting degrade in P S N R performance. F i g . 4.15 shows the average P S N R performance of the video transmission under channels with non-identical m values. Due to the random nature of the propagation channel, it is more '65 Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS 66 v= 100km/h, BdT = 2.8e-3 v = 50km/h, BdT = 1.4e-3 v = 5km/h, BdT= 1.6e-5 30 35 40 Number of Users K Fig. 4.14 60 P S N R Performance with Different M o b i l e Velocities realistic to assume that the fading figures m may be different for different multipaths. This may w e l l be the case in an actual mobile link, since the radio waves take different paths and may undergo different fading before arriving at the receiver. Moreover, it is also reasonable to assume that the signal from the initial path experiences less severe fading. It is because the signal may either come from a direct wave or reflected wave with relatively less scattering than those from subsequent multipaths. A s typical examples, we have considered three multipath fading scenario in terms of m values in the order of signal arrivals namely, {1.0, 1.0, 1.0} , {2.0, 1.5, 1.0} and {3.0, 2.0, 1.0}. F o r the case of multipath m values equal to {1.0, 1.0, 1.0}, it represents a Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS 70 I I I ! ! I o — « — — i — 60> 1 67 I m=[1.0,1.0,1.0] m= 2.0,1.5.1.0 m=[3.0,2.0,1.0] •- 50 5 30 20 10 0 10 15 i i 20 25 30 i 35 40 45 50 55 60 Number of Users K F i g . 4.15 P S N R Performance with Non-Identical Multipath Fading for Each Path channel in which all multipaths exhibit Rayleigh fading. A s observed from the graph, it results in the lowest P S N R performance. Improved P S N R performance of about 7 dB for 25 system users is obtained in the case of multipath m values equal to {2.0, 1.5, 1.0} . It is due to the fact that the first and second paths have relatively less severe fading. Finally, for the case of multipath ra values equal to {3.0, 2.0, 1.0} , it gives the best video performance among all cases in the simulation. Again, it is because of the still lower average fading intensity of the multipath signals. F i n a l l y , we have investigated the impact of imperfect estimation o f the channel tap weights on the video transmission quality. A s previously mentioned, in the IS-95 C D M A forward Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS link, channel estimation is achieved with the aid of a pilot tone sent from the base station to the receiver. In practice, when fading is sufficiently slow, relatively good channel estimates can be obtained. However, in situations where fading is relatively fast, there can be inaccuracies in the estimation process. In order to evaluate the effect of deviations in the channel tap weights estimation on video transmission performance, we assumed an estimation error with Gaussian distribution. In the computer simulations, the exact channel fading amplitudes are deviated with Gaussian variance of different values to generate the imperfect channel tap weight estimations. 1 1 k. _ r <r -fcr—— - \* _ )- •. •• —e—; —*— —-0— • Deviation Variance = 0 Deviation Variance = 0.25 Deviation Variance = 0.5 5 10 15 20 25 30 Number of Users K Fig. 4.16 P S N R Performance with Imperfect Channel Estimation As illustrated in In Fig. 4.16, we compare two cases of imperfect estimation with error of 68 Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS 69 variance 0.25 and 0.5, and include the case of perfect channel estimation for reference purposes. A s seen from the results, imperfect channel tap weights estimation could degrade P S N R performance considerably. For example, let us consider the case of having 50 users i n the system, receiver with perfect channel estimation can provide acceptable video quality of P S N R at about 50 d B . However, the video P S N R values drops to about 7 dB and 18 dB for receivers with channel estimation error of variance of 0.25 and 0.5, respectively. Therefore, we observe that the effect of channel estimation error could have a significant effect, on the video transmission performance of C D M A systems. 4.11 Conclusions In this section, we have investigated the video transmission performance of modified H.263 encoded data in a single-cell C D M A system over correlated Nakagami fading channels for the forward link. The D S - S S C D M A system, fading channel model and R A K E receiver structure were described. Mathematical analysis for the C D M A forward link was presented, and its results were compared with those obtained from computer simulation. It was found that Gaussian approximation provides fairly optimistic B E R prediction for the single-cell forward link system. Through simulations, we illustrated that concatenated coding scheme provides superior B E R performance over non-concatenated scheme for the forward link, thus justifying its applications in the forward link of the IS-95 C D M A systems. In the last section, we presented the P S N R performance of the integrated C D M A forward link video transceiver system through software simulation method. We have shown that the optimal value of L r = L for coherent R A K E receiver. P S N R performance improves with larger values of fading figures, M I P exponents and BT d products. We also illustrated that the effects of unequal fading for individual multipaths and Chapter 4 FORWARD LINK VIDEO TRANSMISSION IN SINGLE-CELL CDMA SYSTEMS imperfect channel estimations can be significant to P S N R performance. Chapter 5 F O R W A R D L I N K V I D E O T R A N S M I S S I O N I N MULTIPLE-CELL CDMA SYSTEMS 5.1 Introduction It is w e l l known that, the cellular concept is being used i n wireless multiple access systems to increase capacity by reusing radio resources i n different cells. W h i l e T D M A and F D M A techniques must provide different frequency allocation for contiguous cells, C D M A systems can reuse the same entire spectrum for all cells, thus greatly simplify frequency planning [5]. Unlike the capacities of T D M A and F D M A systems, which are primarily bandwidth limited, C D M A capacity is primarily limited by interference level [53]. In fact, the capacity is inversely proportional to the amount of interference. For a multiple-cell environment, in addition to the usual A W G N , self-noise interference, and multiple-access interference from users present in the reference cell, there is additional interference contributed from users from neighboring cells. In this chapter, we w i l l investigate the effects of this multiple-cell interference on the video transmission performance of a C D M A forward link. After this introduction, in Section 5.2, we first present the multiple-cell configuration model, which consists of the first two tiers of surrounding cells. In Section 5.3, we introduce the channel model for multiple-cell systems that takes large-scale attenuation into account. In Section 5.4, the D S - S S C D M A system model for multiple-cell environment is described. In Sections 5.5 and 5.6, we present the mathematical analysis of the B E R performance for the multiple-cell C D M A forward link. In Section 5.7, we describe the computer simulation model used in Monte Carlo method. In Sections 5.8 and 5.9, numerical results for the B E R performance of the C D M A forward link and the P S N R performance of the transmitted video over the proposed multiple-cell 71 Chapter 5 FORWARD LINK VIDEO TRANSMISSION IN MULTIPLE-CELL CDMA SYSTEMS 72 C D M A system are presented and discussed. Finally, we complete the chapter with conclusions in Section 5.10. 5.2 Multiple-Cell Configuration Model In this section, we describe the configuration model for multiple-cell systems used in this thesis. F i g . 5.1 shows the C D M A cellular model under consideration w h i c h takes into account the Forward Link Signal From Base Station To Mobile (Reference and 1st Tier Cells) --' »- • Forward Link Signal From Base Station To Mobile (Similar for all 2nd Tier Cells) . Base Station Fig. 5.1 ' Multiple-Cell Configuration M o d e l multiple-cell interference coming from the 18 surrounding cells of the first two tiers. Cells are Chapter 5 FORWARD LINK VIDEO TRANSMISSION IN MULTIPLE-CELL CDMA SYSTEMS 73 numbered from 0 to 18, with number 0 denoting the reference c e l l , and 1 to 18 denoting the surrounding cells. The signals from cells farther away are neglected in the investigation. It is justified by the fact that they suffer greater attenuation due to path loss. To simplify analysis, each cell is assumed to be circular with uniform size, and that each base station (BS) is located at the centre o f the c e l l . A l s o , the reference m o b i l e user is assumed to be equally l i k e l y located anywhere within the reference cell. In F i g . 5.1, the reference mobile, located at the convergence point of arrows representing the forward link signals, receives radio signals from the reference base station (denoted by the square at cell 0) as well as from the surrounding 18 base stations (denoted by the squares at cell 1 to 18). 5.3 Channel Modeling In this section, we describe the radio propagation model for signals travelling in a multiple-cell environment. To model the mobile channel for such a cellular environment, the effects of largescale fading due to shadowing and path loss have to be taken into account. We employ the largescale fading model as discussed in Chapter 2, which assumes that the average signal attenuation is the product of the yth power of distance and a log-normal random variable (see (2.1)). Thus, we can relate the transmitted average signal power P and the received average signal power P q from the <?th B S to the reference mobile as P^ocP^-YioVio, where d q ( 5 .l) is the distance between the reference mobile and the q\\\ base station, y is the path loss exponent, and L is a Gaussian random variable with zero mean and standard deviation G . We q w i l l use y = 4 for the power law and G q = 8 dB for the standard deviation of the log-normal shadowing random variable L , as it has been suggested in past literature [41], [53]. Chapter 5 FORWARD LINK VIDEO TRANSMISSION IN MULTIPLE-CELL CDMA SYSTEMS For s m a l l - s c a l e f a d i n g , we e m p l o y the same frequency-selective f a d i n g m o d e l as described in Section 4.3 for signals transmitted from each base station. Each forward link signal from the qth cell is modeled by a tapped delay line (Fig. 4.3) with statistically independent timevariant tap weights {c (t)}, where q = 0, 1, q[ Q- 1 and / = 0, 1, L- V. The variable Q is the total number of cells (assumed to be 19) and L is the total number of multipaths upper bounded by (4.1). The magnitudes of the tap weights |c /(0| = r\/(0 g f ° thefthsignal paths from r the qth c e l l s are assumed to be N a k a g a m i - m d i s t r i b u t e d and the c o r r e s p o n d i n g phases Zc {t) = Q (t) to be uniformly distributed over [0, 2n). ql q[ 5.4 Multiple-Cell DS-SS CDMA System Model In this section, we describe the direct sequence spectrum spreading scheme for the C D M A systems in a multiple-cell environment, which is similar to that of the single-cell systems as described in Section 4.2. The spreading codes are obtained from the concatenation of Walsh codes with a P N sequence. For a multiple-cell C D M A system, the same P N sequence is shared by all B S . Each B S , however, uses a different time-shifted versions of this common P N sequence for Walsh code concatenation [56]. Mobile receivers detect the amount of time offsets of the P N code to identify one B S from another. We assume the signals transmitted from different B S are chip-synchronized, i.e. signal delay x = kT for some integer k. In the actual IS-95 C D M A network, the system time i s c synchronous to that of Universal Time Coordination ( U T C ) , which uses the same time origin as Global Positioning System (GPS) [55]. A l l base stations in the C D M A system synchronize to the same system time accurately, as it is crucial for proper system operations. Thus, it is justified to assume that the base stations transmit signals in a chip-synchronized manner. 74 Chapter 5 FORWARD LINK VIDEO TRANSMISSION IN MULTIPLE-CELL CDMA SYSTEMS 75 5.5 Multiple-Cell CDMA Forward Link BER Performance Analysis For a C D M A forward link operating in a multiple-cell environment, the transmitted signal for the M i user from the qth cell is a phase-coded carrier expressed as sf{t) = JlT a \t)b \t)co^ t k q +^) k q q c (5.2) q (k) where a q (t) is the spreading code sequence of the kth user from the qth cell, J oo and b q I jPa(t-JT ), j = -oo ? W = a <$ {-1,1} (5.3) ^ e { - l , l } . (5.4) ai k q c G (t) is the data waveform, oo fM I = b j ^ {t-jT), } Pb -oo = In (5.2), P is the transmitted power from the qth base station, co is the common radian carrier q c frequency, and § is the initial phase of the qth base station modulator uniformly distributed over q [0, 2n). In (5.3) and(5.4), T is the chip duration, T i s the data bit duration, p (t) and p (t) are c a b rectangular pulses of unit height and durations of T and T, respectively. c The total signal transmitted by the qth base station, assuming there are K users belonging to the qth base station, is qW= S s JW a %)b (t)cos((Q t k) q q q c + $ ). (5.5) q k =0 For a multiple-cell model, the received signal after the channel is Q-\K„-\L -\ q W = S 9 X £ j— j2P ^ af(t-x )-b \t-x )cos(m t k q qr ql q ql c + y ) + n(t). (5.6) ql = 0/t = 0 / = 0 In (5.6), Q is the total number of cells, K is the total number of users in the qth cell, and L is q q Chapter 5 FORWARD LINK VIDEO TRANSMISSION IN MULTIPLE-CELL CDMA SYSTEMS 76 the number of multipaths for the signal from the qth base station, which is a random value upper bounded by (4.1). Similarly to the single cell system model, P the qth cell as denned in (5.1). q = § + B + ca t • x ql base station's signal, where § q q ql c q[ is the original phase, Q ql is the received signal power from is the phase of the Ith path from the qth is the channel phase shift and x is the ql multipath time delay for the Ith path from the corresponding c e l l , respectively. E a c h path is assumed to fade independently with fading coefficient $ e ql q[ follows a Nakagami-m distribution and the phase Q in w h i c h the amplitude $ q [ follows a uniform distribution over [0,2TC). q[ The final term n(t) is the A W G N process with two-sided power spectral density of r\ /2 0 . Equivalent to the single-cell system analysis in Section 4.6, the response of the coherent R A K E receiver at each sampling time can then be expressed as + nT L -\T r c V = £ n=o j q c c + y )dt n n T c = ^{s L r(t)V J°\t-nT )cos'<ii t ( in) + i^ i + i^ + 5 7 ) i^} n= 0 where L r is the number of r e s o l v i n g R A K E fingers. In (5.7), S component, I^li is the multiple-access interference component, 2^ is the d e s i r e d s i g n a l is the self-noise component, and / „ / is the A W G N component at the output of each R A K E finger, respectively. 5.6 Multiple-Cell BER Performance Analysis with Gaussian Approximation B y following the same procedure of applying the Gaussian approximation to B E R performance analysis as in Chapter 4, we can write the response U of the reference receiver to the received s i g n a l at any s a m p l i n g instant as a c o n d i t i o n a l c o m p l e x G a u s s i a n r a n d o m v a r i a b l e w i t h conditional mean of the desired signal component U , and conditional variance equals to the sum s Chapter 5 FORWARD LINK VIDEO TRANSMISSION IN MULTIPLE-CELL CDMA SYSTEMS 77 of all interference terms, i.e., L -\ r °S = X ( MAI,n + G SI,n + G NI,n)n=0 d 2 (-) 2 5 ' . In order to find the desired signal component U , we modify U s 8 . in (4.21) so that the s large-scale fading effect is taken into account. The original signal power P of the reference user is replaced by the attenuated power P 0 °f the reference user located at cell 0 as given by (5.1) such that ^ A M T - SiPoJ (5.?) 2 n= 0 where (3 0ra is the fading amplitude of the nth resolvable path of the reference base station. 2 To find the total interference 6 S for the multiple-cell systems, we apply certain modifica2 tions on the i n d i v i d u a l components of total interference G s in the single-cell scenario. F o r multiple-access interference ( M A I ) , in conjunction with the amount caused by intra-cell users, multiple-cell systems have additional interference generated by K number of users from each of the surrounding cells, for q = 1, ..., Q - 1 . Moreover, since signals from different cells reach the reference mobile through different channel environment, they each have different attenuation and 2 multipath intensity profiles. To account for these differences, the <s MMn term i n (4.18) is modified by adding a second term to represent the summation of multiple-access interference from cell 1 to cell Q - 1. A l s o , we replace the common power P of the reference c e l l by the attenuated signal power P q for the qth cell as given by (5.1). Finally, the multipath intensity profile Q. for the reference cell in (4.18) is replaced by a more generalized term Q. t the multipath intensity profile for the qth cell, which is defined as ql denoting for Chapter 5 FORWARD LINK VIDEO TRANSMISSION IN MULTIPLE-CELL CDMA SYSTEMS n,/ = * V " ' . (- ) <a where Q, q0 78 5 10 is the first path average signal strength and d is the rate of average power decay, for q the qth base station signal propagation, respectively. After accommodating all these changes, we obtain the conditional M A I variance at the nth R A K E finger for a multiple-cell system as p K -\L -\ 2 n T 0 k=l Q-ip 2 K T 1=0 q=1 i - . l L i - ' 1 *=0 /= 0 where the first term represents M A I from within the reference cell, and the second term represents M A I from the surrounding cells. The summation sign of users changes from k = 1, K- 1 0 for the reference cell user M A I term to k = 0, ..., K - 1 for the surrounding cell M A I term, i n order to account for an additional M A I interferencing users from each of the surrounding cells. Self-interference is obtained by modifying (4.19) in a similar manner. The common signal power P in (4.19) is substituted by the attenuated signal power PQ of the reference cell. The M I P term Q, is replaced by the M I P term for the reference cell denoted as Q, , which is defined in L 0l (5.10), to give '= i The A W G N component d NI ing the fading amplitude (3 by p n n 0 n for the multiple-cell environment is obtained by substitut- to denote for the reference cell, which gives 4/,».= V { P o » } • (5 - 13) The received S N R with a R A K E receiver resolving L paths for a multiple-cell system is defined r Chapter 5 FORWARD LINK VIDEO TRANSMISSION IN MULTIPLE-CELL CDMA SYSTEMS 79 as SNR = (u y (5.14) s L, - 1 2J ^ MAI,n SI,n n= 0 a After substituting t / j , 6 M i 4 / n , &si,n' a n a °NI,ni +a " m t o + (5-14), and performing some straightfor- ward simplifications, we obtain A- 2 I ^ • ^o„> 2 n = OV + II n Q- "o/+ k= 1 / = 0 S-{P n> 2 0/ + 2 0 Q 2 2 K -lL -\ n \ 2 <iW v« = o / 2PnSNR = 1 T - ^ 0 J X ^ • ^oJ 2 2 q q Z 1 2 K -\L -\ (5-15) £ = 0 /=0 ' J In an interference-limited C D M A system, spreading codes of multiple users constitute the dominant source of interference when compared to A W G N . This is especially true for a multiplecell system where typically the number of users is large. In light of the relative insignificance of 2 the A W G N term <5 m n , the last summation term of the denominator of (5.15) can be dropped to simplify the mathematical analysis. Since we are only interested in the relative received power from the qth base station to that of the reference base station, we introduce the attenuation factor of the. signal power from the qth (q ^ 0) base station relative to that of the reference base station (q = 0) as (5.16) as illustrated in Fig. 5.2. In this figure, cell q represents any of the 18 surrounding cells in the first Chapter 5 FORWARD LINK VIDEO TRANSMISSION IN MULTIPLE-CELL CDMA SYSTEMS 80 CellO • Base Station • Mobile Cell 9 Fig. 5.2 C e l l Geometry Model' two tiers, Si is the cell radius and O is the mobile location angle with respect to the base station. In addition, assuming identical M I P and uniform average number of users among different base stations, i.e. (5.17) q K = K {] (5.18) = K then (5.15) can be written as L -\ r SNR = J (K-\L-\ 2N Vjfc = 1 / = 0 n=0 2-1 K-\L-\ q=\ k = 0l = 0 (5.19) L-\ \ 1=1 J Similar to Chapter 4, the received S N R at the output of the receiver may be re-written in a more compact form as Y S , where m m e-i Y m and v = 27V (K - 1 )q(L,0) +^P Kq 1 q(L,6) + (q(L,S) - 1) (5.20) Chapter 5 FORWARD LINK. VIDEO TRANSMISSION IN MULTIPLE-CELL CDMA SYSTEMS °n = .81 0 To obtain the area-averaged B E R , mean values of the attenuation factor p q in (5.20) for different reference mobile locations need to be evaluated. We assume that the power levels transmitted by all base stations are the same, i.e. P — PQ — P for all q. We further assume that the mobiles are always served by the base station with the strongest signal, i.e. p < 1. This assumption is well q justified by the fact that this type of selection diversity is used in practical C D M A systems for soft handoff operations [55]. Under these conditions, we follow the method of F o n g et al. [32] in evaluating the values o f E[p ] q for the first and second tier cells the results o f w h i c h are <b<8&fitecffiffis iM sfig? 0 075M a Mtir ssffl* arc $e m m 0.03315 eg <=s Tic. s<»£ fl§ • -. - ().()24<W Table 5.1 Area Mean of Attenuation Factor summarized in Table 5.1. The detailed derivation of how these values have been obtained can be found in Appendix B . After applying these values of attenuation factor on T , we can then replace Y and S in m (4.28) by T m and S m to obtain the area-averaged B E R for the multiple-cell C D M A system as ( W 1+Y * h 2jnT(m +\) s n + 2 V 2 l.+ Y , Chapter 5 FORWARD LINK VIDEO TRANSMISSION IN MULTIPLE-CELL CDMA SYSTEMS 82 where a new term T ' introduced is defined as s T q(L , 28) Y Q y >_ m s _ 2m 5 v r> ^ 22a) > 2mq(L , 8 ) s r Numerical results w i l l be presented in Section 5.8 with those obtained from Monte Carlo method, of which the simulation model w i l l be described in the next section. 5.7 Computer Simulation Model Description In this section, we describe the software simulation model used for the performance evaluation employing Monte Carlo error counting techniques. Comparing to a single-cell system, additional K users from Q - 1 surrounding cells are needed to generate in the multiple-cell system simula- t i o n . T h e p r e c e d i n g bits b^f k = 0, 1, and present bits b ^ K - 1 and q = 0, 1, q Q of the kth user o f the qth c e l l (for Q - 1) are generated as random i.i.d. variables of values taken from the alphabet { ± 1 } with equal probabilities as in Section 4.8. The signal amplitude P , the phase <p , and the multipath delays % of the Ith path (/ = 0, 1, ?/ ql ql L - 1) from the q gth base station, as well as the A W G N random process are generated in the same manner as in the single-cell scenario. Decision output is obtained by adding the desired signal component S multiple-access interference 7^,-, self-noise interference 7^ from the nth R A K E finger, for n = 0, Lr and A W G N interference term , 7^ 1 where L is the total number of resolvable r fingers. Error count is obtained by checking the detected bit with the transmitted bit. The process is repeated to receive an adequate number of errors in order to achieve a 95% confidence interval of ± 5 % of the average B E R . Chapter 5 FORWARD LINK VIDEO TRANSMISSION IN MULTIPLE-CELL CDMA SYSTEMS 5.8 BER Performance Evaluation Results for Multiple-Cell Systems F i g . 5.3 shows the B E R performance of a C D M A forward link i n a multiple-cell environment obtained by analytical and Monte Carlo simulation methods. A s in the case of single-cell systems, 10 Gaussian Approximation Analysis L=5 Lr=3 Monte Carlo Simulation L=5 Lr=3 Gaussian Approximation Analysis L=5 Lr=5 Monte Carlo Simulation L=5 Lr=5 10" 10 15 20 F i g . 5.3 25 30 35 40 Number of Users per Cell Kq 45 50 55 60 Analysis vs. Simulation for Multi-cell System random spreading sequences are used in both methods for comparison purposes. System parameters are chosen as E /r\ b Q = 30 dB and processing gain N = 64 as before. We assume a fading figure value of m = 2.0 i n the comparison to demonstrate the flexibility o f Nakagami fading channel modeling. We consider two cases of total number of multipaths L = 5 with different 83 Chapter 5 FORWARD LINK VIDEO TRANSMISSION IN MULTIPLE-CELL CDMA SYSTEMS number of R A K E fingers L r = 3 and 5, respectively. A s shown in. the figure, the B E R results evaluated from the analysis using the Gaussian approximation and the Monte Carlo simulation show good agreement. This contrasts with the observation in the case of single-cell systems in Section 4.9 where we found a fair amount of B E R performance discrepancies between the two approach. This improvement can be explained by the fact that i n m u l t i p l e - c e l l systems, forward l i n k signals from different base stations exhibit independent propagation characteristics such as amplitude fluctuations, phase distortions and transmission delays, with respect to other stations. Thus, the overall interference components of a multiple cell system consist of more independent random variables as compared to the single-cell 1 case. A s a result, the central limit theorem in the Gaussian approximation is better satisfied in the process of analyzing B E R performance of the multiple-cell systems. We also observe the trend of decreasing discrepancies between the B E R performance results obtained from the two methods in cases of increasing number of users per cell K Q and higher B E R values. This is similar to that has been observed in the single-cell scenario, which has been discussed and explained in Section 4.8. 5.9 Video Transmission Performance in Multiple-Cell Systems In this section, we present the numerical results obtained from computer simulations of the video transmission performance over the proposed multiple-cell C D M A forward l i n k operating i n Nakagami fading channel environments. The simulated system model is similar to that of Section 4.9, except in this case, we incorporate additional 18 surrounding cells around the reference cell in the system according to the configuration model described in Section 5.2. Each surrounding base station uses a different time-shifted version of the common P N sequence for concatenation with the Walsh codes. The signals transmitted from each base station travel through an independent 84 Chapter 5 FORWARD LINK VIDEO TRANSMISSION IN MULTIPLE-CELL CDMA SYSTEMS frequency-selective channel simulated by the tapped delay line model similar to that of the singlecell case. The clocks of a l l base stations are assumed to be synchronized at the chip level as explained previously in Section 5.4. The results are presented in terms of average P S N R values versus number of users per cell. A s in the previous chapter, each P S N R data point is generated by taking the average of 20 simulation runs using different random seeds. Unless otherwise stated, the default values of the system parameters used to obtain the following numerical results are as follows: • Total number of multipaths L = 3 . This value is chosen to be less than the maximum number of multipaths 5 as evaluated in Section 4.3 for the assumed maximum multipath spread of 5 [is. We w i l l vary L during the course of performance evaluation. • Number of multipath resolving R A K E fingers L r = 3 , as is generally used in practi- cal C D M A systems. • M o b i l e velocity is 100 km/hr and carrier frequency is 2 G H z , resulting in a maximum Doppler frequency B d B T ~ 2.8 x 1 0 d • - 3 of about 180 H z . Thus, the default channel has value of which can be considered as exhibiting fast fading characteristics. Multipath intensity profile is logarithmic with default exponent 5 = 0.2. In F i g . 5.4, we illustrate the average P S N R performance of video transmission under mobile channels with non-identical Nakagami fading figures (the m values) for each multipath. A s examples, we consider three multipath fading m values in the order of signal arrivals namely, {1.0, 1.0, 1.0}, {2.0, 1.5, 1.0} and {3.0, 2.0, 1.0} . The last case gives the best P S N R performance because it has the lowest average fading severity. In F i g . 5.5, we have evaluated the video performance for multiple-cell systems where multipath signals transmitted from base stations of different tiers go through m o b i l e fading 85 Chapter 5 FORWARD LINK VIDEO TRANSMISSION IN MULTIPLE-CELL CDMA SYSTEMS 86 70 r m= 1.0,1.0,1-0 m= '2.0,1-5,1.0' m= '3.0,2.0,1.0' 60<k 50 CD ;o DC Z CO 40 D. <B D) a 30 > < 20 0 i 10 i 15 1 20 1 1 1 25 30 35 40 Number of Users Per Cell K F i g . 5.4 P S N R Performance with Non-Identical m Values for Each Multipath channels m o d e l e d w i t h non-identical set of m values. A s examples, we consider cases o f [{3.0,2.0,1.0}, {2.0, {2.0,1.5,1.0}, {1.0,1.0,1.0}], [{2.0,1.5,1.0}, {2.0,1.5,1.0}, 1.5, 1.0}] and [{2.0, 1.5, 1.0}, {1.0, 1.0, 1.0}, {1.0, 1.0, 1.0}], respectively. F o r each case, the set of m values in the first {•} correspond to the fading figures of the reference-cell channel, the set of m values in the second {•} correspond to the fading figures of the 6 first-tier surrounding-cell channels, and the set of m values in the third {•} correspond to the fading figures of the 12 second-tier surrounding-cell channels. The values in each {•} in turn represent the fading figures for the multipath signals in the order of arrivals. In the first case, we investigate Chapter 5 FORWARD LINK VIDEO TRANSMISSION IN MULTIPLE-CELL CDMA SYSTEMS 10 15 20 25 30 35 87 40 Number of Users K Per Cell F i g . 5.5 P S N R Performance with Non-Identical m Value Set for Channels from Different Tiers the system of which signals transmitted from the reference base station suffers relatively less severe fading than those from the surrounding base stations. The simulation resembles the physical situation in which the reference mobile is most likely closer to the reference base station than the other base stations. Thus, there is a higher probability for the existence of direct line-ofsight signals from the reference base station than from the surrounding base stations. A s a result, it is justified to assume that the channel fading for the reference-cell signal is relatively less severe than the surrounding-cell signals. From the figure, we observe that the video performance of the first case is better than that of the second case where the reference cell has higher m values or less Chapter 5 FORWARD LINK VIDEO TRANSMISSION IN MULTIPLE-CELL CDMA SYSTEMS fading. Next, for the transmission systems in the second and third case, we compare the P S N R performance where the reference-cell channels are assumed to share the same fading figure set but the surrounding-cell channels are assigned with different set of m values. We observe that in the third case where the surrounding-cell channels exhibit more severe fading, better P S N R performance is obtained. This can be explained by the fact that the greater severity of fading in the surrounding-cell channels reduces the amount of multiple-access interference signals from reaching the reference mobile. In F i g . 5.6, the effects of different user load of surrounding cells to video transmission performance are evaluated. In previous investigations, we have assumed that the number of users in the surrounding cells are the same as that of the reference cell. Here, we consider scenario of different surrounding-cell user loads as we evaluate the P S N R values of systems with all the 18 surrounding cells being occupied by 1 / 2 , 1/3 and 1/4 of the. reference-cell user population. A s shown in F i g . 5.6, the P S N R performance improves as the surrounding-cell user load decreases. For example, at the-capacity of 25 users in the reference cell, we observe an increase of about 14dB and 19dB in P S N R performance for the 1/3 and 1/4 capacity cases when compared to the 1/2 capacity case, respectively. The gain in P S N R performance is due to the decrease in the amount of multiple-access interference generated from the neighboring-cell users. It has been clearly seen from this simulation that the user load of surrounding cells could have a significant impact on the video transmission performance in a C D M A system. In F i g . 5.7, we investigate the C D M A video transmission performance with non-identical multipath intensity profile for channels o f cells in different tiers . A s examples, we consider cases of 8 = 0.2, 0.2, 8 = 0.2, 0.4 and 8 = 0.2, 0.8 where the first value corresponds to the logarithmic M I P exponent of the reference-cell channel, and the second value corresponds to that of the 88 Chapter 5 FORWARD LINK VIDEO TRANSMISSION IN MULTIPLE-CELL CDMA SYSTEMS 89 70 Surrounding Cells at 1/2 Capacity Surrounding Cells at 1/3 Capacity Surrounding Cells at 1/4 Capacity 10 Fig. 5.6 15 20 25 30 User Capacity of the Reference Cell 35 40 P S N R Performance with Different Surrounding-Cell User L o a d first and second tier surrounding-cell channels. Since signals transmitted from surrounding cells travel through different kinds of physical environment, it is reasonable to model the channels of the reference cell and surrounding cells with different 8 values. A s shown in the figure, there are significant differences in P S N R results between different cases of 8 values. For instance, at the system capacity of 30 users per cell, we obtain P S N R improvement of approximately lOdB and 20dB in the second and third case (non-identical M I P ) over the first case (uniform MIP), respec- tively. This can be explained by the fact that greater logarithmic M I P exponent in surrounding-cell signals means interferencing multipath signals are weaker when they reach the reference mobile, Chapter 5 FORWARD LINK VIDEO TRANSMISSION IN MULTIPLE-CELL CDMA SYSTEMS 70 90 ! 5 = 0.2, 0.2 5 = 0.2, 0.4 —e— —*— — 5 6( = 0.2, 0.8 - 50 20 10 0 25 Fig. 5.7 I 30 35 40 Number of Users K Per Cell 45 50 P S N R Performance with Non-Identical M I P for Channels from Different Tiers which translates to less multiple-access interference to the received signal. We have also investigated the impact of imperfect channel estimation of the channel tap weights on the video transmission quality. Similar to the single-cell scenario, we simulate the estimation errors as Gaussian random variables with mean of the exact fading amplitude and deviation variance denoted as o . We have compared two cases including perfect channel estimae tion ( a = 0 ) as well as imperfect estimation with error variance o e e = 0.25 and the results are presented in F i g . 5.8. A s seen from this figure, imperfect channel tap weight estimation could noticeably degrades P S N R performance. F o r example, at P S N R value o f 4 0 d B , the system Chapter 5 FORWARD LINK VIDEO TRANSMISSION IN MULTIPLE-CELL CDMA SYSTEMS 91 70 Deviation Variance = 0 Deviation Variance = 0.5 10 F i g . 5.8 15 20 25 30 Number of Users K Per Cell 35 40 P S N R Performance with Imperfect Channel Estimation capacity is reduced from 28 users to 23 users due the channel estimation errors. Finally, we have evaluated the video transmission performance of the C D M A system with non-identical number of multipaths for cells at different tiers. In other investigations, we assume that a l l cells share the same number of multipaths; however, due to the variant nature of the physical environment that different base-station signals have to travel through, it is reasonable to assume that the number of multipaths for cells from different tiers are not necessarily identical. Moreover, it is also logical to assume that the neighboring-cell channels have longer delay spread or more multipaths due to the greater distance that the signals have to travel to reach the reference Chapter 5 FORWARD LINK VIDEO TRANSMISSION IN MULTIPLE-CELL CDMA SYSTEMS 20 Fig. 5.9 25 30 Number of Users K Per Cell 92 40 P S N R Performance with Non-Identical Number of Multipaths for Different Tiers mobile. Typical performance evaluation results are presented in F i g . 5.9, where LO, L I and L 2 correspond to the multipath number of the reference cell, first-tier cells, and second-tier cells, respectively. We consider cases of LO = L I = L 2 = 3 , LO = 3, L I = 4, L 2 = 5 as w e l l as LO = 4, L I = 5, L 2 = 6 . We observe a significant amount o f P S N R degradation as the multipath number of the neighboring-cells increases, which is due to the corresponding increase of multiple-access interference from the extra multipath signals under consideration. Chapter 5 FORWARD LINK VIDEO TRANSMISSION IN MULTIPLE-CELL CDMA SYSTEMS 5.10 Conclusions In this chapter, we have investigated the video transmission performance of modified H.263 encoded data in a multiple-cell C D M A system over correlated Nakagami fading channels for the forward link. The multiple-cell configuration model, channel model and cellular C D M A system m o d e l were e x p l a i n e d i n details. M a t h e m a t i c a l analysis for the C D M A f o r w a r d l i n k was presented, and its results were compared with those obtained from computer simulation. It was found that Gaussian approximation provides better B E R prediction for the multiple-cell forward link system than for the single-cell system. We have presented the P S N R performance of the integrated C D M A forward link video transceiver system through software simulation method. It was shown that video transmission performance improves as the average m value of the multipath channel increases. Moreover, it was demonstrated that the more severe the surrounding-cell channel fading conditions, the better the P S N R performance can be achieved. It was found that neighboring-cell capacity has significant impact on video transmission quality o f a cellular C D M A system. Further, we illustrated that the video performance improves for greater logarithmic M I P exponent for the outer-cell channels. Channel estimation errors have been shown to produce a significant degradation on system performance. Lastly, non-identical multipath channel modeling for cells at different tiers revealed the adverse effect of increased number of multipaths of neighboring-cell channels on the P S N R performance. 93 Chapter 6 CONCLUSIONS AND SUGGESTIONS FOR FUTURE RESEARCH 6.1 Conclusions In this thesis, we investigated the video transmission performance in the forward link of IS-95 based G D M A cellular systems over correlated Nakagami fading channels. The major contributions of the thesis are summarized as follows. 6.1.1 A Modified H.263 Video Codec We have proposed a modified H.263 video codec to improve the average and frame-to-frame P S N R performance of the encoded video bitstream for transmissions i n error-prone mobile channels. The proposed codec incorporates a selective forward error correction coding scheme and a periodic I N T R A frame forced update mechanism to m a x i m i z e video quality. T h i s is achieved w h i l e avoiding excessive time delay critical for real-time video applications and minimizing channel coding redundancy as well as complexity. We have shown that improvement of 3-10 d B i n average P S N R performance can be achieved by the proposed scheme when compared to the original version. 6.1.2 Correlated Nakagami Fading Simulator We have implemented in software a correlated Nakagami fading simulator to model the multipath mobile channels. The simulator is based on the physical model for the radio wave propagation process for different scattering environments. It generates fading output by the summation of complex phasors, with Doppler shift element incorporated to create the time correlation relationship between outputs. It has advantages over conventional Rayleigh and Rician fading simulation 94 Chapter 6 CONCLUSIONS AND SUGGESTIONS FOR FUTURE RESEARCH because Nakagami-m distribution provides a more generalized and versatile fading modeling for the mobile channel. Moreover, the Doppler spectrum generated from this approach is the exact replica of the actual physical situation, as opposed to the idealistic U-tub shape obtained from conventional digital filtering method. 6.1.3 Analysis of CDMA Forward Link We have presented the B E R performance analysis of the C D M A forward link for both single-cell and multiple-cell environment employing the Gaussian approximation. It was observed that for random spreading sequences, the Gaussian approximation does not provide accurate predictions for B E R performance of single-cell systems. On the other hand, it predicts the B E R results quite accurately for multiple-cell systems. These observations are explained by the fact that in singlecell systems, the interference signals transmitted from the base station to the reference mobile exhibit the same amplitude fading, phase distortions and time delays, thus do not satisfy the random variable independency requirement of the central limit theorem. In the case of multiplecell environment, although the multiple-access interference from users within the same cell shares common channel characteristics, interference signals from different neighboring cells are statistically independent from each other. Therefore, the multiple-cell interference components are more accurately described by the Gaussian distribution as dictated by the Central L i m i t Theorem. 6.1.4 End-to-end Video Transmission Performance Evaluation We have presented the P S N R performance of the modified H.263 coded video transmitted through the proposed IS-95 based C D M A system over correlated Nakagami fading channels, for both single-cell and multiple-cell environment. We have observed that the coherent R A K E receiver 95 Chapter 6 CONCLUSIONS AND SUGGESTIONS FOR FUTURE RESEARCH functions as a maximal-ratio combiner such that the optimal number o f resolvable paths equals the total number of multipaths. In general, the P S N R performance improves as the m values of the reference-cell channel increase due to the corresponding reduction in fading severity. The system also performs better for larger logarithmic M I P exponents due to the net gain i n interference reduction versus desired signal power reduction. We also observed that the sensitivity of the video transmission performance on the channel estimation error can be significant. F o r multiple-cell systems, we found that better P S N R performance is achieved when the neighboring cells have more severe fading channel conditions, lower user capacity, and greater l o g a r i t h m i c M I P exponent. Lastly, we illustrated that increasing number of multipaths in surrounding-cell channels can degrade the video transmission performance in a C D M A system considerably. 6.2 Suggestions for Future Research • One of the interesting topic for future research would be the evaluation of video transmission performance for the reverse link of the C D M A systems. This w i l l complete . the investigation of the performance of a two-way video communication system. • Employing soft-decision Viterbi decoding at the receiver is also a worthwhile task for further research as it should improve the video transmission performance. • B y adopting the latest H.263+ compression standard for the video codec, new optional features from the standard can be employed to improve the video codec performance. 96 Glossary m Nakagami Fading Figure DCT Discrete Cosine Transform MV Motion Vector QCIF Quarter Common Intermediate Format Y Luminance Component C, B C R Chrominance Component GOB Group of Blocks MB Macroblock d Distance between M o b i l e and Base Station y Path Loss Exponent of the Power L a w P Received Signal Power R P Transmitted Signal Power \\ Gaussian Random Variable of Log-normal Distribution a Standard Deviation of £ T . T m M a x i m u m Multipath Spread in seconds B co Coherence Bandwidth in H z MIP Multipath Intensity Profile B M a x i m u m Doppler Spread in H z d cs a Root Mean Squared Value of Signal Amplitude in Rayleigh Fading <j Average Power of Rician Fading Signal I (-) Modified Bessel Function of the first kind and zero-order K Rician Factor b 0 97 98 A Peak Amplitude of the Dominant Signal T(m) Gamma Function Q. Second moment of Nakagami Fading Signal Amplitude B Spread Spectrum Bandwidth in H z B Original Signal Bandwidth PN Pseudo-noise Z M Output of the M Correlator of a R A K E Receiver a M R A K E Weight for the M Correlator Z' Total Output of R A K E Correlators n Period of a Sequence ss b The ith Binary B i t of a Sequence i Autocorrelation Function ARQ Automatic Repeat Request FEC Forward Error Correction AWGN Additive White Gaussian Noise I-frame I N T R A frame P-frame Prediction Frame QCIF Quadrature Common Intermediate Format PSNR Peak Signal-to-Noise Ratio of, rf Original and Reconstructed Pixel Components P B i t Error Probability e {c {t)} Frequency-Selective Channel Complex Tap Weights { P„(0 } Channel Tap Weight Amplitude n 99 {6 (f)} Channel Tap Weight Phase W Real Bandpass Signal Bandwidth n(t) Gaussian Noise Process — White Noise Power Spectral Density L Number of Multipaths n T Chip Duration T B i t Duration N Processing Gain L M a x i m u m Number of Multipaths v M o b i l e Velocity Vector A Wavelength c a Incident Angle of y'th Partial Waves Relative to M o b i l e Velocity Vector ; E (t) Electric Field of Radio Wave e •, <p°j jth Partial Wave Amplitude and Phase X Number of Partial Waves k Multiple User Index z L Number of Resolvable Paths t Receiver Sampling Time U Receiver's Sampling Output r n s (k) (t) The kth User's Transmitted Signal (k) a (t) b \t) P (k . The kth User's Spreading Code The ytth User's Data Average Transmitted Power 100 Common Carrier Frequency § Common Phase of the Modulator p (t) Rectangular Pulses of Chips p (t) Rectangular Pulses of Data Bits r(t) Received Signal After the Channel K Total number of users a b cp Phase of the/th Path Signal x Multipath Time Delay of the/th Path ; l :,s (n) S Desired Signal Amplitude at the nth Finger (n) I Multiple-Access Interference at the nth Finger 1^ Self-Noise Interference at. the nth Finger mai White Noise Interference Term at the nth Finger (k) b Information B i t of the kth User to be Detected Q (k) b_{ Precedent Information B i t of the M i User x , ni cp Multipath Time Delay Difference Between the nth and /th Paths Multipath Phase Difference Between the nth and /th Paths w/ R [x], k0 CQ Continuous-Time Partial Cross-Correlation Functions Between the kth and the Reference User Discrete Aperiodic Cross-Correlation Functions Between the kth and the Reference User k 2 <5 Variance Sum of A l l Interference Terms S G R-koit] 2 M A • 2 G SI / n 2 <J N I N n Conditional Variance of Multiple-Access Interference at the nth Finger Conditional Variance of Self-Noise Interference at the nth Finger Conditional Variance of White Noise at the nth Finger Average Cross-Correlation Parameter Q.[ Average Signal Power of the /th Path U Desired Signal Component of the Receiver Response E Signal Energy Per B i t 8 Exponent of Logarithmic Multipath Intensity Profile S Random Component of the S N R T Deterministic Component of the S N R F(a, x) Incomplete Gamma Function P Average Received Power from the qh\ Base Station P Transmitted Signal Power from the qth Base Station s b q q d Distance Between the Mobile and the qth Base Station q C, , a Gaussian Random Variable and its Standard Deviation for the Signal Path from the qth Base Station s \t) The jfeth User's Transmitted Signal of the qth Cell a \t) The kth User's Spreading Code of the qth Cell b \t) The kth User's Data of the qth Cell q q ( q { q { q P, P Transmitted and Received Power of Signal from the <^th C e l l , 0 Carrier Phase of Signal from the <?th Cell K Number of Users in the qth Cell r{t) Received Signal After the Channel for Multiple-Cell Environment Q Total Number of Cells L Number of Multipaths of Signal from the gth Cell n n q (p g/ P' e Phase of the /th Path from the gth Cell Average-Averaged Probability of Error for Multiple-Cell 102 Channel Phase Shift for the /th Path from the gth Cell Multipath Time Delay for the /th Path from the qth C e l l Fading Amplitude for the /th Path from the qth Cell Receiver Response at Sampling Time for Multiple-Cell Desired Signal Amplitude at the nth Finger for Multiple-Cell Multiple-Access Interference at the nth Finger for Multiple-Cell Self-Noise Interference at the nth Finger for Multiple-Cell White Noise Interference Term at the nth Finger for Multiple-Cell Large-Scale Fading Attenuation Factor Desired Signal Component of the Receiver Response for Multiple-Cell Variance Sum of A l l Interference Terms for Multiple-Cell Conditional Variance of Multiple-Access Interference at the nth Finger for Multiple-Cell Conditional Variance of Self-Noise Interference at the nth Finger for Multiple-Cell Conditional Variance of White Noise at the nth Finger for M u l t i p l e Cell Average Signal Power of the /th Path from the qth Cell Decay Exponent for the qth Cell's M I P Random Component of the S N R for Multiple-Cell Deterministic Component of the S N R for Multiple-Cell Bibliography [I] R . Kohno, R. Meidan and L . B . Milstein, "Spectrum spectrum access methods for wireless communications," IEEE Communications Magazine, pp. 58-67, Jan. 1995. [2] D . L . Schilling, R. L . Pickholtz, and L . B . Milstein, "Spread spectrum goes commercial," IEEE Spectrum, pp. 40-45, A u g . 1990. [3] R. L . Pickholtz, L . B . Milstein, and D . L . Schilling, "Spread spectrum for mobile communications," IEEE Trans. Vehicular Tech., VT-40, pp. 313-322, M a y 1991. [4] A . Fukasawa, T. Sato, Y. Takizawa, T. Kato, M . Kawebe and.R. E . Fisher, "Wideband C D M A system for personal radio communications," IEEE Communications Magazine, Oct. 1996. [5] T. S. Rappaport, Wireless Communications Principles & Practice. N e w Jersey: Prentice H a l l , 1996. [6] R. Price and P. E . Green, " A communication technique Proceedings of the IRE, pp. 555-570, Mar. 1958. [7] W. C . Y . Lee, "Overview of cellular C D M A , " IEEE Trans. Vehicular Tech., vol. 40, pp. 291-302, M a y 1991. [8] D . C . C o x , R. Murray and A . . Norris, "800 M H z attenuation measured in and around suburban houses," AT&T Bell Laboratory Tech. Journal, vol. 673, no. 6, July-Aug. 1984. [9] EIA/IS-95, "Dual mode mobile station-base station wideband spread spectrum compatibility standard," PN-3119, Electronics Industries Association, Engineering Department, Dec. 1992. [10] N . Morinaga, M . Nakagawa and R. Kohno, "New concepts and technologies for achieving highly repliable and high-capacity multimedia wireless communications systems", IEEE Communications Magazine, pp. 34-40, Jan. 1997. [II] P. Bahl and B . Girod, "Wireless video," IEEE Communications Magazine, June 1998. [12] I T U Telecom. Standardization Sector of I T U , "Video coding communication," ITU-T Recommendation H.263, Mar. 1996. 103 for multipath for channel," low bitrate Bibliography 104 [13] M . Khansari, A . Jalali, E . Dubois and P. Mermelstein, " L o w bit-rate video transmission over fading channels for wireless microcellular systems," IEEE Trans, on Circuits Syst. Video Technol, vol. 6, no. 1, Feb. 1996. [14] R . Stedman, H . Gharavi, L . Hanzo and R. Steele, "Transmission of subband-coded images via mobile channels," IEEE Trans. Circuits Syst. Video Technol, vol. 3, no. 1, Feb. 1993. [15] R . Talluri, "Error-resilient video coding in the I S O M P E G - 4 Communications Magazine, June 1998. [16] M . Khansari, A . Zakauddin, W . Y. Chan, E . Dubois and P. Mermelstein, "Approaches to layered coding for dual-rate wireless video transmission," Proc. of Int. Conf. on Image Processing, 1994. [17] N . Farber and B . Girod, "Robust H.263 compatible video transmission for mobile access to video servers," Proc. of Int. Conf. on Image Processing, 1997'. [18] L . Hanzo and J. Streit, "Adaptive low-rate wireless videophone schemes," IEEE Trans. Circuits Syst. Video Technol., vol. 5, no. 4, A u g . 1995. [19] J. Streit and L . Hanzo, "Quadtree-based reconfigurable cordless videophone systems," IEEE Trans. Circuits Syst. Video Technol, vol. 6, no. 2, Apr. 1996. [20] P. Cherriman and L . Hanzo, "Programmable H.263-based wireless video transceivers for interference-limited environments," IEEE Trans. Circuits Syst. Video Technol, vol. 8, no. 3, June 1998. [21] H . L i u and M . E . Zarki, "Transmission of video telephony images over wireless channels," Wireless Networks, vol. 2, no. 3, 1996. [22] H . Suzuki, " A statistical model for urban radio propagation," IEEE Trans, on Comm., vol. C O M - 2 5 , pp. 673-680, July 1977. [23] T. E n g and L . B . Milstein, "Coherent D S - C D M A performance i n Nakagami multipath fading," IEEE Trans, on Comm., vol. 43, no. 2/3/4/, February/March/March/April 1995. [24] W . Braun and U . Dersch, " A physical mobile radio channel model," IEEE Trans. Veh. Technol, vol. 40, no. 2, pp. 472-482, 1991. standard," IEEE Bibliography 105 [25] I T U - T Recommendation 601, Encoding Parameters of Digital Television for Studios, 1982 [26] J. G . Proakis, Digital Communications. New York: M c G r a w H i l l , 1983. [27] M . Yokoyama, " B P S K system with sounder to combat Rayleigh fading in mobile radio communication," IEEE Trans. Veh. Tech., vol. VT-34, pp. 35-40, Feb. 1985. [28] F. Davarian, " M o b i l e digital communication via tone calibration," IEEE Trans. Veh. Tech., vol. VT-36, pp. 55-62, M a y 1987. • ' [29] G . P. Efthymoglou, V. A . A a l o and H . Helmken, "Performance Analysis of coherent D S C D M A Systems in a Nakagami fading channel with arbitrary parameters," IEEE Trans. Veh. Tech., vol. 46, no. 2, M a y 1997. [30] G . L . Turin, "Introduction to spread-spectrum antimultipath techniques and application to urban digital radio," Proc. IEEE, vol. 68, pp.328-353, Mar. 1980. [31] M . B . Pursley, "Performance evaluation for phase-coded spread-spectrum multiple-access communication - Part I: System Analysis," IEEE Trans, on Comm., vol. C O M - 2 5 , no. 8, July 1977. [32] M . H . Fong, V. K . Bhargava and Q. Wang, "Concatenated orthogonal/PN spreading sequences and their application to cellular D S - C D M A systems with integrated traffic," IEEE J. Select. Areas Commun., vol. 14, no. 3, Apr. 1996. [33] M . H . Fong, Q. Wang and V. K . Bhargava, "Concatenated orthogonal/PN codes for D S C D M A systems in a multi-user and multipath fading environment," Proc. of Global Telecomm. Conf., 1994. [34] D . V . Sarwate and M . B . Pursley, "Crosscorrelation properties of pseudorandom related sequences," Proc. IEEE, vol. 68, pp. 593-619, M a y 1980. [35] M . Nakagami, "The m-distribution - a general formula of intensity distribution of rapid fading," in Statistical Methods in Radio Wave Propagation, W. G . Hoffman, E d . Oxford, England: Pergamon, pp. 3-36, 1960. [36] S. A . Abbas and A . U . Sheikh, " A geometric theory of Nakagami fading multipath mobile radio channel with physical interpretations," Proc. Vehecular Tech. Conf., 1996. their and Bibliography 106 [37] W. R. Brauri and U . Dersch, " A physical mobile radio channel model," IEEE Trans. Veh. Tech., vol. 40, no. 2, M a y 1991. [38] G . Turin, F. Clapp, T. Johnston, S. Fine and D . Lavry, " A statistical model of urban multipath propagation," IEEE Trans. Veh. Tech., vol. V T - 2 1 , no. 1, pp. 1-9, 1972. [39] M . V. Clark, L . J. Greenstein, W. K . Kennedy and M . Shafi, "Matched filter performance bounds for diversity combining receivers in digital mobile radio," IEEE Trans. Veh. Tech., vol. 41, no. 4, pp. 356-362, 1992. [40] H . X i a n g , "Binary code-division multiple-access systems operating in multipath fading, noisy channels," IEEE Trans. Comm., vol. C O M M - 3 3 , pp. 775-784, A u g . 1985. [41] W. C . Jakes, E d . , Microwave Mobile Communications. New Jersey: I E E E Press, 1993. [42] E . Geraniotis, "Direct-sequence spread-spectrum multiple-access communications over nonselective and frequency-selective rician fading channels," IEEE Trans, on Comm., vol. C O M - 3 4 , pp. 756-764, A u g . 1986. [43] E . A . Geraniotis and M . B . Pursley, "Performance of coherent direct-sequence spreadspectrum communications over specular multipath fading channels," IEEE Trans, on Comm., vol. C O M - 3 3 , pp. 502-508, June 1985. [44] K . Yao, "Error probability of asynchronous spread spectrum multiple access communication systems," IEEE Trans, on Comm., vol. C O M - 2 5 , pp. 803-809, A u g . 1977. [45] A . Andreadis, G . Benelli, A . Garzelli and S. Susini, " F E C coding for H.263 compatible video transmission," Proc. of Int. Conf. on Image Processing, 1991. [46] E . Steinbach, N . Faerber, B . Girod, "Standard compatible extension of H.263 for robust video transmission in mobile environments", IEEE Trans, on Circuits Syst. Video Technol., vol. 7, pp. 872-881, Dec. 1997. [47] H . Ibaraki et at, "Mobile video communication techniques and services", Proc. ofSPIE Conf. on Visual Commun. and Image Proc, pp. 1029-1033, 1995. [48] U . Horn, B . Girod and B . Belzer, "Scalable video coding with multiscale motion compensation and unequal error protection", Proc. Int. Symp. on Multimedia Commun. and Video Coding, Oct. 1995. Bibliography 107 [49] T. Yang, J. Chalidabhongse and C . C . J. K u o , "Performance study of robust wireless video transmission over fading channels," Conf. Rec. of the 30th Asilomar Conf. on Signals, Systems and Computers, 1997. [50] P. Bahl and I. Chlamtac, "H.263 based video codec for real-time visual communications over wireless radio networks," Proc. of Int. Conf. on Universal Personal Comm., 1997. [51] U . Dersch and R. J. Ruegg, "Simulations of the time and frequency selective outdoor mobile radio channel," IEEE Trans. Veh. Tech., vol. 42, no. 3, A u g . 1993. [52] A . H . Wojnar, "Unknown bounds on performance in Nakagami channels," IEEE Trans. Commun. vol. C O M M - 3 4 , pp. 22-24, Jan. 1986. [53] K . S. Gilhousen, I. M . Jacobs, R. Padovani, A . J. Viterbi, L . A . Weaver, Jr., and C . E . Wheatley III, " O n the capacity of a cellular C D M A system," IEEE Trans. Veh. Tech., vol. 40, no. 2, M a y 1991. [54] K . R. Rao and J. J. Hwang, Techniques and Standards for Image, Video, and Audio Coding, New Jersey: Prentice Hall P T R , 1996. [55] W. C . Y . Lee, Mobile Cellular Telecommunications: Analog and Digital Systems, M c G r a w - H i l l International Editions, 1995. [56] A . Salmasi and K . S. Gilhousen, " O n the system design aspects of code division multiple access ( C D M A ) applied to digital cellular and personal communications networks," Proc. of Veh. Tech. Conf, 1991. [57] R. L . Peterson, R. E . Ziemer and D . E . Borth, Introduction to Spread Sectrum Communications, New Jersey: Prentice H a l l , 1995. [58] F. Davarian, "Channel simulation to facilitate mobile-satellite communications research," IEEE Trans. Commun. vol. C O M M - 3 5 , no. 1, Jan. 1987. Appendix A. Derivation of Multiple-Access Interference Variance for the Gaussian Approximation From (4.9b), the multiple-access interference component in the output response of the receiver is 1 —1 #i=J-X I K — P B P / - { ^ i ( ^ ) +4%i(^/)}-cos(q) ) ( A 1 ) n / k=1 1 = 0 A s s u m i n g that the period o f the C D M A spreading code sequence (see E q . (4.5)) is large as compared to the processing gain N = T/T , we can model the sequences as random binary c sequences, which are mutually independent. A s such, I^ can be treated as a summation of ai independent random variables. In light o f the presence o f the cosine term, each summation component has zero mean. Thus, its variance conditioned on (3 becomes a ^ - = £ [ ( 7 ^ ) ] 2 n ; which can be expressed as Y cj ~ E (n) r- K-\L -l W ( k= 1 /=0 LV (A2) p„P/'{fti(\i)+Ai(y}'cos((p„ ) f- I I J (k) From (A2), assuming same number of multipaths for every user (i.e., L (3 are considered as constant, we have = L) and since P and 1 n K-\L-\ < # J / = £- I J,{h} E[W ]E[{b^R ,(xJ 2 b^ 2 k + (A3) k=\l = 0 2 (k) (k) Since 7i[cos (cp )] = 1 / 2 and by assuming that b_{ and b n/ 0 are i.i.d. random variables with equal probability taking values from the alphabet { ± 1 } , we have E[{b^R (x ) kl nl + b R (x )} ] k) 0 2 kl nl = E[R (x ) 2 kl n[ R ,(x )] 2 + k nl 2 = f- \RUx) + R (x) dx. 1 T kl 0 108 ( A 4 ) Appendix A. Derivation of Multiple-Access Interference Variance for the Gaussian Approximation The above integral can be evaluated by using the results in [31] to give (A5) 3N 3 •r {N) K where N-l r'kiW = J^iCKl-W 1=0 + + Cud-WCud-N+D cui) + c (i)c (i ki ki + + CKl-N+l) (A6) i) + cl(i+i)}. Substituting (A5) into (A3), we obtain the expression shown in (4.17): K-XL-l *MA1 = f - f k i iPn> • EiW ! 2 2 •^ = 1 /= 0 k= 1 3 • r (N) • K 1 (A7) 1=0 10 110 Appendix B. Evaluation of the Mean of Attenuation Factor Appendix B. Evaluation of the Mean of Attenuation Factor The evaluation of the mean of attenuation factor p is based on the cell model illustrated in F i g . q 5.2. In order to simplfy the mathematical analysis, the cells are assumed to be circular and uniform in size. To reiterate, in (5.16) we have denned rd \y p, = ( ^ J l O ^ - W . . 0 In (B1), d q (Bl) /10 is the distance between the reference mobile to the gth base station, which is a random "variable dependent on the reference mobile location, y is the path loss exponent, and C, is a q Gaussian random variable with zero mean and standard variance a . The mobile location is a function of both the distance and the angle with respect to the reference base station as shown in F i g . 5.2. It is assumed that the reference mobile are equally likely to be located anywhere within the reference c e l l . L e t the random distance between the mobile and reference base station be d Q = R. Then, the P D F of R for a circular cell is a function directly proportional to the circle circumference and can be written as p (r) R = kr, 0<r<9t (B2) f ,5K where k is a constant and is the cell radius. B y definition, p (r)dr R = 1, thus we have 'o P (r) R = ^ 2 (B3) The P D F of the angle 6 is a uniform distribution over [0,2TC ), i.e. P(Q) = ^ . Since the attenuation factor p q is composed of random variables r and 9 , we have, (B4) Appendix B. Evaluation of the Mean of Attenuation Factor 111 2KR E[p ] = J jp(r)p(Q)^yE[10^-^ ]drdQ (B5) /l0 q 0 0 which, after substitutions of p(r) and p ( 6 ) , becomes 271/? E [ ^ ] ^ J l ( f > = [ 1 O ( C o o ° " g / 1 O ] ^ 0 (B6) - F r o m [53], we have £[10 ( ? °-^ ) / I 0 ] = exp | glnlO^ 10 2 1-2 V2G 10 In 10 (B7) B y applying cosine law to the geometry of cell configuration illustrated in F i g . 5.2, we obtain d q for the 6 first-tier cells, the 6 farther second-tier cells, and the 6 closer second-tier cells, respectively, as follow: d = J(j3R) + d + 2V3/W cose 2 2 q l 0 for. 1 < q < 6 d = ,J(3R) + d^ + 6Rd cosQ . for 7 < q < 12 2 q J 0 = J(2j3R) + d^ + 4j3Rd cosQ for 13 < ^ < 18 . 2 0 (B8) (B9) (BIO) Substituting (B7)-(B10) to (B6), and using y = 4 , O" = 8 d B for the integration, we obtain the numerical results as presented i n Table 5.1.
- Library Home /
- Search Collections /
- Open Collections /
- Browse Collections /
- UBC Theses and Dissertations /
- Efficient video transmission over correlated nakagami...
Open Collections
UBC Theses and Dissertations
Featured Collection
UBC Theses and Dissertations
Efficient video transmission over correlated nakagami fading channels for IS-95 CDMA systems Chan, Norman 1999
pdf
Page Metadata
Item Metadata
Title | Efficient video transmission over correlated nakagami fading channels for IS-95 CDMA systems |
Creator |
Chan, Norman |
Date Issued | 1999 |
Description | This thesis deals with the problem of efficient transmission of video signals over generalized fading channels in Direct Sequence-Spread Spectrum (DS-SS) Code Division Multiple Access (CDMA) systems. The video codec is based upon the ITU H.263 compression algorithm which targets at providing low bit-rate video telephony services suitable for wireless transmission. In order to reduce the overall impact of errors due to mobile channels on the video sequence, a modified version of the original H.263 codec is proposed incorporating a selective Forward Error Correction (FEC) coding scheme combined with a forced INTRA frame update mechanism. This modified version of H.263 codec provides improvement in both average video and frame-to-frame performance. We further consider a coherent DS - CDMA system based upon the IS-95 standard for the forward link (base-to-mobile) in both single-cell and multiple-cell environment. We provide performance evaluation results by both analysis, employing the Gaussian approximation, and computer simulations, using Monte Carlo error counting techniques. The proposed CDMA system uses concatenated Walsh/maximal-length coding scheme for spectrum spreading. The resulting spread codes maintain orthogonality while reducing inhomogenous cross-correlations among Walsh sequences. The frequency-selective fading channel is modeled by a tapped delay line model with channel coefficients of each path following an independent Nakagami-m distribution. We have implemented in software a correlated Nakagami fading simulator based upon the principle of superposition of complex partial waves with a (possibly strong) component resulting from the direct waves. The time correlation is generated by assigning each partial wave a Doppler shift as a function of time. This approach is an exact replica of the actual physical situation as it reproduces the wave propagation process, as opposed to the Doppler spectrum approximation approach used in other simulators. The received signal is demodulated coherently using a RAKE receiver with variable resolving fingers, where multipath components are maximal-ratio combined'. For our analysis, we assume perfect knowledge of the channel, which could be accomplished either by the usage of pilot tone or some type of channel-parameter estimation circuits. However, for the computer simulation, such perfect knowledge of the channel is not necessary. In terms of performance evaluation results, we first present the improved performance of the modified H.263 codec as a function of Peak Signal-to-Noise Ratio (PSNR) transmitted in additive white Gaussian noise (AWGN) environment. Then, the analytical and computer simulated results for the bit error rate (BER) performance of CDMA forward link in Nakagami fading channels for both single-cell and multiple-cell environment are presented. Further, we present the PSNR performance results for the video transmission featuring the modified H.263 coding scheme over the proposed CDMA systems. Finally, a variety of performance evaluation results, both in single-cell and multiple-cell environment, are presented for different number of resolving paths, signal propagation characteristics, cell user capacity, as well as for the presence of channel estimation errors. In all cases, heuristic explanations and interpretations of the trend of the obtained results are also given. |
Extent | 5601812 bytes |
Genre |
Thesis/Dissertation |
Type |
Text |
File Format | application/pdf |
Language | eng |
Date Available | 2009-06-16 |
Provider | Vancouver : University of British Columbia Library |
Rights | For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use. |
DOI | 10.14288/1.0065286 |
URI | http://hdl.handle.net/2429/9359 |
Degree |
Master of Applied Science - MASc |
Program |
Electrical and Computer Engineering |
Affiliation |
Applied Science, Faculty of Electrical and Computer Engineering, Department of |
Degree Grantor | University of British Columbia |
Graduation Date | 1999-05 |
Campus |
UBCV |
Scholarly Level | Graduate |
Aggregated Source Repository | DSpace |
Download
- Media
- 831-ubc_1999-0180.pdf [ 5.34MB ]
- Metadata
- JSON: 831-1.0065286.json
- JSON-LD: 831-1.0065286-ld.json
- RDF/XML (Pretty): 831-1.0065286-rdf.xml
- RDF/JSON: 831-1.0065286-rdf.json
- Turtle: 831-1.0065286-turtle.txt
- N-Triples: 831-1.0065286-rdf-ntriples.txt
- Original Record: 831-1.0065286-source.json
- Full Text
- 831-1.0065286-fulltext.txt
- Citation
- 831-1.0065286.ris
Full Text
Cite
Citation Scheme:
Usage Statistics
Share
Embed
Customize your widget with the following options, then copy and paste the code below into the HTML
of your page to embed this item in your website.
<div id="ubcOpenCollectionsWidgetDisplay">
<script id="ubcOpenCollectionsWidget"
src="{[{embed.src}]}"
data-item="{[{embed.item}]}"
data-collection="{[{embed.collection}]}"
data-metadata="{[{embed.showMetadata}]}"
data-width="{[{embed.width}]}"
async >
</script>
</div>
Our image viewer uses the IIIF 2.0 standard.
To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0065286/manifest