UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Design of a simple reading machine for the blind Ramsay , William Desmond 1968

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-UBC_1969_A7 R35.pdf [ 5.59MB ]
Metadata
JSON: 831-1.0104050.json
JSON-LD: 831-1.0104050-ld.json
RDF/XML (Pretty): 831-1.0104050-rdf.xml
RDF/JSON: 831-1.0104050-rdf.json
Turtle: 831-1.0104050-turtle.txt
N-Triples: 831-1.0104050-rdf-ntriples.txt
Original Record: 831-1.0104050-source.json
Full Text
831-1.0104050-fulltext.txt
Citation
831-1.0104050.ris

Full Text

DESIGN OF A SIMPLE READING MACHINE FOR THE BLIND by WILLIAM DESMOND RAMSAY B. Eng., Carleton U n i v e r s i t y , 1966 A .THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF APPLIED SCIENCE i n the Department of E l e c t r i c a l Engineering We accept t h i s t h e s i s as conforming to the required standard Research Supervisor Members of Commitee Head of Department Members of the Department of E l e c t r i c a l Engineering THE UNIVERSITY OF BRITISH COLUMBIA October, 1968 In p r e s e n t i n g t h i s t h e s i s i n p a r t i a l f u l f i l m e n t o f t h e r e q u i r e m e n t s f o r an a d v a n c e d d e g r e e a t t h e U n i v e r s i t y o f B r i t i s h C o l u m b i a , I a g r e e t h a t t h e L i b r a r y s h a l l make i t f r e e l y a v a i l a b l e f o r r e f e r e n c e and S t u d y . I f u r t h e r a g r e e t h a t p e r m i s s i o n f o r e x t e n s i v e c o p y i n g o f t h i s t h e s i s f o r s c h o l a r l y p u r p o s e s may be g r a n t e d b y t h e Head o f my De p a r t m e n t o r by h i s r e p r e s e n t a t i v e s . It i s u n d e r s t o o d t h a t c o p y i n g o r p u b l i c a t i o n o f t h i s t h e s i s f o r f i n a n c i a l g a i n s h a l l n o t be a l l o w e d w i t h o u t my w r i t t e n p e r m i s s i o n . D e p a r t m e n t o f <t I fc T/?SC/}^ £w 6/A/f^/VM£~ -The U n i v e r s i t y o f B r i t i s h C o l u m b i a V a n c o u v e r 8, Canada D a t e M u / i T ABSTRACT A compact reading machine ("Lexiphone") has been designed and constructed to convert p r i n t e d l e t t e r s i n t o a p a t t e r n of sounds. The machine reads by d i r e c t t r a n s l a t i o n of v e r t i c a l s e c tions of the l e t t e r s , according to a r e c e n t l y developed code. In t h i s code, the "melody" produced i s independent of the v e r t i c a l p o s i t i o n of the reading head ; however the user i s given an i n d i c a t i o n (mean p i t c h ) of the v e r t i c a l p o s i t i o n t o - f a c i l i t a t e t r a c k i n g along a l i n e of p r i n t . , The d i s c r e t e nature of the d i r e c t t r a n s l a t i o n process l i m i t s the t h e o r e t i c a l l y p o s s i b l e reading r a t e s . Tests with a r t i -f i c i a l l y generated codes were performed to i n v e s t i g a t e t h i s l i m i t , and i t i s expected that the l i m i t w i l l be above that f o r Morse Code— 60 to 70 words per minute. This would be adequate f o r p r a c t i c a l use. Tests performed at Haskins Laboratories p r e d i c t e d s i m i l a r performance f o r other machines, such as the o p t o p h o n e . However, p r a c t i c a l users of the " B a t t e l l e Optophone", the most r e f i n e d v e r s i o n of the optophone, a t t a i n e d only 25 words per minute (on Grade I reading m a t e r i a l ) a f t e r an extensive course . I t i s suggested that t h i s was due to the d i f f i c u l t y i n the e a r l i e r machines of producing repeatable versions of the code. Code sounds from the present Lexiphone prototype were found to be very consistent and repeatable, and should allow the predicted reading rates to be approached. P r a c t i c a l reading r e s u l t s w i t h the machine are presented. At the time of w r i t i n g , a subject t r a i n i n g with the machine i s reading two-page passages of Grade I I I m a t e r i a l at .30 words per minute, and her. performance i s s t i l l improving. TABLE OF CONTENTS Page ABSTRACT i i TABLE OF CONTENTS i i i . LIST OF ILLUSTRATIONS v LIST OF TABLES . . ." v i i ACKNOWLEDGEMENT . v i i i 1. INTRODUCTION . . 1 1.2 The Problem 1 . 1.2.1 H i s t o r y ' 2 1.2.2 Reading Machine Codes...' 5 1.2.2.1 General Considerations 5 1.2.2.2 The Lexiphone Code 7 2. THE HARDWARE OF THE LEXIPHONE 9 2.1 Organization of the Machine 9 2.2 The Input Section 10 2.2.1 The Array 11 2.2.2 The Schmitt Triggers • . .. 14 2.2 .3 The Mechanical Page Scanner 17 2.2.4 The Imaging System 17 2.2.5 Page I l l u m i n a t i o n 18 2 . 3 The D i g i t a l Section 19 2 .3.1 P r i n c i p l e of f-Computation 19 2 .3.2 D i g i t a l Section Organization 21 2 . 3 . 3 D i g i t a l Section C i r c u i t D e t a i l s 26 2 . 3 . 3.1 Clock and Timer : 28 2 . 3 . 3.2 Scanning C i r c u i t 29 i i i 2 . 3 . 3 . 3 Edge and "Count By" 30 2 . 3 . 3 . 4 Counting Control 31 2 . 3 . 3 . 5 f-Counter 31 2 . 3 . 3 . 6 The Output Buffers 33 2 . 3 . 3 . 7 E r r o r P r o t e c t i o n 3 3 ' 2 . 3 . 3 . 8 Top Edge C i r c u i t 33 2 . 3 . 3 . 9 I n t e n s i t y Modulation 36 2 . 3 . 3.10 D/A Converter '. 38 2 . 4 The Analog Section . 38 2 . 4 . 1 . Exponential Function Generator 4 0 2 . 4 . 2 Voltage C o n t r o l l e d Generator 4 0 2 . 4 . 3 Amplitude Modulation.. ... 4 0 3 . PERFORMANCE CHECKS .' 41 3 . 1 Machine Tests 41 3 . 1 . 1 Upper Outline 44 . 3 . 2 Code Tests 47 3 . 3 Reading Results with the Machine.. 54 3 . 4 Continuing Work 55 4 . SUMMARY AND CONCLUSIONS, • . 60 4 . 1 Remaining Questions 61 APPENDIX 63 REFERENCES 64 i v LIST OF ILLUSTRATIONS Figure • . Page 1- 1 Production of the Lexiphone Code 7 2- 1 Block Diagram of the Lexiphone 10 2-2 E f f e c t s of Alignment Change 12 2-3 Testing with Prepared Films 13 2-4 Type Face Measurement 14 2-5 Lexiphone Prototype... 16 2-6 Events During a T y p i c a l Scan 20 2-7 F u n c t i o n a l Diagram of the D i g i t a l Section 22 2-8 Timer Outputs. . 23 2-9 Schmitt Triggers .. 15 2-10 Clock and Timer 22 2-11 Scanning C i r c u i t 28 2-12 Edge and "Count By".' ' 30 2-13 Counting C o n t r o l 32 2-14 f-Counter 32 2-15 Conditions f o r Loading Output Buf f e r 33 2-16 Production of Top Edge Pulse 35 2-17 T y p i c a l f-Function f o r the " I " 35 2-18 Black Pulse Counting 36 2-19 D/A Converter 37 2- 20 Analog Section 39 3- 1 Raw S i g n a l from a Photodiode 42 3-2 E f f e c t of the Schmitt Triggers 42 3-3 P r i n t Viewed by the Lexiphone 43 3-4 Top Edge Recognition 45 v 3-5 Spurious Pulses..... 45 3-6 Unavoidable " E r r o r s " 46 3-7 Tests of the f-Computation 48 3-8 Measured f-Function - Test #1 49 3-9 Measured f-Function - Test #2 49 3-10 Measured f-Function - Test #3 50 3-11 Sample of P r a c t i c a l f-Function 50 3-12 Comparison of Codes 53 3-13 .Reading Results with the Lexiphone 56 3-14 Bat t e l l e Reading Speeds 57 vi LIST OF TABLES Table Page 3-1 T r a n s l i t e r a t i o n s to Produce Wuhzi 52 3-2 Wuhzi Equivalents of Words i n the Eight Word Test , 52 v i i ACKNOWLEDGMENT Gr a t e f u l acknowledgment i s made to the many people who a s s i s t e d during the course of the p r o j e c t . P a r t i c u l a r thanks i s due to my supervisor, Dr. M. P. Beddoes, f o r h i s encouragement and advice. I would.like to thank Professor F. K. Bowers f o r h i s c a r e f u l reading and c o r r e c t i o n of the manuscript. -B l i n d subjects Peggy Spencer and Linda Jentsch' deserve r e c o g n i t i o n f o r the many hours they spent working with the L e x i -phone. A l s o , thanks go to the volunteers from the department who subjected themselves to the Lexiphone and other code sounds. I would also l i k e to thank Peter Dewdney, W i l l i s Martin and Hans Paul f o r proofreading the t h e s i s , Grant Murray f o r h i s work b u i l d i n g equipment, and Avis Hopkins f o r the ty p i n g . v i i i 1. INTRODUCTION This t h e s i s deals with the design, c o n s t r u c t i o n and t e s t i n g of the Lexiphone, a r e l a t i v e l y simple reading machine f o r the b l i n d . The machine reads the p r i n t e d page and produces a musi-c a l code as output. The musical code was proposed by Beddoes and tested on computer s i m u l a t i o n . The present machine produces a repeatable v e r s i o n of t h i s code and has been quite dependable. The t h e s i s a l s o includes r e s u l t s of code t e s t s that were performed i n an attempt to e x p l a i n the f a i l i n g s of previous machines. 1.2 The Problem The problem of enabling the b l i n d to read the standard p r i n t e d page i s a very o l d one, but i s s t i l l e s s e n t i a l l y unsolved. The b l i n d person who wants to read the l o c a l news, notes from f r i e n d s , b u l l e t i n s , or any of much s i m i l a r m a t e r i a l not yet t r a n s l a t e d i n t o b r a i l l e has no choice other than to use a sighted reader. This i s a tremendous inconvenience, and the l a c k of independence and pr i v a c y i s very d i s t a s t e f u l to many b l i n d people. Any r e a d i l y a v a i l a b l e device that would allow even the reading of personal correspondence would be a great improvement indeed. Almost since i t was f i r s t t e c h n i c a l l y p o s s i b l e , attempts have been made by various workers to make such a simple machine. Although many machines have been constructed, reading with'them has been such a tedious and p a i n f u l l y slow process that none has yet gained acceptance as a u s e f u l a i d . The 'Lexiphone belongs to the group known as " d i r e c t t r a n s -l a t i o n machines". A " d i r e c t t r a n s l a t i o n machine" considers only the 2 patterns of black and white p r e s e n t l y under i t s scanner, and using some transformation, produces a sound that represents what i t sees. The d i r e c t t r a n s l a t i o n machine thus produces an i n s t a n t - b y - i n s t a n t t r a n s l a t i o n of p r i n t patterns i n t o sound. I t i s then the task of the user to f i n d out which sound patterns correspond to p a r t i c u l a r l e t t e r s and words, the same way as a sighted reader must l e a r n which patterns of ink correspond to given l e t t e r s and words. Such a machine can obviously be used f o r any reading task. The problem i s whether or not the sounds i t makes can be e a s i l y and q u i c k l y i n t e r p r e t e d . Due to t h e i r p o t e n t i a l low cost and p o r t a b i l i t y , d i r e c t t r a n s l a t i o n machines have been the ones most widely considered i n study of the problem of p r o v i d i n g a personal machine. Previous machines have been s u f f i c i e n t l y l i m i t e d by technology that designers concentrated on making a sound from i n k , hoping that i t would produce a s u i t a b l e code when p r i n t was scanned. In the Lexiphone, the p h i l -osophy, has been that one should f e e l f r e e to f i n d a good code, con-f i d e n t that i t can be instrumented. Another problem t y p i c a l of older machines was poor r e s o l u t i o n which caused the code to be very depen-dent on alignment. This r e s u l t e d i n machines that were d i f f i c u l t to operate and i n code sounds that lacked r e p e a t a b i l i t y . Some of the older machines are described b r i e f l y i n the next s e c t i o n . 1.2.1 H i s t o r y • (2) A complete h i s t o r i c a l review has been given by Caple . The p r i n c i p a l machines b u i l t to date are- as f o l l o w s : 1914: White Sounding Optophone - The work of Fournier d'Albe, i t had .eight pulsed areas of l i g h t on the page and gave a tone 3 (3) when any area was r e f l e c t e d to the c e l l . 1920: Black Sounding Optophone - Barr and Stroud, a B r i t i s h f i r m , put a f i v e tone optophone i n a bridge arrangement so that the sound was heard only when p r i n t was under the scanner. Bridge balance was poor. 1928: Visagraph - R.E. Naumburg proposed a machine that embossed a (12) r a i s e d v e r s i o n of the l e t t e r on.aluminum-foil. I t turned out to be harder to read than Optophone code. During World War I I , the U.S. Veterans A d m i n i s t r a t i o n formed the Committee on Sensory Devices. They commissioned Raskins Laboratories and RCA to look i n t o the problem of reading machines. 1944-49: RCA A-2 Reader - A swept audio o s c i l l a t o r coupled to a swept spot of l i g h t and keyed on when the spot was over (13) p r i n t . Produces an optophone-like sound but has some noise due to keying. RCA Recognition Machine - Output from a f l y i n g - s p o t scanner was modulated depending on the t o t a l amount of l i g h t r e f l e c t e d from the page as the l i g h t s l i t was passed over i t . ^ Sound was turned o f f when no l e t t e r was present. Reading was very slow. (In a d d i t i o n to the work on the PM scan machine, the Haskins Laboratories evaluated other e x i s t i n g machines and tested d i f f e r e n t codes. This i s discussed l a t e r . ) 1952: The Argyle Reader - Argyle of B r i t i s h Columbia made a machine which had a spot o f l i g h t scanning the l e t t e r v e r t i c a l l y at 200 Hz, and the r e f l e c t e d l i g h t was picked up by a c e l l and 4 a m p l i f i e d . The r e s u l t was an e r r a t i c buzz at harmonics of 200 Hz that could be read no b e t t e r than the optophone. 1957 onward: Since 1957, the B a t t e l l e Memorial I n s t i t u t e , under the auspices of the Veterans A d m i n i s t r a t i o n , modernised the o r i g -i n a l optophone, using a p h o t o c e l l f o r each tone r a t h e r than (8) the pulsed areas of l i g h t . They b u i l t s e v e r a l of these improved machines and developed extensive t r a i n i n g programs. In s p i t e of t h i s , the optophone has not been s u c c e s s f u l . Visotoner, V i s o t a c t o r , Cognodictor - Mauch Laboratories i n Ohio have developed optophone-like machines with audible out-put (Visotoner) and with t a c t i l e output ( V i s o t a c t o r ) . ^ They have a l s o done some work on producing s p e l l e d speech output (18) (Cognodictor). The Cognodictor i s c u r r e n t l y s t i l l i n the experimental s t a t e . Such a machine might provide an output which would be simple to l e a r n . T a c t i l e Machine - L i n v i l l and B l i s s have produced a machine which gives a t a c t i l e d i s p l a y of the l e t t e r on an 8 x 5 array of v i b r a t i n g r e e d sZ 1 5 ^ 1 6 ^ Note that of these machines, those producing an audible code output are s t i l l s i m i l a r to the o r i g i n a l optophone i n at l e a s t two important aspects. They produce g e n e r a l l y a c h o r d - l i k e output, and are notably l a c k i n g i n r e s o l u t i o n . I t i s suggested that i n these f a c t o r s can be found the p r i n c i p a l reasons f o r l a c k of success of the e a r l y machines. Although a l l the machines made i t p o s s i b l e to decipher a word e v e n t u a l l y , with enough t r a i n i n g and enough time, no machine was s u f f i c i e n t l y easy to use. In the Lexiphone, some weaknesses of previous machines have been removed. I t i s hoped that t h i s work w i l l help answer the question of whether the d i r e c t t r a n s l a t i o n machine deserves f u r t h e r c o n s i d e r a t i o n . 1.2.2 Reading Machine Code The machine described i n t h i s t h e s i s has been designed to produce the Lexiphone code as proposed by B e d d o e s . T h e code has (2) been tested i n computer simulations at MIT and a l s o here at UBC. I t d i f f e r s from l e s s s o p h i s t i c a t e d codes i n some s i g n i f i c a n t respect The melody produced i s independent of the v e r t i c a l alignment of the scanner and c o n s i s t s of only one note at a time. Chords are never played. In a d d i t i o n , there i s amplitude modulation determined by a separate c r i t e r i o n . 1.2.2.1 General Considerations Several workers, notably Nye, Cooper, and Zahl, have examined i n d e t a i l the problem of a u r a l outputs.(4)(6) ^ e conclu-s i o n seems to be that no a r t i f i c i a l l e t t e r - b y - l e t t e r code has any hope of competing with the n a t u r a l speech code. However, there seems to be no fundamental reason why the r a t e of reading of such a code should not be f a s t enough to make i t of use. Cooper and Zahl decided that the l i m i t i n g f a c t o r was the blending of the code elements i n t o a b l u r r or a buzz, and used the r e c e p t i o n of I n t e r n a t i o n a l Morse code as a case i n p o i n t . Much the (2) same idea was expressed by Caple when he r e l a t e d the reading speed to the maximum channel capacity of the subject i n b i t s / s e c . In speech, the r a t e s vary from about 120 wpm, t y p i c a l of r a d i o announcing, to between 150 and 200 wpm f o r normal conversation 6 In the case of I n t e r n a t i o n a l Morse, the only encoding i s i n ampli-tude (on-off keying). A l e t t e r has from one to four elements, the average value being three elements per l e t t e r . I f we assume that the speed at which elements s t a r t to blend i n t o a buzz i s 20 Hz, i t gives a reading rate of 60-70 wpm as an absolute maximum f o r the Morse. Speeds of t h i s order have been recorded i n e x c e p t i o n a l l y g i f t e d and experienced operators. Commercial operators handle code comfortably and dependably at speeds of about 35 wpm. The DOT requires a speed of 15 wpm f o r the Advanced Amateur l i c e n c e . Thus the f i r s t plateau i n reading speed, a t t a i n e d a f t e r a f a i r l y r igorous formal course, i s about 20% of the absolute maximum, and the u s e f u l r a t e f o r experienced operators is'about one h a l f the maximum value. Cooper and Z a h l ^ ^ extended t h i s experience to reading machine codes. They conclude that there are about the same number of elements per l e t t e r i n the optophone code as i n the Morse, and as a r e s u l t assume that we cannot expect much more by way of p e r f o r -mance wit h the optophone than with the Morse code. The Lexiphone code i s modulated i n both amplitude and frequency. The frequency i s modulated with a composite upper o u t l i n e of the l e t t e r s , and the amplitude i s modulated by making the sound louder when the scanning s l i t i s over a v e r t i c a l l i n e , or " r i s e r " . In constrast to the Morse where only two elements, the dot and the dash, are used to code the l e t t e r s , the Lexiphone code has a host of frequencies to choose from, and, i n a d d i t i o n , has amplitude modulation. The d i f f e r e n c e s in. modu-l a t i n g schemes should produce very d i f f e r e n t problems i n a s s i m i l a t i o n . Results of some t e s t s w i t h t h i s and other codes are presented l a t e r i n the t h e s i s . When l e t t e r s are presented slowly i n Lexiphone code, a wealth of d e t a i l i s perceived. This d e t a i l may be used by an 7 operator i n i d e n t i f y i n g an i s o l a t e d l e t t e r . The l e t t e r s are, however, o v e r s p e c i f i e d . Thus, when heard at a f a s t e r r a t e , c e r t a i n word i s l i s t e n e d to at high speed, what i s perceived i s a character-i s t i c melody due to the dominant high and low frequencies i n i t s component l e t t e r s , and a c h a r a c t e r i s t i c rhythm due to the r i s e r thumps which occur only i n c e r t a i n l e t t e r s . A s t r i k i n g demonstration of the low number of p s y c h o l o g i c a l l y perceived elements i n a word i s the f a c t that a user f i n d s that he himself s t a r t s to mimic Lexiphone coded words. I t i s due to t h i s formation of "gestalten" ( s i n g l e , whole images formed from d i s c r e t e elements) that decoding can proceed at a r a t e above the Cooper-Zahl buzz, frequency.(9)(4) 12.2.2 The Lexiphone Code Transformation mean values may be s u f f i c i e n t f o r i d e n t i f i c a t i o n . (1) When an e n t i r e composite upper o u t l i n e c a l c u l a t e d from: f = y + ZAy y. m (Locus of bottom Datum, _of scanning s l i t ) F i g . 1-1 Production of the Lexiphone Code As mentioned above, the Lexiphone code i s frequency modu-l a t e d by a composite upper o u t l i n e of the l e t t e r s and amplitude modulated by.the presence of a v e r t i c a l l i n e or " r i s e r " (e.g., "1", 8 "d", "h", "k", e t c . ) . The a c t u a l transformation, proposed by Beddoes,^"^ i s shown i n F i g . 1-1. The frequency values range from 200 Hz to _ a maximum of 3000 Hz. The amplitude can have only two values, loud or s o f t . The composite upper o u t l i n e , or f - f u n c t i o n , i s determined only by the p o s i t i o n of the upper edges of the p r i n t i n f ormation. The transformation i s thus independent of l e t t e r t h i c k n e s s . To form the f - f u n c t i o n , the distance from the bottom of the scanning s l i t to the highest upper edge (y ) i s added to the sum of the distances down to the other upper edges from the topmost o u t l i n e . (Ay's i n F i g . 1-1) The f - f u n c t i o n at any point i n the l e t t e r represents the musical note that w i l l be heard'when the scanner i s at that point i n the l e t t e r . The r e l a t i o n of output frequency to f - f u n c t i o n has been arranged (Section 3«4) so that i f t h e . l e t t e r i s s h i f t e d up (or down) i n the scanning s l i t , i t j u s t moves the f-func-t i o n up or down on the musical s t a f f , and e x a c t l y the same melody i s heard, but i n a d i f f e r e n t key. The changing key provides an immediate clue to the alignment of the scanner, while the constant melody means that comprehension i s not a f f e c t e d . By merely l i s t e n i n g to the key of the melody of the p r i n t , the operator knows immediately i f he i s d r i f t i n g up or down on the l i n e . B l i n d subjects using the present Lexiphone prototype depend on t h i s p i t c h clue to determine the cor-r e c t v e r t i c a l alignment. The i n t e n s i t y modulation due to the r i s e r s provides the rhythmic p a t t e r n , which, i n conjunction with mean l e t t e r frequencies, are h e l p f u l i n formation of g e s t a l t e n , the key. to r a p i d a s s i m i l a t i o n of code sounds. 9 2. THE HARDWARE OF THE LEXIPHONE 2.1 Organization of the Machine The o r g a n i z a t i o n of the machine can be seen from the block diagram of F i g . 2-1. The image of the p r i n t i s focused on to the p h o t o c e l l array. This gives r i s e to 54 analog c u r r e n t s , each vary-i n g i n accordance w i t h the instantaneous i l l u m i n a t i o n of that par-t i c u l a r c e l l . The analog photocurrents are then passed i n t o Schmitt t r i g g e r s which give the d i g i t a l s i g n a l s f o r the next s e c t i o n . The heart of the Lexiphone i s the block l a b e l l e d "f-com-puter". This u n i t takes the s i g n a l s from the Schmitt t r i g g e r s and c a l c u l a t e s the f - f u n c t i o n according to the Beddoes transformation ( F i g . l - l ) . ^ ~ ^ P a r a l l e l to the f-computer we have the r i s e r detector which decides when the array i s over a v e r t i c a l r i s e r and produces the required amplitude modulating s i g n a l . The D/A converter puts the f - f u n c t i o n i n t o analog form convenient f o r modulating the audio generator. The frequency modulating s i g n a l i s produced by passing the analog f - f u n c t i o n through an exponential element. The exponential t r a n s f e r f u n c t i o n assures that f i x e d voltage i n t e r v a l s of the f-func-t i o n produce f i x e d musical i n t e r v a l s i n the audible output. The amplitude modulation i s three valued, i . e . , OFF when no l e t t e r s are seen, LOUD when a r i s e r i s i n d i c a t e d , and SOFT otherwise. A more d e t a i l e d d e s c r i p t i o n of each s e c t i o n f o l l o w s . 10 54 l i n e s p h o t o c e l l array sohml tt triggers i l l u m i n a t o r - page INPUT SECTION 54 l i n e s f d i g i t a l T " f " - f u n c t i o n computer 8 l i n e s X r i s e r detector D/A DIGITAL SECTION f - f u n c t i o n . (analog) amp. mod. s i g n a l exponential element freq. mod. s i g n a l ^ audio generator sound r i / . ] \ °"riv*+ ANALOG - O C P J j output S E C T I 0 N P i g . 2-1 Block'Diagram of the Lexiphone. 2.2 The Input Section One of the biggest problems i n instrumentation of a usable reading machine i s t r a n s l a t i o n of a normal p r i n t e d page i n t o a s i g n a l s u i t a b l e f o r processing by a cede generating u n i t . A method used i n previous reading machines, and the one chosen f o r t h i s prototype of the Lexiphone, i s to focus the image of the p r i n t to be read on to an array of p h o t o c e l l s . While simple i n p r i n c i p l e , t h i s poses many p r a c t i c a l problems. 11 2.2.1 The Array The array used i n the machine i s the F a i r c h i l d FPA-500 l i n e a r i n t e g r a t e d photodiode array. I t has two columns of photo-diodes, each column having 63 diodes on 12 m i l centres. F i f t y - f o u r c e l l s of one column are used i n the Lexiphone. Each diode i s 35 m i l i n e f f e c t i v e area, with a 4 x 6 m i l region of uniform l i g h t s e n s i t i v i t y . An elemental diode of the array has a dark current (at room temperature) of l e s s than 1 na, and the photocurrent r i s e s l i n e a r l y w ith i r r a d i a t i o n to about 750 na at 30 mw/cm . Each c e l l of the array was measured by F a i r c h i l d before 2 d e l i v e r y at 5 mw/cm i r r a d i a t i o n * and gave a t y p i c a l photocurrent of 150 na. A c t u a l l y , the l i g h t l e v e l s required f o r t h i s amount of i r r a -d i a t i o n are r a t h e r i m p r a c t i c a l , and i t was necessary to operate the array i n the Lexiphone at much lower photocurrents. Previous machines, had used only a small number of c e l l s , ( t y p i c a l l y 10) and, as a r e s u l t , r e s o l u t i o n was not good and the code was very dependent on alignment. In the case of the optophone, there was a good reason f o r not going to a l a r g e r number of c e l l s . The optophone code c o n s i s t s of a tone f o r each c e l l , and becomes pro-h i b i t i v e l y confusing as the number of c e l l s i s made l a r g e . The Lexiphone code, on the other hand, depends only on the measurements of the l e t t e r i t s e l f , and the purpose of the scanner i s to get an accurate measurement of the l e t t e r . The p i c t u r e s of Figure 2-2 * As the photodiode array i s quite s e n s i t i v e to wavelengths i n the long v i s i b l e and i n f r a r e d regions of the spectrum, i t i s more s u i t -able to measure the s e n s i t i v i t y i n terms of the t o t a l i r r a d i a t i o n (measured i n , say, mw/cm^) than the i l l u m i n a t i o n , which i s v i s i b l e only. For a given incandescent source, the f r a c t i o n of the t o t a l power r a d i a t e d that f a l l s i n the v i s i b l e i s a f u n c t i o n of the colour temperature. For example, one- mw/cm2 of i r r a d i a t i o n from an u r i f i l -t ered tungsten source operating at 2870°K c o n s t i t u t e s about 20 f t . -candles of i l l u m i n a t i o n . 12 G (a) G G • B 54 c e l l (c) • o 6 -o c e l l 0 - o o F i g . 2-2 E f f e c t s of alignment change on high and low reso-l u t i o n scanners, (a) Source m a t e r i a l of 3 alignments. (b) Picked up by high r e s o l u t i o n scanner. (c) Picked up by low r e s o l u t i o n scanner. 13 provide a simple demonstration of one of the b e n e f i t s of a high r e s o l u t i o n scanner. They show "G" as seen by the present 54 c e l l scanner,' and by an array of only 6 c e l l s t y p i c a l of e a r l i e r machines. As we see, the low r e s o l u t i o n scanner sees quite a d i f f e r e n t l e t t e r when i t s v e r t i c a l alignment i s changed (The f a c t that i t does not r e a l l y look l i k e a "G" would not matter i f i t could be made to come out the same f o r a l l alignments). The nature of the 54 c e l l array has another advantage i n connection with the s i g n a l from an i n d i v i d u a l c e l l . An i n d i v i d u a l c e l l of a high r e s o l u t i o n array covers a much smaller area of the l e t t e r and as a r e s u l t gives a sharper t r a n s i t i o n when moving from a l i g h t to a dark region. Later i n the t h e s i s i s shown the waveform from a c e l l , taken as the machine scans a row of v e r t i c a l l i n e s . The w e l l defined white-to-black dip i s very evident ( F i g . 3-1)• To reduce the number of v a r i a b l e s i n the t e s t i n g of pre-l i m i n a r y c i r c u i t r y , t e s t m a t e r i a l was prepared on high contrast 35 mm f i l m and p u l l e d d i r e c t l y across the surface of the array ( F i g . 2-3) This method provides a nearly i d e a l o p t i c a l s i g n a l to the array. I t a l s o puts a l l the important f a c t o r s f o r t e s t i n g , such as i n t e n s i t y of i l l u m i n a t i o n , s i z e , p o s i t i o n , and s t y l e of type . completely under c o n t r o l . FPA-500 p r i n t on f i l m (35 mm) crank to v a r i a b l e supply F i g . 2-3 Testing with prepared f i l m s , The s i z e of a given type face i s measured from the top of a r i s e r to the bottom of a descender as depicted i n F i g . 2-4. This measurement i s g e n e r a l l y given i n " p o i n t s " , where 72 points f o r most of the t e s t m a t e r i a l measured 0.125" (9 point), and required a m a g n i f i c a t i o n of four to f i l l 80^ of the FPA-500*s 640 m i l a c t i v e area. The m a g n i f i c a t i o n should never be such that the type face f i l l s the e n t i r e array, as t h i s would leave no l a t i t u d e i n alignment. Another array has been purchased f o r use i n future machines. I t i s the FPA-203, which has only 63 diodes on 4 m i l centres. -It has only one quarter the s e n s i t i v i t y of the FPA-500, but the 4 m i l centres w i l l reduce m a g n i f i c a t i o n requirements by a f a c t o r of three, which more than makes up f o r the decrease i n s e n s i t i v i t y . equals approximately one i n c h . (17) The type on the t y p e w r i t e r used s i z e type descender "1 ine" of p r i n t F i g . 2-4 Type Size i s measured i n " p o i n t s " from the top of r i s e r to bottom of descender. (72 points = .9962") 2.2.2 The Schmitt Triggers ( F i g . 2-9) E s s e n t i a l l y , the f u n c t i o n of the r e s t of the machine i s to take the raw s i g n a l s from the p h o t o c e l l array and produce the musical 15 code that -would "be a r r i v e d at by measurement of the p r i n t . The Schmitt t r i g g e r s accept the c e l l outputs and produce a l o g i c a l 1 (ground) on c e l l s that have l e s s than some c e r t a i n value of l i g h t current, chosen to represent the highest value l i k e l y to be obtained when, the _ c e l l .is. over i n k . There are many instances, say j u s t on an edge or over a t h i n or hazy l i n e , that the current can be i n an indeterminate•region between l i g h t and dark. The Schmitt t r i g g e r s e l i m i n a t e t h i s u n c e r t a i n t y . ( F i g . 3-1 and 3-2). F i g . 2-9 Schmitt Triggers - Since-the -photodiode i s such a high-impedance device, an FET input has been used i n the t r i g g e r c i r c u i t . This u n i t has a hys-t e r e s i s of 0.2v, which i s about l / 5 of•the swing from black ink to the white page. An i n d i v i d u a l s e t t i n g pot has been included on each t r i g g e r . This compensates f o r v a r i a t i o n s i n FET c h a r a c t e r i s t i c s , ' and 2-5 (a) (b) present prototype compact e l e c t r o n i c s f o r f u r t h e r machines 17 a l s o f o r non-uniformities i n page i l l u m i n a t i o n . 2.2.3 The Mechanical Page Scanner ( F i g . 2-5) In t h i s prototype, instead of moving, the p h o t o c e l l s over the page, the page i s moved under the p h o t o c e l l s . -The same mechan-(2) ism was used by Caple i n h i s Lexiphone. Two m o d i f i c a t i o n s were made: the a d d i t i o n of the adjustment screws f o r alignment of the t a b l e during setup, and the a d d i t i o n of an i n s t a n t - s t o p l e v e r so the operator may c o n t r o l the t a b l e himself. This method i s , of course, r e l a t i v e l y u n s a t i s f a c t o r y f o r a f u n c t i o n a l reading machine, but i t does serve as a convenient t e s t device. The a c t u a l product would probably use a hand-held reading probe such as the type used by (5)(18) Mauch i n h i s Visotoner and V i s o t a c t o r . This probe i s moved across the page by the operator and w i l l give him complete c o n t r o l of reading speed and alignment. A system employing f i b e r o p t i c s has been proposed that may s i m p l i f y the reading head (see Section 4 - l ) . 2.2.4 The Imaging System At present, a camera lens (Elgeet Cine N a v i t a r 12 mm, F1.2) i s used, to produce the desired image. Such a lens i s f a r too bulky to i n c l u d e i n a hand-held probe. I t was used merely because i t i s v e r s a t i l e and was a v a i l a b l e f o r use at the time. The lens chosen should have a short f o c a l length to keep the p h y s i c a l s i z e of the system down. I t should have as wide an aperture as p r a c t i c a l (the FI.2 of t h i s lens i s q u i t e good) to reduce i l l u m i n a t i o n requirements. One immediately apparent drawback of a wide lens opening was •the very shallow depth of f i e l d . Even i r r e g u l a r i t i e s i n the page caused a l o s s of focus and i t became necessary to put a glass p l a t e 18 over the t e s t paper. A hand-held probe would provide d i r e c t con-t a c t w i t h the page and focus would not be a problem. 2.2.5 Page I l l u m i n a t i o n As discussed i n Section 2.2.4, the d i s t i n c t disadvantage of a h i g h - r e s o l u t i o n scanner i s the small amount of l i g h t that any one c e l l of the array i s able to gather. To get u s e f u l amounts of photocurrent from the array, very high i n t e n s i t i e s are required on the page. At present, i l l u m i n a t i o n i s supplied by a C a r l Zeiss mic-roscope i l l u m i n a t o r with the d i f f u s i o n screen removed so that the s p e c i a l 15 watt bulb's filament i s focused on the page. This par-t i c u l a r bulb (6v 15w Zeiss #38 0177) has a f l a t wound filament that focuses n i c e l y i n t o a b r i g h t rectangle (about 0.2" x 0.4") which i s convenient f o r i l l u m i n a t i n g a l e t t e r . This p a r t i c u l a r arrangement provides an i l l u m i n a t i o n on the page of about 12,000 foot .candles, which i s more than b r i g h t s u n l i g h t , but only a small part of the l i g h t s h i n i n g on the page makes i t s way to .the photodiode array. Of the l i g h t s c a t t e r e d by the page, only a small amount w i l l actual3y be gathered by the l e n s , and t h i s i s f u r t h e r reduced by the m a g n i f i c a t i o n . I = k (Pm) 2 I (3.1) page array where P i s the lens opening (Fl.2 here), m the d e s i r e d m a g n i f i c a t i o n , and k a constant determined by the r e f l e c t i v i t y of the page and the manner i n which the l i g h t i s s cattered by the page. With the above i n mind, i t i s not s u r p r i s i n g that the pho-tocurrent f r o n i a diode of the array i s only 60 na over a white page, 2 i n d i c a t i n g an i r r a d i a t i o n of some 2.5 mw/cm (about 50 f t . candles with the lamp used). Even though t h i s value of l i g h t i s i n the lower 19 end of the array's u s e f u l region, i t i s not f e l t that i t would be p r a c t i c a l to attempt to boost the i l l u m i n a t i o n on the page any f u r -t h e r , a t l e a s t c e r t a i n l y not by i n c r e a s i n g the s i z e of the lamp. Since the array only "looks a t " a t h i n v e r t i c a l s e c t i o n , the e f f e c -t i v e i l l u m i n a t i o n might be increased by using, a c y l i n d r i c a l l e n s , to focus the l i g h t i n t o a l i n e . With the present machine such mea-sures were not warranted. 2.3 The D i g i t a l Section The d i g i t a l s e c t i o n must take the s i g n a l s from the Schmitt t r i g g e r s of the input s e c t i o n and produce the f - f u n c t i o n and r i s e r s i g n a l s needed by the f i n a l analog s e c t i o n . The d i g i t a l s e c t i o n c o n s t i t u t e s the bulk of the machine. 2.3•1 P r i n c i p l e of f-Computation As noted i n the s e c t i o n d e s c r i b i n g the code, the frequency of the audible output i s : co = k e f ' (3.2) where .f i s a f u n c t i o n c a l c u l a t e d from: f - y m + H Ay (3.3) During operation of the machine, the c e l l s of the array are scanned from top to bottom, each completed scan producing a value f o r the . f - f u n c t i o n . The frequency of top to bottom scanning i s approximately 10 KHz, y i e l d i n g a v i r t u a l l y continuously changing f - f u n c t i o n as•the array i s moved across a l e t t e r . During any one scan, each c e l l i s i n t e r r o g a t e d i n turn u n t i l the f i r s t white to black t r a n s i t i o n (top edge) i s encountered. Each f u r t h e r i n t e r r o -20 BOTTOM p o s i t i o n of scanning s l i t count by EVENT A-B-D-E-Sta r t of scan. "Count by" r e g i s t e r contains 4 - the edge count plus one. No-counting t a k e s place. 1st edge encountered. "Count by" i s reduced by one, b r i n g -i n g i t to the correct value of three. Counting by three commences. 2nd edge encountered. "Count by" reduced to two. Counting continues by two. 3rd.edge encountered. Counting continues by one. End of scan. Check i s made that counting i s by one and the value of count s h i f t e d i n t o output b u f f e r . M-l f = Jl Ay. i = l where M = number of black areas i n scanning s l i t s f i e l d . P i g . 2-6 Events During a T y p i c a l Scan 21 ga t i o n , from t h i s time u n t i l the next top edge i s detected, adds to a counter a number N, equal to the number of top edges p r e s e n t l y under the s l i t (N was determined by the previous scan). Prom the second to the t h i r d top edges, N - l i s added to the counter f o r each c e l l sampled. This sequence continues to the bottom of the array. At the bottom, a check i s made to see i f the counting i s now pro-ceeding by one. I f i t i s , i t means that the number of top edges, N (borrowed from the previous scan), was, i n f a c t , the co r r e c t value to use f o r t h i s scan. This being the case, the t o t a l count w i l l be the d e s i r e d f, and may be s h i f t e d out. That t h i s procedure a c t u a l l y does produce the f u n c t i o n f = y m + Z A y " ( 3 . 3 ) can be seen from the example of Fig. 2-6. From B to C, each c e l l spacing counts three u n i t s , from C to D,two u n i t s , and from D to the end (E), one u n i t . This forms the same t o t a l as adding the three vectors y^, Ay and Ay 2. 2 .3-2 D i g i t a l Section Organization Reference to the block diagram of F i g . 2-7 shows how the above scheme has a c t u a l l y been r e a l i z e d using l o g i c c i r c u i t r y . The machine operates on a 64-step c y c l e , during which time the 54 c e l l s are scanned and. the f - f u n c t i o n c a l c u l a t e d . The t i m i n g of a cycle i s d e t a i l e d i n Figures 2-6 and 2-8. The f u n c t i o n i s accumulated i n the f-counter, which i s simply an 8-bit up counter which can be toggled on the 4's, 2's, o r l ' s b i t . Once enabled, the counting con-t r o l c i r c u i t d i r e c t s pulses to the appropriate inputs depending on the contents of the "count by" r e g i s t e r . For example, i f the machine To CRT clock timer RAMP 3 to one Load mge Counter < > GATES "COUNT BY" r u r t r u I — COUNTING 1 1 CONTROL -n c l e a r . P 2 =&> l's Count ( E p) Loud enable > 8 f-Counter X X MUTING PP 8 l i n e s Output b u f f e r :LP) =I E.L+E.L 8 l i n e s D/A •9 O o 1 ON EP LP £ Muting i n t e n s i t y ANALOG SECTION > D O 1 ) S o u n d output F i g . 2-7 Fu n c t i o n a l Diagram of the D i g i t a l Section 23 i s counting by three, a pulse i s f i r s t applied to the 2's and then to the l ' s b i t during each clock period. On F i g . 2-7 i t i s noticed that the edge counter i s set to 1 at the s t a r t of each c y c l e , r a t h e r than being cleared to zero. The reason f o r t h i s i s that the f i r s t top edge encountered i n a scan should s t a r t the counting by the number of top edges under the s l i t . However, since each top edge pulse a l s o reduces the "count by" by one, i t i s necessary to have the i n i t i a l content one higher than the desired s t a r t i n g count. This has been accomplished by having each scan s t a r t with the edge counter already at one. The master timer c o n t r o l s the sequence of a l l operations of the machine. I t generates three c o n t r o l pulses, P-^ , , P^, and an analog voltage p r o p o r t i o n a l to the cycle step number. I t i s i n turn c o n t r o l l e d by the clock. H — I — I 1 1—I L -0 1 2 3 4 5 6.7 4—\ h H 1 — l h ; i I I I I >-'t 60 63 • .1-msec 0 1 2 3 4 5 6 7 — i — i — i — I I I 0 1 2 3 4 5 6 7 c e l l 1 _array scanned 60 63 c e l l 54 l i i l >-£ 60 63 0 1 2 3 4 5 £ 7 8 9 RAMP 60 63 F i g . 2-8 Timer Outputs 24 The L e t t e r Present (LP) f l i p - f l o p i s set by the black pulse, P^, and hence i n d i c a t e s the presence of black under the scanner. The Edge Present (EP) f l i p - f l o p i s set by the top edge pulse, P^, and hence i s 1 when a white-to-black t r a n s i t i o n (top edge) has been detected. The muting FF i s used to s i l e n c e the output of the machine during the space between l e t t e r s (The l e t t e r present may not be used f o r t h i s purpose d i r e c t l y , as i t i s reset at the s t a r t of each scan). The sequence' of operations i s as f o l l o w s : P^: -Loads output b u f f e r with contents of the f-counter i f the l e t t e r present and edge present are i n the same s t a t e and i f the "count by" i s at one. -Loads muting FF from the LP. -Stores output of a . c i r c u i t counting the number of c e l l s black (used f o r r i s e r d e t e c t i o n ) . ?2: -Clears the f-counter to zero. -Resets EP FF ( t h i s d i s a b l e s the counting). -Resets LP FF. -Resets the "black count" FF. -Loads edge count i n t o the "count by" r e g i s t e r . P^: -Loads i n t e n s i t y FF. -Sets edge count to one. - S t a r t s new scan. A feature not yet discussed i s the i n t e n s i t y FF. I t i s used to make the i n t e n s i t y of the code output greater when only one edge i s present and more than eight c e l l s are i n d i c a t i n g black. At 25 time P^, the "count by" i s loaded with the contents of the edge counter which w i l l be one greater than the number of top edges under the s l i t . At time P^, then, the c i r c u i t looks at the n=2 l i n e i n the counting c o n t r o l and i f i t i s HI, i n d i c a t i n g only one edge under the s l i t , then the i n t e n s i t y PP i s set to make the sound loud. I f , i n a d d i t i o n , the output of the black counting c i r c u i t i s at ground, i n d i c a t i n g more than eight c e l l s black,'the sound i s i n t e n -s i f i e d . The time from P^ to the end of the scan i s occupied by the Black Pulses, P^, produced when c e l l s over black are sampled, and the top edge pulses, P^ _, produced by the required white-black sequence i n P, . b P^: -Sets the L e t t e r Present FF. -Removes the blanking l e v e l from the CRT. - A l l P^ go to the top edge d e c i s i o n c i r c u i t where.the sequence i s examined and P^ produced at the .right time. P^: -Sets the Edge Present FF ( t h i s enables counting). -Reduces "count by" by one. -Adds one to the edge count. I t was stated before that there are c e r t a i n c o nditions which must be met before the value of " f " i n the counter can be t r a n s -f e r r e d to the output b u f f e r and hence used f o r the code. The pro-v i s o that the edge count must be equal to one has already been d i s -cussed and i s due to the f a c t that the top edge count used i s the one from the previous scan, and we must protect against a wrong value when the number of top edges changes. In some cases, very t h i n l i n e s , 26 or l i g h t parts of p r i n t , a c o n d i t i o n may e x i s t where black i s seen by the scanner, but no top edge i s i n d i c a t e d . In t h i s case the t r a n s f e r of " f " w i l l be i n h i b i t e d by another c i r c u i t , as the value i s obviously wrong. The c i r c u i t would a l s o protect us i n case of a P^ without a P^, although t h i s w i l l never occur i n p r a c t i c e . On good p r i n t , these precautions produce v i r t u a l l y no change i n the sound of the f i n a l output. Before the machine was i n i t s f i n a l form, the muting was done by l o o k i n g at the contents of the output b u f f e r , and the output was muted when t h i s was zero. In the case of a top edge dropout (P^ with no P^) t h i s caused the output to be muted completely during the dropout. The r e s u l t was a very raspy sound on p r i n t where such dropouts were a problem. With the muting done by l o o k i n g f o r i n k , the output i s not muted i n the case of a dropout, and the problem i s not so acute. Nevertheless, the p r o t e c t i o n i s very simple and has been l e f t i n . S i m i l a r l y , the e r r o r due to a change i n top edge number i s s l i g h t and occurs only on that scan i n which the number changes, but, again, p r o t e c t i o n i s so simple i t has been l e f t i n . The analog s e c t i o n has been shown as a box i n Figure 2-7. I t i s d e t a i l e d l a t e r . I t produces the desired audio output from the f - f u n c t i o n , the muting, and the i n t e n s i t y l e v e l s . 2-3-3 D i g i t a l Section C i r c u i t D e t a i l s Motorola MC700P s e r i e s i n t e g r a t e d c i r c u i t s have been used e x t e n s i v e l y i n the d i g i t a l s e c t i o n . Some F a i r c h i l d 900 s e r i e s were used i n the e a r l y stages of the machine but supply became d i f f i c u l t , and machines p r e s e n t l y being b u i l t use Motorola throughout. 27 clock -2>-rUT_ clock jTj-ir to the desired combination on gates g i v i n g P-j_, 72, P3 A A A A > 0 A 1 -1-6 T — , 1 4: T 2R< 2R R AA/V-•2R 2R 2R' 2R; R R AAAr R A/vV R •AAAr RAMP R = 1.2K 2R d e t a i l 3.3K j> ><J10K (1) (4) <-> 4> (7) & p. .» p, 3 * p. F i g . -'2-10 Clock and Timer 28 2.3-3.1 Clock and Timer ( F i g . 2-10) The clock i s a free-running m u l t i v i b r a t o r which gives square wave output at about 600 KHz. The ti m i n g i s done by the 6-bit up counter, g i v i n g a 64-step c y c l e . The counter was conveni-ent f o r t h i s purpose, as the D/A ladder could be connected to give s h i f t l i n e s (5 groups of s h i f t ( 5 x l l ) 0_ o_ to corresponding gates 5 > * JT_ P B * #1 FF presets to '"ONE" others to "ZERO" F i g . 2-11 Scanning C i r c u i t 29 an output voltage p r o p o r t i o n a l to the step i n the c y c l e . With t h i s analog ramp on the v e r t i c a l d e f l e c t i o n of a 'scope and the black pulse connected to modulate the beam i n t e n s i t y , we can get a good d i s p l a y of what i s being seen by the scanner. This turned out to be a very valuable a i d indeed i n s e t t i n g up the machine. I f the a r -ray i s not e x a c t l y v e r t i c a l , or i f there i s poor focus or uneven l i g h t i n g , t h i s i s e a s i l y detected by l o o k i n g at the d i s p l a y of the p r i n t on the 'scope. In a f i n a l .model of the machine, t h i s feature would probably be omitted. (Figure 3-3 shows an example of the above 'scope d i s p l a y . ) 2.3.3.2 Scanning C i r c u i t The purpose of the scanning c i r c u i t i s to sample each c e l l of the array i n t u r n and to produce a pulse on the c e l l s that are i n d i c a t i n g black. I t i s e s s e n t i a l l y a s h i f t r e g i s t e r and a set of gates connected to the 54 l i n e s from the input s e c t i o n . The scanner i s d e t a i l e d i n Figure 2-11. On P^, the scan i s begun by s t a r t i n g a "one" out at the s t a r t of the s h i f t r e g i s t e r . The ground enable l e v e l i s thus app l i e d to each gate i n turn and every time the cor-responding t r i g g e r i s i n d i c a t i n g a black, a pulse i s produced at the gate output. A l l the gate outputs are fed to a diode OR c i r c u i t . The t r i g g e r l e v e l s of.the Schmitt c i r c u i t s i n the input s e c t i o n are adjusted as f o l l o w s : a white page i s examined, but w i t h reduced i l l u m i n a t i o n l e v e l . The potentiometers are then set so that each of the c i r c u i t s j u s t i n d i c a t e s black. This gives a P^ s i g n a l that i s an accurate reproduction of the black areas passing under the array. The 'scope d i s p l a y s obtained from the P^ s i g n a l s (see t e s t sec-t i o n , Figure 3-3) a t t e s t to the r e s o l u t i o n of the array and the 30 cleanness of operation of t h e ' t r i g g e r s . This i s an important point i n the present work, as i t i s f e l t that the f a i l u r e of previous attempts' to make d i r e c t t r a n s l a t i o n machines has been, i n l a r g e p a r t , due to the poor r e s o l u t i o n and poor r e p e a t a b i l i t y of the scanning arrangement. 2.3•3•3 Edge Counter and "Count By" Re g i s t e r This uses the c i r c u i t of Figure 2-12.- ' P^ sets the. edge count to one; P^ adds one to the count each time; t r a n s f e r s the value of the edge count to the count by r e g i s t e r ; and P^ al s o dec-reases the "count by" by one each time. s e t 3- — UP COUNTER DOWN COUNTER i o I o , o " _ "COUNT BY" OUTPUT F i g . 2-12 Edge and "Count By" 31 '2.3-3-4 Counting Control (Figure 2-13) The counting c o n t r o l produces the appropriate pulses to. the 4's, 2's, and l ' s b i t s of the f-counter as the "count by" i s changed. The s t r i n g of monostables produces a sequence of. three pulses during each clock period and they are d i r e c t e d to the appro-p r i a t e counter i n p u t s . I t i s important that the pulses be s e q u e n t i a l so that any carry i n the f-counter w i l l have a chance to propagate before the next b i t i s added. The four gates decode the "count by" and the diode matrix then d i s a b l e s the appropriate counting b i t s . For example, i f the count i s to be "by two" the n=2 l i n e from the gates w i l l - be HI, and the diodes w i l l apply d i s a b l e l e v e l s to the l ' s and 4's b i t s . The counting c o n t r o l a l s o provides the necessary n=l and n=2 outputs. 2-3-3-5 f-Counter (Figure 2-14) The f-counter, as was mentioned before, i s an 8-bit binary up-counter so wired that i t can be toggled on the 4's, 2's, or l ' s b i t . C a p a c i t i v e coupling i s used so that the l e v e l on a FF output does not d i s a b l e the gate used to OR the carry with the e x t e r n a l i n -put on that b i t . In t h i s c i r c u i t , any carry w i l l propagate immedi-a t e l y , but there i s a recovery time, set by the period of the mono-s t a b l e s , during which time no f u r t h e r carry can propagate. I t i s therefore important that the recovery time of the f-counter be l e s s than the time between successive pulses from the counting c o n t r o l . In the present machine, the clock period i s 1800 ns. The len g t h of each monostable i n the counting c o n t r o l has been adjusted to approx-imately 400 ns, and the pulse length of the coupling monostables i n the f-counter i s 100 ns. 32 Prom "Count By" > '—^ " X X • X, X > X X -o 4's 2's l ' s Count Enable (E.P.) Disables 4 — . to f-counter 4's F i g . 2-13 Counting Contro'l i t : 2's b i t 1 Clear to output b u f f e r b i t 8 F i g . 2-14 f-Counter 33 '2.3-3.6 The Output Buf f e r The output b u f f e r stores the value of the f - f u n c t i o n during a scan u n t i l the new value i s ready. I t i s composed of 8 JK f l i p -f l o p s connected to the f-counter and loaded by p u l s i n g t h e i r toggles. 2.3-3-7 E r r o r P r o t e c t i o n The p r o t e c t i o n against l o a d i n g the output b u f f e r i f n / l or' i f the LP and EP are not i n the same s t a t e , i s shown i n Figure 2.15. This w i l l prevent the l o a d i n g of an erroneous value of f i n the case of a top edge dropout or when the number of top edges changes. F i g . 2-15 Conditions f o r Loading Output Buf f e r 2.3-3.8 Top Edge C i r c u i t The top edge c i r c u i t deserves a s p e c i a l d i s c u s s i o n , as i t i s i n t h i s region, as w e l l as i n the area of d e c i d i n g which c r i t e r i o n to use f o r the i n t e n s i t y modulation, that a good deal of experimen-34 • t a t i o n and t e s t i n g was necessary. In the case of the top edge, the problem i s e s s e n t i a l l y thi s : , we must make a d e c i s i o n as to which sequence of black-white to produce a top edge pulse on. I f we decide j u s t one black i s a u t o m a t i c a l l y an edge, then we get each l i t t l e b i t of d i r t and a l s o get an edge i n d i c a t e d i f there i s a s o l i -t a r y white i n the middle of a huge i s l a n d of black. On the other hand, i f the d e c i s i o n i s made to look f o r a l a r g e number of black c e l l s , then we w i l l have trou b l e on l e g i t i m a t e h o r i z o n t a l l i n e s of t h i n p r i n t . A good general method, and the one that was e v e n t u a l l y decided upon, to produce a top edge pulse - i s to feed the P^'s i n t o a short b-bit s h i f t r e g i s t e r where we can examine the l a s t b c e l l s sampled and decide when a top edge has occurred (see Figure 2-16). The f i n a l model used b=5 and a code of 11000, which means we must have two white c e l l s and two black c e l l s , separated by a "don't care" c e l l . B l i n d subjects p r e s e n t l y using the machine f i n d the code quite consistent on a v a r i e t y of p r i n t s . When the array i s passing o f f the edge of a v e r t i c a l l i n e , there i s a s i t u t a t i o n i n which the s t a t e of the c e l l s i s more or l e s s random. This w i l l occur f o r an i n s t a n t on the sides of v e r t i c a l l i n e s no matter how c l e a r they are. I f i t were p o s s i b l e , the top edge code should a l s o be one that helps prevent a. huge number of top edges from being i n d i c a t e d when the c e l l s are i n such a c o n d i t i o n . A c t u a l l y , the d u r a t i o n of such bursts i s so short that i t i s not important to suppress them. Figure 2-17 shows the f - f u n c t i o n f o r the c a p i t a l " I " . 35 C l O G K — • j - i r u gates connected to give P T on desired P B sequence -¥ i I P i g . 2-16 Production of Top Edge Pulse l J L % P i g . 2.17 T y p i c a l f - f u n c t i o n f o r the " I " . Various features are worth noting here. Notice that the " I " was s l i g h t l y t i l t e d . The scanner f i r s t encountered the high top edge, then the two s e r i f s , next the r i s e r (same height as the f i r s t edge), then the two s e r i f s again and f i n a l l y the lower s e r i f . Notice that the e r r o r s i g n a l s as the scanner leaves the v e r t i c a l edge are of very short duration. 36 2.3.3-9 The I n t e n s i t y Modulation The main use of i n t e n s i t y modulation i s to provide a d d i -t i o n a l clues to the i d e n t i t y of a l e t t e r or word, other than those provided by the frequency modulation alone. Without i n t e n s i t y modu-l a t i o n , some of the l e t t e r s are j u s t about impossible to t e l l apart. For example, the lower case "e" and the lower case "s" have almost i d e n t i c a l f - f u n c t i o n s . "M" and "Y" are i d e n t i c a l i n f - f u n c t i o n . The sound i s made louder when the scanner i s over a v e r t i c a l l i n e , or " r i s e r " , such as the r i s e r i n the l e t t e r "d", or the three r i s e r s i n the l e t t e r "m". The machine i n d i c a t e s a r i s e r whenever only one black i s l a n d i s present under the scanning s l i t and more than eight c e l l s are black. The presence of a s o l i t a r y i s l a n d of black ( i n d i -cated i n the machine by only one top edge pulse) i s almost s u f f i c i e n t to define a r i s e r . The proviso that more than eight c e l l s be black must be added to prevent i n d i c a t i n g a " r i s e r " on a s i n g l e h o r i z o n t a l l i n e . - T U T . —c>r c . -r o c l e a r T 1 -• 0 r '~U f 1 * T 1 u T V o P 2 >8 black F i g . 2-18 Black Pulse Counting f o r R i s e r Detection 37 Since the i n t e n s i t y FP i s set by P^, the information as to whether or not there was enough black present to i n d i c a t e a r i s e r must be a v a i l a b l e at t h i s time. The method i s to set a FF when the count i n a 4 - b i t up counter passes 8 and t o store t h i s information f o r use i n the next scan. (Figure 2-18) +15 '2K 8K 16K •32K <>64K - H - l >128K 256K -44-(a) op amp input +3.6 from t \ 0-bit — [ > / V A/V-(+1.5, -1.5) R to diodes (c) R F i g . 2-19 D/A Converter, (a) Weighted currents are switched to the op e r a t i o n a l a m p l i f i e r by the diodes. (b) D e t a i l of a b i t switcher. R = 470 i n a l l but MSB where i t i s 220 so that the e n t i r e current can flow w i t h the a v a i l a b l e voltage d i f f e r e n c e . (c) D e t a i l of a r e s i s t o r of the D/A. 38 2.3-3.10 The D/A Converter ( P i g . 2-19) (7) The D/A converter i s the one described by George Austin. In the converter, each b i t of the 8-bit word can switch the appro-p r i a t e l y weighted current i n t o the input of the o p e r a t i o n a l ampli-f i e r . The a m p l i f i e r then sums a l l these currents and produces the word i n the output b u f f e r . In the Lexiphone, t h i s analog voltage i s the f - f u n c t i o n i t s e l f , which we must use to produce the sounds of the Lexiphone code. This work i s done by the analog s e c t i o n of the machine. 2.4 The Analog Section In a musical s c a l e , frequency r a t i o s , not frequency i n t e r v a l s , are important. For example, on the piano keyboard, a given spacing of keys produces the same musical i n t e r v a l (octave, f i f t h , t h i r d , e t c . ) , no matter where on the keyboard i t i s played. The Lexiphone i s to be a musical machine. I f we were to imagine the lowest value of the f - f u n c t i o n p o s s i b l e , as sounding the bottom key on the keyboard, then we would expect as the f - f u n c t i o n increased l i n e a r l y i n value, the note sounded would move l i n e a r l y up the key-board. Not i c e , however, what t h i s means i n terms of voltage to frequency t r a n s l a t i o n . I f each time the f u n c t i o n increased by 2 v o l t s , i t moved us up the keyboard one octave, then the frequency i s doubling f o r each 2 v o l t increase. This i s an exponential r e l a -t i o n . Regardless of the exact s c a l e chosen, the requirement f o r a musical r e l a t i o n i s the same - a c e r t a i n increment i n the f - f u n c t i o n must always produce the same r a t i o i n frequency. In the Lexiphone, the musical r e l a t i o n has been obtained by feeding the f - f u n c t i o n to to a c i r c u i t w ith an exponential t r a n s f e r f u n c t i o n and the output of t h i s to a simple l i n e a r voltage c o n t r o l l e d generator (VCG). D/A GAIN i rv> o P H O (ft 0 3 (D O c+ H-O -15 RANGE.-, (EXF.G. Gain) •T to VCG 40U J 27K 22K OUTPUT >1.5K .02-L -15 T ON 27KJ=r LOUD 1 -15 40 2.4-1 The Exponential Function Generator The exponential t r a n s f e r f u n c t i o n of t h i s c i r c u i t i s obtained from the current-voltage c h a r a c t e r i s t i c of a diode. The voltage i s fed to the diode, and the r e s u l t a n t , e x p o n e n t i a l l y r e l a t e d current is converted to an appropriate voltage by the oper-a t i o n a l a m p l i f i e r . . The exponential element i s provided mainly to give a p o s i t i o n independence to the code, and any t e s t of i t s q u a l -i t y w i l l be more q u a l i t a t i v e than q u a n t i t a t i v e . As such, the requirement i s not an exacting one. The current voltage charac-t e r i s t i c of a diode turned out to be more than adequate. 2.4.2 The Voltage C o n t r o l l e d Generator (VCG) The VCG i s r e a l l y a modified UJT r e l a x a t i o n o s c i l l a t o r . The c a p a c i t o r on the emitter of the UJT i s charged from a t r a n s i s -t o r a c t i n g as a constant current source of adjustable value. Since the peak point voltage (at which the ca p a c i t o r i s discharged), remains constant, the period of o s c i l l a t i o n w i l l be l i n e a r l y depen-dent on the charging current, and hence on the input voltage, A. FF i s used to produce square wave output. The pleasant oboe-like sound of the square wave i s pre f e r r e d by subjects. 2.4-3 Amplitude Modulation The output of the reading machine can be loud, s o f t , or o f f . Both the on-off and the i n t e n s i t y s w itching functions are performed by FET switches. 3. PERFORMANCE CHECKS 3.1 Machine Tests The f i r s t point of i n t e r e s t i n a check i s the a c t u a l photocurrent from a c e l l of the array. Figure 3-1 i s the raw photo-c e l l output as the machine scans a row of brackets. This t e s t was found to be very good f o r f o c u s i n g the machine. The Lexiphone i s set scanning the l i n e and the focus adjusted to give the deepest dips as brackets are passed over. Notice that even on the raw s i g -n a l , the white to black t r a n s i t i o n , i s very d e f i n i t e . Figure 3-2 i s the s i g n a l taken a f t e r the Schmitt t r i g g e r s . A l l the gray range has been eliminated and we have a d e f i n i t e white, or a d e f i n i t e black. I t i s these h i g h - q u a l i t y d i g i t a l s i g n a l s that are used to get the black pulse and f i n a l l y the f - f u n c t i o n . The waveforms before and a f t e r the Schmitts were taken w i t h the machine scanning at the rat e of a l i n e of 70 brackets i n 15 seconds. At 5 per word, t h i s would represent 50 wpm. The rat e was then speeded up to see i f slowness of the response was any prob-lem. L i t t l e or no d e t e r i o r a t i o n of the t r i g g e r output was no t i c e d u n t i l the e n t i r e page width was being scanned i n something l e s s than 0.5 second. Response time should be no l i m i t a t i o n . Figure 3-3 shows some standard t y p e w r i t t e n t e x t as i t i s picked up by the Lexiphone's p h o t o c e l l array. This sample came from a portable machine with standard c l o t h ribbon. The ramp output from the timer i s displayed v e r t i c a l l y on an o s c i l l o s c o p e , and the black pulse s i g n a l (P^) from the Lexiphone i s appl i e d to the z-axis to .modulate the beam i n t e n s i t y . Notice the extremely f i n e r e s o l u t i o n that i s c h a r a c t e r i s t i c of the 54 c e l l reading head. 4 2 P i g . 3-1 Raw s i g n a l from a photodiode. This i s taken from one c e l l as the array scans a row of "brackets ( ( ( ( ( ( . Signal i s about 50 na p-p, with a mini-mum of approximately 10 na. F i g . 3 - 2 E f f e c t of the Schmitt t r i g g e r s . Output of the t r i g g e r r e c e i v i n g the s i g n a l F i g . 3 - 1 . F i g . 3-3 (a) Magnified view of standard typewriter p r i n t . (b) Typewritten text as seen by the Lexiphone. The f i n e r e s o l u t i o n i s t y p i c a l of the 54 c e l l scanner. 44 3.1.1 Upper Outline The top edge d e c i s i o n c i r c u i t must produce a pulse when a l e g i t i m a t e white-to-black t r a n s i t i o n has occurrrd. Figures 3-4, 3-5, and 3-6 are waveforms taken with the P^ c i r c u i t set to produce a P^ when a 00011 sequence has occurred. When the page i s i n the dark, a l l the c e l l s are i n d i c a t i n g black (Figure 3-4) and a top edge pulse i s produced at the s t a r t of the black i s l a n d . This i s the simplest, most s t r a i g h t f o r w a r d case, and v i r t u a l l y any c r i t e r i o n would work here. Figure 3-5, top t r a c e , shows a P^ sequence much more t y p i -c a l of the problems encountered when a c t u a l l y reading a page. The s i n g l e c e l l black pulse second from the l e f t could e a s i l y be caused by a speck of dust, or d i r t from an erasure. In a d d i t i o n , we no t i c e a s o l i t a r y white c e l l i n the r i g h t hand i s l a n d of black. This could have been due to p r i n t imperfections. The P^ output i s shown below. The s o l i t a r y c e l l s have been ignored, and the corr e c t two top edge pulses have been produced. (P^ i s not produced u n t i l the 00011 sequence i s encountered and i s hence delayed two pulses from the edge i t i n d i c a t e s . ) Regardless of the c r i t e r i o n used to determine an upper o u t l i n e , there w i l l s t i l l be c e r t a i n unavoidable " e r r o r s " . When passing o f f the edge of a v e r t i c a l l i n e , a momentary burst of ran-domly spaced top edges occurs. The v e r t i c a l " l i n e " i s a c t u a l l y q u i t e i r r e g u l a r when examined c l o s e l y , and even i f i t were'perfect, e r r o r s would s t i l l occur when the c e l l s are h a l f over the edge. In t h i s s i t u a t i o n they are presented with grey, and due to inherent i n d i v i d u a l d i f f e r e n c e s , the Schmitts turn o f f i n a random order. F o r t u n a t e l y , t h i s i s of a very short d u r a t i o n , and i s not s i g n i f i c a n t F i g . 3-4 Top edge r e c o g n i t i o n . This p i c t u r e shows the simplest case. There i s a huge i s l a n d of black ana only one edge i s present. Note that the P-t i s not produced u n t i l two black c e l l s have been confirmed. F i g . 3-5 Spurious pulses and dropouts are ignored by the P-k c i r c u i t . The i s o l a t e d black c e l l and the i s o -l a t e d white c e l l seen i n the top trace are ignored and the correct two top edge pulses are produced. 46 P i g . 3-6 Unavoidable " e r r o r s " must s t i l l occur. The c e l l s are caught i n a random stat e i n the upper waveform. This leads to a random burst of top edges. ( t h i s c o n d i t i o n e x i s t s when the array i s passing o f f the side of a v e r t i c a l r i s e r ) 47 i n the output. To give an o v e r a l l check of the machine performance from page to f - f u n c t i o n , a set of t e s t s with one to four edges i n random arrangements was typed (Figure 3-7)• The t h e o r e t i c a l value of the f - f u n c t i o n i s e a s i l y determined from these s t r a i g h t l i n e patterns and can be compared to the values c a l c u l a t e d by the machine. The r e s u l t s are very good, and v e r i f y that the machine i s c a l c u l a t i n g the corre c t f - f u n c t i o n . In the d i s c u s s i o n on , an f - f u n c t i o n f o r the " I " was shown. The "a" of Figure 3-11 i s a f u r t h e r i l l u s t r a t i o n of . the type of f - f u n c t i o n a c t u a l l y obtained. 3.2 Code Tests The problem of f i n d i n g the best a u r a l output f o r a reading machine has been given much c o n s i d e r a t i o n . I t i s g e n e r a l l y agreed that the very nature of a d i r e c t t r a n s l a t i o n machine introduces l i m i t a t i o n s on the a t t a i n a b l e reading r a t e s . To give a general idea of how the Lexiphone code compares with other codes, some simple t e s t s were run. The t e s t i n g p a t t e r n used by Haskins l a b s ^ and N y e ^ to t e s t comparative code performances was used. A set of eight simple four l e t t e r words are expressed i n the code to be tested and pre- . sented i n a code-response-code-voice format. The standard of performance i s a speech-like code o r i g i -nated at Haskins Labs and nicknamed " W u h z i " . W u h z i i s produced by a human speaker. The formation of Wuhzi i s shown i n Table 3-1. The l e a r n i n g curve f o r Wuhzi forms a basis f o r comparison. I t i s i n t e r e s t i n g to note that the Wuhzi l e a r n i n g curve obtained here i s i d e n t i c a l to that of Nye.^^ This gives us at l e a s t some confidence that the conditions of the t e s t were d u p l i c a t e d . 48 F i g . 3-7 Tests cf the f-computing c i r c u i t r y . (a) as typed, (b) recorded through the reading head. These s t r a i g h t l i n e patterns of one, two, three, and four top edges t e s t the generation of the f - f u n c t i o n from the page to the end of the f-computer. The r i g h t answers are easy to c a l c u l a t e and can be q u i c k l y com-pared to the measured ones, (see the f o l l o w i n g p i c t u r e s) 49 F i g . 3-8 Measured f - f u n c t i o n from t e s t #1 F i g . 3-9 Measured f - f u n c t i o n from t e s t #2 50 3-10 Measured f - f u n c t i o n , t e s t #3. These three t e s t s provide a f i n a l t e s t of the operation of the bulk of the machine. The proper production of these f - f u n c t i o n s not only requires that the f-computer be f u n c t i o n i n g c o r r e c t l y but also those parts of the machine supplying inputs to the f-computer ( i l l u m i n a t i o n , scanning, focus, array, Schmitts). 3-11 Sample of p r a c t i c a l f - f u n c t i o n . This f - f u n c t i o n of "a" i s t y p i c a l of those obtained during use of the machine on p r i n t . 51 The source f o r the Lexiphone t e s t s was the machine i t s e l f . I t i s f e l t that t h i s point cannot be over emphasized, as the output of previous machines had been so unpredictable that i t was necessary to record a master tape of sounds and use i t as the source. In t h i s case, the words of the t e s t were simply typed on a sheet of paper and read by the Lexiphone d i r e c t l y on to the f i n a l t e s t tape. The Morse t e s t was made using i n t e r n a t i o n a l morse at 50 -wpm, the same speed at which the Lexiphone was run. The l i s t was made by sending on an e l e c t r o n i c key at 25 wpm,-. and running the tape at double speed. . This was necessary because 50 wpm i s a very high speed indeed to send Morse. The r e s u l t s of the t e s t s were more or l e s s as expected (Figure 3-12). The Wuhzi t e s t was i n s u r p r i s i n g l y close agreement wit h the curve of Nye et a l . The Lexiphone i s , not s u r p r i s i n g l y , somewhat below Wuhzi. The important point i s that the Morse i s markedly lower than the Lexiphone at t h i s speed of 50.wpm. This bears out the proposal put f o r t h e a r l i e r i n the t h e s i s that the Lexiphone code (at l e a s t at higher speeds) c e r t a i n l y does not have the same number of p s y c h o l o g i c a l l y perceived elements as Morse, and that decoding can, and does, proceed at a speed much above the Buzz Rate suggested by Cooper and Zahl. I t i s i n t e r e s t i n g to note the r e s u l t when two of the b l i n d subjects already using the machine were asked to t r y the Lexiphone t e s t . One missed a word on t h e ' f i r s t t r y , but a f t e r that the scores were p e r f e c t . The f a c t that the corr e c t answers could be c a l l e d o f f without h e s i t a t i o n by these expert g i r l s shows, at l e a s t , that s u f f i c i e n t information i s present f o r the l e a r n i n g curve to reach per f e c t score. In fact,, the Lexiphone curve •52 a - o (ton) (Write as uh) n - r b - j 0 - a (tap) c - th P - t d - g q - ch e - i ( t i p ) r - n f - s s - f g - d t - k h - w u - a (top) (write as ah) i - e (eat) V - z j - b w - y k - P X - zh (azure) 1 - m y - h m - 1 z - V Table 3 - 1 T r a n s l i t e r a t i o n s used to produce Wuhzi word wuhzi pronounced wi t h yekw yeekwuh w i l l yemm yeem were y i n i y i n i ( t i p ) from snal s n a l (tap) been j i i r j e e r have wozi wuhzi t h i s kwef kweef that kwok kwuck Table 3 - 2 Wuhzi equivalents of the words used i n the 8-word t e s t P e r f e c t Score A — Z T " X sr y. Wuhzi (NYE) A Wuhzi (UBC) * Lexiphone o Morse +. Optophone (Haskins) 4- 5" 6 T r i a l Number P i g . 3-12 Comparison of Code 54 was c a r r i e d beyond that shown on the graph and s e v e r a l p e r f e c t scores were recorded among group members, although the average was only up to 20 a f t e r 18 t r i a l s . Two important points emerge from these code t e s t s . F i r s t , by comparison with the Morse, we suggest that the Lexiphone code has fewer p s y c h o l o g i c a l l y perceived elements than had been assumed by other workers, and i s hence amenable to higher speed decoding. Sec-ondhand more important from the point of view of t h i s t h e s i s , a l e a r n i n g curve roughly the same as the best codes tested at Haskins Laboratories (Optophone curve) has been obtained with a set of t e s t s recorded d i r e c t l y from the machine. The consistency of the code sounds from pre s e n t a t i o n to p r e s e n t a t i o n i s here due to the dependable performance of the machine i t s e l f , and not to the c e r t a i n r e p e a t a b i l -i t y of hearing the same tape recorded sounds over and over. 3 . 3 Reading Results w i t h the Machine . During the design and c o n s t r u c t i o n of the machine, a group of subjects was t r a i n e d using a computer s i m u l a t i o n of the Lexiphone code. The s i m u l a t i o n was. quite approximate i n that i t was made with data from r e l a t i v e l y coarse manual measurement of the l e t t e r s . The s i m u l a t i o n had a quantized sound, and, as was discovered when the subjects were exposed to the r e a l time machine, c e r t a i n e r r o r s had been made i n the o r i g i n a l l e t t e r measurements. Nevertheless, the sounds were b a s i c a l l y the Lexiphone, and i t i s f e l t that t h e i r exper-ience with the s i m u l a t i o n prepared the subjects w e l l f o r the task of l e a r n i n g to read with the hardware machine. At the time of w r i t i n g , only two subjects have had a s i g -n i f i c a n t exposure to the machine. They are making very encouraging 55 progress (Figure 3 - 1 3 ) i n reading speed. ( c . f . Figure 3 - 1 4 ) By way of comparison, an extensive, 2 0 0 - l e s s o n " t r a i n i n g program was developed by the B a t t e l l e Memorial I n s t i t u t e f o r i n s t r u c t i o n i n the (8) use of the B a t t e l l e Optophone. At the end of f i s c a l year 1 9 6 3 , when the f i n a l report was given, they say of t h e i r top student i n the c l a s s of ' 6 3 : "On the t e s t of comparable f i r s t grade l e v e l ma-t e r i a l at the conclusion of t r a i n i n g , subject A read 2 1 . 9 wpm i n c l u d -(8) i n g l i n e change time, and 2 5 . 4 wpm excluding l i n e change time". The present Lexiphone i s reading grade I I I m a t e r i a l at 30'wpm and the r a t e s are s t i l l improving. 3 . 4 Continuing Work The best method of assessing the present design and of suggesting any refinements- ' i s t e s t i n g by b l i n d users. To t h i s end, work i s under way on 1 0 Lexiphone machines, which are "to be given out f o r e v a l u a t i o n . At the time of w r i t i n g , the e l e c t r o n i c sections of these machines are nearing completion. The new machines have a much higher parts d e n s i t y , and can be f i t t e d , e x c l u s i v e of the power supply i n t o l e s s than a 6" cube. Such compact u n i t s r e q u i r e great care i n the layout and w i r i n g . S t i l l to be constructed are s u i t a b l e o p t i c a l and mechanical systems f o r the op e r a t i o n a l machines. Such features as the a b i l i t y to t r a c k i n a s t r a i g h t l i n e and at a constant speed, and the a b i l i t y to change s c a l e , are extremely d e s i r a b l e . Such.a system must a l s o be capable of p r o v i d i n g strong i l l u m i n a t i o n on the page. In s p i t e of these many necessary f e a t u r e s , the reading head must be s m a l l . The problems of t r a c k i n g and v a r i a b l e s cale have been handled s u c c e s s f u l l y by e a r l i e r workers and i t i s p o s s i b l e to adopt proven techniques i n •Hours on Machine F i g . 3-13 Reading Progress with the Lexiphone Test Number P i g . 3.14 Best Two Subjects of B a t t e l l e FY 1963 E v a l u a t i o n C l a s s . Each t e s t represents 10 hours of i n s t r u c t i o n . 58 these areas. The Lexiphone, however, has p a r t i c u l a r l y demanding i l l u -mination requirements, which may be hard to meet while s t i l l main-t a i n i n g a small p h y s i c a l s i z e . One p o s s i b l e s o l u t i o n would be to i l l u m i n a t e the page v i a f i b e r o p t i c s . This would mean the h i g h - i n t e n s i t y l i g h t could be loc a t e d i n the f i x e d pasfe^of the machine. Losses, p r i m a r i l y i n g e t t i n g l i g h t i n t o a l i g h t guide, would increase the i n t e n s i t y of the lamp requ i r e d , but since i t would no longer be part of the reading head, t h i s may not be a serious problem. In c o n s i d e r i n g the f i b e r o p t i c s , the p o s s i b i l i t y of t r a n s -m i t t i n g the image i t s e l f through a coherent bundle a r i s e s . This would make p o s s i b l e a f u r t h e r r eduction i n the s i z e of the reading head i t s e l f . Requirements have been given to the f i b e r o p t i c s d i v i s i o n of Bausch and Lomb and they are preparing a device with a coherent inner bundle f o r image transmission and a non-coherent outer bundle f o r i l l u m i n a t i o n . I t i s suggested that page scanning be done with a hand-held probe guided by a mechanical t r a c k i n g a i d . This arrangement i s s m aller and more v e r s a t i l e than the present moving platform page scanner. In a d d i t i o n , i t enables the user to slow down e a s i l y over u n f a m i l i a r words, and to quicken the pace over f a m i l i a r t e x t . With such easy c o n t r o l , the beginning user can develop h i s own techniques of d i s t i n g u i s h i n g s i m i l a r l e t t e r s or decoding d i f f i c u l t words. When the e l e c t r o n i c s , page scanner, and o p t i c s are ready, people must be t r a i n e d i n the use of these machines. At present, work i s proceeding, i n c o l l a b o r a t i o n with the F a c u l t y of Education, on development of a s u i t a b l e t r a i n i n g program f o r users of the L e x i -phone. Prepared by such a course, they w i l l be i n a p o s i t i o n to 59 evaluate the merits and f a i l i n g s of the machine. To summarise, the f o l l o w i n g items must be taken care of to obtain the f i n a l r e s u l t s of the Lexiphone p r o j e c t : The present c i r c u i t must be packaged i n t o a compact u n i t . The f i r s t of ten such u n i t s i s now almost completed. A s u i t a b l e o p t i c a l system must be b u i l t . I t must be com-p a t i b l e with a reading head of small p h y s i c a l s i z e , and provide suf-f i c i e n t page i l l u m i n a t i o n . P r o v i s i o n f o r scale change ( v a r i a b l e mag-n i f i c a t i o n ) i s d e s i r a b l e . The p o s s i b i l i t y of using f i b e r o p t i c s i s being i n v e s t i g a t e d . A thorough, p r o f e s s i o n a l l y managed, t r a i n i n g program i s e s s e n t i a l to meaningful r e s u l t s . Such a program i s now being d e v e l -oped to t r a i n users of the 10 t e s t machines. F i n a l l y , the 10 machines may be d i s t r i b u t e d to t h e i r users, and the t r a i n i n g - e v a l u a t i o n program begun. The r e s u l t s of t h i s w i l l guide the course of future work. 60 4. SUMMARY AND CONCLUSIONS E a r l y reading machines met with such meagre success that many workers concluded that simple reading devices could never o f f e r s a t i s f a c t o r y reading r a t e s . This t h e s i s reviews the older machines and suggests t h a t , while there are fundamental l i m i t a t i o n s , the r e a l f a i l i n g of previous machines was i n l a c k of r e p e a t a b i l i t y and an over-dependence on alignment. I t i s f e l t that c r i t i c s of the d i r e c t t r a n s -l a t i o n machine have been discouraged by c a l c u l a t i n g t h e o r e t i c a l max-ima i n reading speed that were somewhat low, and by being overly demanding i n t h e i r estimates of an absolute minimum " u s e f u l " speed (12) (Cooper sets the absolute minimum at 50-60 wpm and the value f o r an average good user at 100 wpm). Beddoes^^ has proposed a code (Lexiphone code) which i s not dependent on alignment, a l l o w i n g a considerable l a t i t u d e i n t r a c k i n g , and which provides a constant clue (mean p i t c h ) to the c o r r e c t alignment. Tests w i t h computer simula-t i o n s showed that t h i s code could be learned at a r a t e f a s t enough to warrant b u i l d i n g a machine. This t h e s i s has described the design and c o n s t r u c t i o n of a machine to produce the Lexiphone code dependably • and repeatably. A high r e s o l u t i o n (54 c e l l ) scanner has been used to take f u l l advantage of the p o s i t i o n independence inherent i n the Lexiphone code. D i g i t a l c i r c u i t s have been used throughout the code generating u n i t , and a high degree of d e p e n d a b i l i t y , as w e l l as a p h y s i c a l compactness, has been attained.. The prototype machine has received extensive use over the past three months without r e q u i r i n g adjustment. Reading r e s u l t s with t h i s prototype (Section 3-3) have shown i t to be already markedly superior to the Optophone, and per-formance has been s t e a d i l y improving. 61 A speech-like code (Wuhzi), the Lexiphone code, and Morse code were subjected to an eight word screening t e s t used to compare code performances i n the p a s t . ( 4 0 ( 6 0 (Section 3.2). The Wuhzi curve corresponded e x a c t l y with the curves of other workers. (4) (6) ^ e Lexiphone was roughly the same as the Optophone (as tested at Haskins Labs), and the Morse was w e l l below both. By comparison w i t h the Morse, it'was suggested that at the speed of these t e s t s (50 wpm), the machine code has l e s s p s y c h o l o g i c a l l y perceived elements than Morse, and hence lends i t s e l f to more r a p i d a s s i m i l a t i o n . The close correspondence between the Lexiphone and Optophone l e a r n i n g curves served to confirm the suggestion put f o r t h e a r l i e r i n the t h e s i s that the true f a i l i n g of simple machines had not been i n the code per se, but r a t h e r i n the c a p a b i l i t y to produce the code with a r e a l machine manned by a b l i n d operator. The l e a r n i n g curves obtained with a r t i -f i c i a l l y repeated sounds would i n d i c a t e the Optophone could have been learned at 50 wpm, whereas t h i s value was not approached i n p r a c t i c e . I t i s concluded that the Lexiphone i s w e l l s u i t e d to r e a -l i z e the p o t e n t i a l speed of the code i t s e l f , and that t h i s value i s above the somewhat p e s s i m i s t i c values c a l c u l a t e d p r e v i o u s l y . 4-1 Remaining Questions Net reading speeds of 30 wpm are being obtained on Grade • I I I m a t e r i a l with the present Lexiphone. I t i s f e l t that i t i s not unreasonable to expect 40 wpm as a maximum. I t i s pointed out that such speeds are near the upper l i m i t f o r unprocessed s p e l l e d speech (2) (each l e t t e r spoken). Golden showed t h a t , by use of a time com-pressor, code sounds could be understood at a much f a s t e r r a t e than 62 the uncompressed v e r s i o n . He a l s o optimized the parameters f o r t h i s compression. The e f f e c t of compression on the Lexiphone code from the machine deserves f u r t h e r c o n s i d e r a t i o n . I f r e s u l t s s i m i l a r to Golden's are obtained, i t could make the Lexiphone-compressor u n i t comparable i n reading speed to the obvious next step i n c o m p l e x i t y — the character recogniser with a spelled-speech output. A prototype (Cognodictor) of such a machine i s being t r i e d by Mauch Labora-(18) tories . However, the t r a c k i n g tolerances are very small and reading rates only of the order of 25 wpm have been recorded. The advantage of the r e c o g n i t i o n machine i s , of course, that i s r e q u i r e s very l i t t l e p r a c t i c e to use, but i f s i m i l a r reading rates can be obtained with the much smaller, l e s s c r i t i c a l machine, i t would warrant spendin time l e a r n i n g the code. More research on code compression could make t h i s p o s s i b l e . We a l s o should ask i f the present Lexiphone code i s the best. Reading performance has shown the Lexiphone to be superior to the Optophone (Section 3 - 3 ) but i t i a not known p r e c i s e l y why. There i s room here f o r more i n v e s t i g a t i o n of the codes. T h e o r e t i c a l work on the two codes, and t h e o r e t i c a l p r e d i c t i o n s of reading r a t e s , may r e v e a l a b e t t e r code to use. Work towards a model of code per-ception would be very h e l p f u l i n determining the optimum codes f o r future machines. I t i s d e s i r a b l e that the reading machine be portable and of "reasonable" p r i c e . The present Lexiphone meets the small s i z e requirement, but component cost i s approximately $600 . While the primary c o n s i d e r a t i o n i s a machine that works w e l l , i t i s important to determine i f the present machine could be produced at a "reasonable p r i c e . 63 APPENDIX Logic Level Symbology •O Ground Level f o r A s s e r t i o n •O P o s i t i v e Level f o r A s s e r t i o n Pulse to P o s i t i v e Level f o r A s s e r t i o n _^ Pulse to Ground Level f o r A s s e r t i o n —> Nonstandard S i g n a l 64 REFERENCES 1 . Beddoes, M.P., "An Inexpensive Reading Instrument With Sound Output f o r the B l i n d , " IEEE Transactions on Bio-Medical Engineering, V o l . BME-15, No. 2 , A p r i l 1 9 6 8 . 2 . Caple, C.G., "The Lexiphone: An Experimental Reading Machine f o r the B l i n d " , M.A.Sc. Thesis, U.B.C., February, 1 9 6 6 . 3 . Fournier d'Albe, E.E., "The Type-Reading Optophone," Nature, 2 4 , P- 4 ; 1 9 1 4 . 4 . Cooper, F.S., and Zahl, P.A., "Research on Reading Machines f o r the B l i n d , " Progress Report to the Committee on Sensory Devices of the work done at the Haskins Lab o r a t o r i e s , New York, June 3 0 , 1 9 4 7 . 5. Mauch, H.A., and Smith, G.C., "The Development of a Reading Machine f o r the B l i n d , " Summary report to the Veterans Admin, from Mauch Labo r a t o r i e s , Inc., Dayton, Ohio, June 3 0 , 1 9 6 6 . 6. Nye, P.W., "An I n v e s t i g a t i o n of Audio Outputs f o r a Reading Machine," Research B u l l e t i n , No. 1 0 , American Foundation f o r the B l i n d , New York, J u l y , 1 9 6 5 . 7. A u s t i n , G.M., "Design and Construction of an Opaque O p t i c a l Countour Tracer f o r Character Recognition Research," M.A.Sc. Thesis, U.B.C., December, 1 9 6 7 . 8. Coffey, J.L., and McFarland, R.R., "The e v a l u a t i o n and Stan-d a r d i z a t i o n of S e l e c t i o n and T r a i n i n g Procedures f o r the B a t t e l l e A u r a l Reading Device," Research Report to the Veterans A d m i n i s t r a t i o n from the B a t t e l l e Memorial I n s t i -t u t e , Columbus, Ohio, June 3 0 , 1 9 6 3 -9 . Sayers, B.M., and Cherry, E.C., "Mechanism of B i n a u r a l Fusion i n the Hearing of Speech," Journal of the A c o u s t i c a l S o c i - ety of America, V o l . 2 9 , 9 7 3 - 9 8 7 , September, 1 9 6 7 . 1 0 . Golden, B.P., "Auditory Displays f o r D i r e c t - T r a n s l a t i o n Reading Machines," M.Sc. Thesis, M.I.T., May, 1 9 6 6 . 1 1 . Fournier d'Albe, E.E., "The Optophone: An Instrument f o r Reading by Ear," Nature, 1 0 5 , 2 9 5 - 2 9 6 , 1 9 2 0 . 1 2 . Cooper, F.S., "Research on Reading Machines f o r the B l i n d , " i n P.A. Zahl (ed.), Blindness: Modern Approaches to the Unseen  Environment, Pr i n c e t o n Univ 1. Press, P r i n c e t o n , N.J., 5 1 2 -5 4 3 , 1 9 5 0 . 13- Zworykin, V.K., and F l o r y , L.E., "Reading Aid f o r the B l i n d , " E l e c t r o n i c s , No. 19, 84-87, August, 1946. 65 14. Clowes, M.B., E l l i s , K., Parks, J.R., Rengger, R., Communication from the Na t i o n a l P h y s i c a l Laboratory, U.K., Feb. 22, 1961. 15. B l i s s , J.C., " K i n e s t h e t i c - T a c t i l e Communications," IRE Tran-sactions on Information Theory, IT-8, 02-98, 1962. 16. , "Medical E l e c t r o n i c s f o r B l i n d Readers," E l e c t r o n i c s , 18, 35-36, January 25, 1965-17. Simon, 0., Int r o d u c t i o n to Typography, Harmondsworth, Middlesex, Penguin Books Lt d . , 1954. 18. Mauch, H.A., and Smith, G.C., "The Development of a Reading Machine f o r the B l i n d , " Summary Report to the Veterans A d m i n i s t r a t i o n from Mauch Labo r a t o r i e s , Inc., Dayton,. Ohio, June 30, 1968. 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

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>
                        
                    
IIIF logo 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.831.1-0104050/manifest

Comment

Related Items