Open Collections

Energy-efficient compressed sensing frameworks for the compression of electroencephalogram signals

University of British Columbia

The use of wireless body sensor networks (WBSNs) is gaining popularity in monitoring and communicating information about a person's health. In such applications, the amount of data transmitted by the sensor node should be minimized. This is because the energy available in these battery-powered sensors is limited. In this thesis, we study the wireless transmission of electroencephalogram (EEG) signals. We propose novel, energy-efficient compressed sensing (CS) frameworks that take advantage of the inherent structure present in EEG signals (both temporal and spatial correlations) to efficiently compress these signals at the sensor node in WBSNs. We first present a simple CS-based framework that is adapted to the EEG WBSN setting. We optimize the sparsifying dictionary and demonstrate that using a fixed sparse binary sensing matrix offers similar performances to optimal matrices while requiring far fewer computations. We then add an energy-efficient Independent Component Analysis (ICA) preprocessing block to the simple CS framework to exploit the spatial correlations among EEG channels. We show that the proposed framework provides significant energy savings as compared to the state-of-the-art method. As well, for a fixed compression ratio, our system achieves similar normalized mean square error performance as the state-of-the-art method, which is better than that achieved by the simple CS framework. We further improve on the energy performance of the framework by replacing the ICA preprocessing block by a simpler, correlations-based interchannel redundancy module and by using entropy coding. On the energy front, our proposed CS framework is up to 8 times more energy-efficient than the typical wavelet compression method. We also show that our method achieves a better reconstruction quality than the state-of-the art BSBL method. We further demonstrate that our method is robust to measurement noise and to packet loss, and that it is applicable to a wide range of EEG signal types. We finally apply our CS framework to compress EEG signals in the context of a brain computer interface application and evaluate its impact on the performance of the system. We show that interesting energy savings can be realized at the cost of a mild decrease in classification accuracy.

Text

2013-10-24

Attribution-NonCommercial-NoDerivs 2.5 Canada

http://hdl.handle.net/2429/45359

Electrical and Computer Engineering

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/2.5/ca/