A Simple but Efficient EEG Data Compression Algorithm for Neuromorphic Applications

Abstract

Widespread use of Multichannel Electroencephalograph (MCEEG) in diversified fields ranging from clinical studies to Brain Computer Interface (BCI) application, has put in a lot of thrust in data processing concepts, for effective storage and transmission. The paper proposes a computationally simple and novel methodology Normalized Spatial Pseudo Codec (n-SPC) to compress MCEEG signals. The signals are first normalized followed by two operations namely the spatial coding and pseudo coding operating on integer part and fractional part of the normalized data respectively. The proposed method was evaluated on publicly available EEG databases and results indicate that the algorithm exhibits good storage efficiency with average Compression Ratio (CR) of 4.61 with a computational complexity of only O(zN). The algorithm offers significantly a better decompressed signal quality, quantified by average Peak Signal to Noise Ratio (PSNR) of 21.42 dB. The average encoding and decoding time per sample is 0.3 and 0.04 ms, respectively with an average Percentage Root Mean Square Deviation (PRD) of 5.33. The efficacy was further evaluated using the decompressed signal to detect sleep spindle, from an excerpt of EEG recording and was compared with the visual scoring of two experts, available at the DREAMS Sleep Spindles Database. Hence, the proposed compression scheme can be used in practical MCEEG recording, archiving and BCI and neuromorphic systems.