Optimal Transform of Multichannel Evoked Neural Signals Using a Video Compression Algorithm

Abstract

One of the most important problems in the field of biomedical engineering is how to record a multichannel neural signal. This problem arises because recording produces a large amount of data that must be reduced to transfer it through wireless transmission, and data reduction must be made without compromising data quality. Video compression technology is very important in the field of signal processing, and there are many similarities between multichannel neural signals and video signals. Therefore, we use motion vectors (MVs) to reduce the redundancy between successive video frames and successive channels. We also test what transform for neural signal compression is best. Our novel signal compression method gives a signal-to-noise ratio (SNR) of 25 db and compresses data to 5% of the original signal.

Acquisition of action potentials with ultra-low sampling rates

“…Unfortunately, spike acquisition based on Nyquist sampling is hard-limited by the action potential bandwidth (several kHz) rather than the spiking rate. Various algorithms have been explored to reduce data transmission rates compared to raw waveform transmission in wireless spike acquisition, but these methods do not reduce sampling rates [1]- [3]. Initially, spikes are band-pass-filtered and sampled at the usual Nyquist rate, typically more than 10 kHz.…”