A neural data lossless compression scheme based on spatial and temporal prediction

A lossless compression scheme for extracellular recordings is presented in this paper. A linear neural network (LNN) is used to predict future values of the signal based on its past samples and data from its neighboring channels. The difference between the signal and the predicted value is transmitted achieving lossless compression. The LNN can effectively be used to exploit spatial and temporal correlation in neural signals, achieving about 2 times more compress ratio (CR) when compared to lossless delta compression. It is also found that using the compressor on small groups of neighboring channels helps achieving higher CR for the same amount of memory and multiplications.

[1]  Mohamad Sawan,et al.  A Mixed-Signal Multichip Neural Recording Interface With Bandwidth Reduction , 2009, IEEE Transactions on Biomedical Circuits and Systems.

[2]  Maurits Ortmanns,et al.  Delta compression in time-multiplexed multichannel neural recorders , 2016, 2016 12th Conference on Ph.D. Research in Microelectronics and Electronics (PRIME).

[3]  Maurits Ortmanns,et al.  Real-time data compression of neural spikes , 2014, 2014 IEEE 12th International New Circuits and Systems Conference (NEWCAS).

[4]  Refet Firat Yazicioglu,et al.  22.7 A 966-electrode neural probe with 384 configurable channels in 0.13µm SOI CMOS , 2016, 2016 IEEE International Solid-State Circuits Conference (ISSCC).