A superposition-based analog data compression scheme for massively-parallel neural recordings

Abstract

An analog data compression scheme for massively-parallel observation of neural activity is presented, that simultaneously reduces the total number of physical transmission lines. The compression is realized by superimposing signals of different recording sites on one common transmission line, a reconstruction of spatial information is achievable by correlation, for which three simple mathematical operations are presented. The generalized system-theoretical description allows to apply the concept to any multidimensional analog sensing network, and grants to extract the relation between compression and SNR reduction. Simulation-based examples of neural recording arrays revealed possible transmission line reductions of 92% in a 25 × 25 array with 11.5dB higher noise floor, or 70.8% in a 6×8 array with an SNR reduction equivalent to only 1 bit of resolution.