Data Compression versus Signal Fidelity Trade-off in Wired-OR ADC Arrays for Neural Recording

Pumiao Yan, Nishal P. Shah, Dante G. Muratore, Pulkit Tandon, E. J. Chichilnisky, Boris Murmann

Abstract: This paper investigates the efficacy of a wired-OR compressive readout architecture for neural recording, which enables simultaneous data compression of action potential signals for high channel count electrode arrays. We consider a range of wiring configurations to assess the trade-offs between compression ratio and various task-specific signal fidelity metrics. We consider the fidelity in threshold crossing detection, spike assignment, and waveform estimation, and find that for an event SNR of 7-10 the readout captures at least 80% of the spike waveforms at ∼150x data compression.