A low-power analog spike detector for extracellular neural recordings

Abstract

This paper discusses a low-power spike detection circuit, which reduces bandwidth from neural recordings by only outputting a short pulse at each neural spike time. Communication bandwidth is dramatically reduced to the number of spikes. The principal idea is to use two low pass filters, one with a higher cutoff frequency to remove high frequency noise and the other with a lower cutoff frequency to create a local average. When the difference between the signal and the local average rises above a threshold, a spike is detected. The circuit uses subthreshold CMOS to keep the power consumption low enough for integration of many channels in an implanted device. This spike detection method shows promising results towards a robust and unsupervised algorithm that is lower power and more compact than existing spike detection methods.