Real-time adaptive discrimination threshold estimation for embedded neural signals detection

Multi-electrode array systems used in neurological applications produce large amount of data because of the simultaneous continuous high-rate sampling on a large number of channels. This data flow must be reduced to envision compact data acquisition systems with wireless transmission for body implantation. In spike-related applications, the useful data is sparse due to the relative low neurons firing rate combined to the high sampling rate. High compression ratio can be achieved by detecting, extracting and storing only the relevant spike occurrences. The first step is to provide a simple yet robust discrimination threshold based on the characteristics of the noise distribution. This article presents both a method and its hardware implementation for adaptive spike detection.