Neural spike detection and localisation via Volterra filtering

Abstract

The spike detection problem is cast into a delay estimation. Using elementary operational calculus, we obtain an explicit characterization of the spike locations, in terms of short time window iterated integrals of the noisy signal. From this characterization, we derive a joint spike detection and localization system where the decision function is implemented as the output of a digital Volterra filter. Simulation results using experimental data shows that the method compares favorably with one of the most successful one in the literature.