Simulation of a fully automated system for optimal on-line recognition of neural waveforms in extracellular recordings

We present a software simulation of a fully automated system that can optimally detect the occurrence, classify and resolve superpositions of neural spike waveforms from multiple neurons recorded with the same extracellular electrode. The algorithms in this simulation are designed for on-line operation without human supervision. Therefore, the whitening filter required for optimality, waveform templates and discrimination thresholds are generated automatically.