Unsupervised Recognition of Neuronal Discharge Waveforms for On-line Real-Time Operation

Fast and reliable unsupervised spike sorting is necessary for electrophysiological applications that require critical time operations (e.g., recordings during human neurosurgery) or management of large amount of data (e.g., recordings from large microelectrode arrays in behaving animals). We present an algorithm that can recognize the waveform of neural traces corresponding to extracellular action potentials. Spike shapes are expressed in a phase space spanned by the first and second derivatives of the raw signal trace. The performance of the algorithm is tested against artificially generated noisy data sets. We present the main features of the algorithm aimed to on-line real-time operations.