A new action potential classifier using 3-Gaussian model fitting

Spike sorting is a prerequisite for all researches on multi-channel extracellular neural signal recordings. In this paper, we develop a new method for action potential classification. We introduce a mathematical model consisting of three Gaussian waveforms, which appropriately represents the general shapes of action potentials. Then we search for the best-fit waveform for each noise-corrupted spike based on the model, using peak fitting method. These processes result in increased separability among different classes of action potentials. The performance of the proposed method is assessed with synthesized neural recordings composed by spike templates and white Gaussian noise in various SNR environments.