Neural spike sorting using mathematical morphology, multiwavelets transform and hierarchical clustering

Neural spike sorting is an indispensable first step for the analysis of multiple spike train data. As it is very common that noises degrade the performance of most of the available spike sorting method, in this paper we have proposed a novel spike sorting algorithm framework which seems less susceptible to heavy noises. At first, mathematical morphology operation is used to facilitate the spike event detection process, especially in strong noisy situations. Then, multiwavelets transform is performed to the detected spike waveforms to extract discriminative features. Finally, hierarchical clustering with an outlier removal process proceeds to separate the first 10 distinguishable multiwavelets coefficients. The results show that our spike sorting method performs quite well even for the heavy noisy simulated spike data.