Spike sorting based on radial basis function network with overlap decomposition

Spike sorting is the essential step in analyzing recording spike signals for studying information processing mechanisms within the nervous system. Overlapping is one of the serious problems in the spike sorting of multi-channel recordings. In this paper, a modified radial basis function (RBF) network is proposed to decompose the overlapped signals and separate spikes within the same RBF network. A modified radial basis function based on Gaussian function is employed in the method to improve the accuracy of overlap decomposition. In addition, the improved constructing algorithm reduces the calculation cost by taking advantage of the symmetry of the RBF network. The performance of the presented method is tested at various signal-to-noise ratio levels based on simulation data coming from the University of Leicester. Experiment results show that our method successfully resolves the fully overlapping problem and has better accuracy comparing with the Gaussian function.