Adaptive Filtering of Neuronal Spike Train Data

A method of analyzing neuronal spike train stimulus-response data which enhances temporal features and reduces nonstationarities is described. First, a Parzen estimate of the post-stimulus density function is computed by convolving spike events with Gaussian kernels. Second, successive segments of the spike train are correlated to a template, and the temporal relationship between segments is adjusted for maximum correlation. This method has been applied for the identification of high-frequency rhythms in spike train data from the cat optic nerve.