Wavelet-analysis in application to studying spike separation and information encoding in neuron dynamics

Abstract

We study how the noise statistics influences the performance of separation of extracellularly recorded spikes by principal component analysis and wavelet-based technique. We show that the two approaches have different robustness against the frequency band of the experimental noise and an appropriate filtering of the spike waveforms can significantly improve the results of separation. For the wavelet technique we suggest filter parameters optimizing spike separation. Finally we discuss a hypothesis that information encoding in neural dynamics may sometimes be considered in terms of frequency modulation.