Wavelet-based analysis of spectral rearrangements of EEG patterns and of non-stationary correlations

In this paper we present a novel technique of studying EEG signals taking into account their essential nonstationarity. The bursts of activity in EEG rhythm are modeled as a superposition of specially designed elementary signals against the background of a real EEG record at rest. To calculate the time variation of quantitative characteristics of EEG patterns we propose the algorithm based on continuous wavelet transform (CWT) followed by the analysis of spectral integral dynamics in a given frequency range. We introduce new quantitative parameters to describe the dynamics of spectral properties both for each burst of brain activity and for their ensemble. Based on the given model we have identified the appearance and disappearance of patterns in EEG rhythm. The problem of non-stationary correlation of different EEG channels is solved. The use of the techniques for analyzing and classifying transient processes related to the activity of human central nervous system is also discussed.