ABSTRACT

This chapter provides a theoretical framework with which it can interpret electroencephalogram (EEG) power and coherence. The chapter considers the factors that determine the magnitude of peaks in the EEG power spectrum. Second-order blind identification is a blind source separation (BSS) method developed for time-series analysis in array processing which explicitly incorporates the temporal structure of the second-order statistics of the signals in developing a BSS from the EEG electrode. When applied to EEG signals, they can be used to obtain a multivariate spectral analysis that can identify brain networks operating in different frequency bands. The effects of volume conduction on EEG coherence are severe for closely spaced electrode, and limit the use of coherence to estimate spatial properties of brain networks. In EEG applications, spectral analysis provides a means to assess statistical properties of oscillations in different frequency bands.