Coherence Analyses of Event-Related Potentials Using Fourier and Wavelet Transforms

Abstract

The EEG coherence is beneficial to investigate cortical connectivity of different brain regions in response to specific events. Recently, the wavelet-based coherence is suggested for this purpose due to that it can capture the time-varying characteristics of signals from which the Fourier transform would suffer. However, the quantitative comparison between Fourier and wavelet-based coherence is less addressed. In this paper we used simulated signals under various signal-to-noise ratios and real evoked electrophysiological signals to assess the coherence estimation of these two methods. Our results showed that Fourier-based coherence is more affected by the Guassian noise than wavelet-based coherence, in particular at the frequency band of 18-50 Hz