Multidimensional spectral analysis of the electroencephalogram

he thesis investigates the hypothesis that specific brain states can be identified from spatial patterns on EEG topographic maps. Preliminary work showed that topographic maps of rms EEG data are representative of the actual potential distribution on the scalp and that errors in maps are not significantly affected by the method of interpolation used in their construction. Spatial patterns on the maps were subsequently investigated using spectral analysis. With simulated EEG data, results indicated that maximum entropy (ME) power spectrum estimates (PSEs) are consistently superior to Bartlett and Blackman-Tukey PSEs in terms of error in peak-position and the minimum separation in frequency required to identify two sinusoids. The ME PS analysis of sinusoids in white noise showed that acceptable PSEs could be obtained if the SNR was...