Optimal feature subset selection for pattern classification and recognition of multichannel eeg data using spectral entropy as a complexity measure applications to alcoholic and sleep eeg data
