A Procedure For Automatic Classification Of EEG Genetic Variants

A procedure using spectral parameters for automatic classification of inherited EEG variants was investigated. Spectral parameters were calculated by extracting the poles of the transfer function derived from a fifth order AR model of the EEG. The visible form of each genetic EEG variant appeared to be determined by a dominant pole of the transfer function. Using classification rules developed from observing the spectral parameters of clearly defined EEG samples of each variant, twenty two EEG samples were processed. In a strict sense, there was a 59 percent agreement between the classification procedure and the impressionistic cl ass i f i cat i on indicating that the use of these spectral parameters is a viable technique for classifying EEG variants.