Segmentation of the epileptic EEG by means of a finite state automaton.

The aims and merits of the syntactical decomposition of the epileptic EEG are discussed in some detail. General assumptions in the design of segmenting algorithm--an essential part of the reduction of data program--are considered and a working example of segmenting routine is presented. This routine is designed as a finite state automaton, equivalent to a formal regular grammar. The performance of the routine on two sets of experimental data is presented and its intrinsic stability is shown in terms of cluster analysis of the segments produced. Means of extension of the proposed scheme to other cases of potential interest as well as possible applications in the higher level syntactical analysis are pointed out.