IMPROVED STRUCTURE OF PRE-SEGMENTED EEG-S ECTIONS BY CLUSTERING WITH GLOBAL OPTIMIZATION - A METHODICAL STUDY

The automatic pattern classification in EEG recordings realized by the subsequent processing steps of segmentation, feature extraction and c lassification, is one opportunity for the realization of decision supp orting systems in computer-assisted EEC analysis. As a rule cluster al gorithms are used as classifiers (explorative approach). The aim of th is study is to replace local optimizing cluster algorithms by a cluste r approach with global optimization. This results in a definite and re producible classification of EEC segments in single recording channels . It can be shown that on the basis of such optimal classification mor e precise decision making is possible.