Dynamic Time Warping based connectivity classification of Event-Related Potentials

Abstract

Human brain electrical responses measured as Electroencephalogram epochs have different characteristics by means of amplitude and frequency content depending on the conditions and stimuli. Event-related potentials are the responses given to the stimuli and can be measured using the EEG. The average of these epochs are computed to remove the background activity and helps to exhibit the response to stimuli solely. In the concept of this study, dynamic time warping based connectivity features are used to classify the single-trial ERP epochs. Color Stroop test was implemented and ERP data are collected from 10 subjects. Support vector machine and K-NN classifiers are used and accurate classification results are achieved with the use of DTW metrics.