Power Spectral Density Analysis in Alfa, Beta and Gamma Frequency Bands for Classification of Motor EEG Signals

摘要

The main idea of the brain-computer interface (BCI) systems is to facilitate the lives of individuals whose cognitive functions are healthy but who have difficulty in moving their muscles due to motor nervous system disorders. The BCI systems are generally EEG-based and their success depends on the preprocessing of the signal, the detection of distinctive features, the use of appropriate classifiers and the selection of effective channels. In this study, power spectral analysis based feature extraction was performed for alpha, beta and gamma frequency bands in classification of motor tasks, and the classification performance was evaluated by applying the extracted features to the k-nearest neighborhood (k-EYK) and support vector machines (DVM) classifiers. Accordingly, 99.92% classification accuracy was obtained in the case where k-EYK was used together with the Burg method for the beta band. Thus, it is possible to say that the proposed methods are successful in recognizing motor imagery tasks.