Classification of Motor Imagery EEG Signals for Using in Neuro-Rehabilitation Applications

Brain computer interfaces which are developed for the rehabilitation systems decode motor imagery EEG signals to control external devices. However, the extraction of the features from the EEG imagery signals and classification of it is an important problem. In this paper, common spatial pattern analysis, which is widely used in motor image applications, was preferred for getting features. As a classifier, the accuracy performances of Artificial neural network (%93), Convolutional Neural Network (%91), Support Vector Machine (%84) and K-Nearest Neighbour Algorithm (%90) were compared. As a result of the comparison, Artificial Neural Network method was the most successful classifier with %93.9 accuracy.