Brain-Computer Interface System for Hand Rehabilitation

This paper describes a brain-computer interface (Bel) system for hand rehabilitation. The system is composed of an EEG amplifier, a desktop computer, and a hand orthosis. A healthy subject participated in the experiment. When performing hand movement tasks, his EEG signals were recorded. A support-vector machine (SVM)-based classifier was trained using the processed data. Then, the system with the learned SVM-classifier was tested using test data. As a result, it was confirmed that the classifier had test accuracy of 73.3 %, and that fingers movements (extension and flexion) of the experimental subj ect were performed by the hand orthosis which was activated by the subject's intention (motor imagery).