부산대학교 도서관

EEG signals contain information of limb motion patterns that are difficult to be extracted using conventional methods. The purpose of this study was to determine the feasibility of novel machine learning algorithms in extraction of hand motion intents from EEG. In this paper, 3 hand motion patterns (hand opening, closing, resting) were recognized using the filter bank common spatial pattern (FBCSP) method. The spatial filtering method and the CNN classifier were used to decode and classify the EEG signals. 15 healthy subjects' EEG datasets were used. Results demonstrated that the average correct rate of motion pattern recognition was 83.61%. The performance of these algorithms was validated using off-line computer-controlled robotic hand system, the average accuracy of a subject was 77.32%, and the average time of the system recognition action was 28.4ms. This study demonstrated that hand motion pattern recognition can be extracted from EEG signals and can be utilized by brain computer interface (BCI) to trajectory control a robotic hand motion.