Electroencephalogram Recognition of Imaginary Right and Left Hand Movements by brain-computer Interface
摘要
Based on the peak value of power spectral density (PPSD) and corresponding frequency (CF), an approach that performs electroencephalogram (EEG) feature extraction during imaginary right and left hand movements was proposed. The data were gained from brain computer interface competition in 2003 provided by Graz University of Technology. The EEG signals between 8-16 Hz were decomposed by db3 wavelet packet at three levels. The PPSD and CF of electrodes C3 and C4 were defined as the EEG feature vectors and calculated respectively. The left and right hand motor imaginary tasks were distinguished by the time-variable linear classifier. The proposed method was applied to the test data for 140 trials. The satisfactory results were obtained with the highest classification accuracy 89.29%. The maximum mutual information was 0.622 8 bit, and the signal-to-noise ratio (SNR) was 1.3713. The PPSD and its CF on electrodes C3 and C4 between 8 and 16 Hz were coincident with event-related desynchronization (ERD) and event-related synchronization (ERS). This method is simple, quick, and promising for on-line brain computer interface system.