The Study Of EEG Feature Extraction And Classification Based On Movement Imagination

Brain-computer interface is a new way of human-computer interaction developed in recent years, it does not depend on the brain peripheral nerve and muscle tissue. The brain can directly communicate with the outside world by this way. It can judge the person’s real thoughts through the human brain wave signals. The human brain wave signals commonly referred to the EEG signal.The main work of this paper is the study based on EEG signals.This article focuses on the study of human brain thinking activity during movement imagination. The main tasks of this paper are the feature extraction and classification.This paper raised the combination feature extraction method of the power spectral density and wavelet packet in frequency-domain. First, The paper adopt the power spectral density method and get the average periodogram of EEG spectrum. Then the more obvious features appear in the spectrum near10Hz. In the experiments, the8-12Hz is the best extraction frequency band of left and right the spectrum range. The signal is decomposed by wavelet packet. The (42) node coefficient is extracted as the feature value after Wavelet packet decomposition of4layers. After that, The eigenvalues of C3and C4electrodes were squared and summed separately. The result is the energy difference. Feature vector of energy values are obtained based on event-related principle. Based on the energy difference of C3and C4electrode, the two-dimensional feature vector is generated, it is the input of classification. The characteristic value from the two methods and improved methods, is put into the Nearest neighbor classifier, Discriminant function of Fisher, Support vector machine classification respectively. comparing the experimental results, the linear classifier got the best classification result. The paper also raised improved method of the linear classifier to reduce complexity. Finally, This paper described design process and results of the Emotive brain electrode cap equipment, developed a brain computer interface system.In this paper, the experiments use the international standard BCI competition data. Data are stored mainly by the form of MATLAB. Data includes training data, test data, and the correct results of the data. By simulation of the experimental data, Most of the recognition rate are higher than80%, the highest recognition rate is85.71%.