WCSE

Classification Method of Motor Imagery EEG Signal Based on Wavelet Packet and Common Spatial Pattern

Abstract— The traditional EEG signal feature extraction method based on frequency characteristics only extracts the energy feature of each channel, ignoring the correlation information of EEG signals between different channels. To gain a better effect of feature extraction, this article puts forward a feature extraction method based on wavelet packet and common spatial patterns (CSP), namely taking advantage of wavelet packet’s multi-resolution characteristics to run the orthogonal decomposition within the all frequency field to extract the motor imagery  rhythm and  rhythm, from the motor imagery EEG signals of the left hand and the right foot, and further extract the features by doing the spatial filtering through CSP. In combination with the advantages of wavelet packet and CSP, this method could play the relevant information among different channels to the full and classify the two kinds of motor imagery EEG signals by making use of support vector machine (SVM). Better EEG classification results achieved via this experiment are that the maximum classification accuracy of  rhythm is 92.8571% and the maximum classification accuracy of  rhythm is 93.5714%. To achieve the practical application of BCI standards.

Index Terms— brain computer interface (BCI), motor imagery (MI), Wavelet packet, common spatial patterns (CSP)

Manzhen Ma, Libin Guo, Kuifeng Su
Department of Control Engineering, Academy of Armored Force Engin, CHINA

[Download]

Cite: Manzhen Ma, Libin Guo, Kuifeng Su, "Classification Method of Motor Imagery EEG Signal Based on Wavelet Packet and Common Spatial Pattern," Proceedings of 2017 the 7th International Workshop on Computer Science and Engineering, pp. 730-735, Beijing, 25-27 June, 2017.