Abstract

Info

Title:: Pattern classification of motor imagery EEG signals based on ABC-SVM algorithm

Author(s):: MAYuliang; LIUWeixing; ZHANG Songjie; WANG Zhenjie; ZHANG Qizhong; Institute of Intelligent Control and Robotics, Hangzhou Dianzi University, Hangzhou 310018, China

Keywords:: electroencephalogram signal; artificial bee colony algorithm; support vector machine; regularization common spatial pattern; pattern classification

Abstract:: Due to the problems of traditional support vector machine (SVM) classification method in electroencephalogram (EEG) signal processing, such as high complexity of searching the optimal parameters, heavy workload and low classification accuracy, a new SVMclassification method based on artificial bee colony (ABC) algorithm is proposed in this study to improve the accuracy of motor imagery EEG recognition. Firstly, the regularization common spatial pattern is used for EEG feature extraction. Then penalty factor and kernel function of SVM are optimized by ABC algorithm. Finally, the optimized SVM classifiers is trained and tested by two kinds of EEG data of right foot and right hand movements. The final results show that the accuracy of ABCSVM classifier for EEG classification is averagely 2.5% higher than that of non-parameter-optimized SVM classifier, which proved that the proposed algorithm is feasibility and achieves a high accuracy in motor imagery EEG recognition.