Pattern Classification of Motor Imagery EEG-NIRS Based on SVM with Artificial Fish Swarm Algorithm

To address the problems of parameter setting and the poor classification accuracy in the traditional support vector machine (SVM), this paper proposes an approach to optimize the penalty factor and kernel parameter of SVM classifier based on artificial fish swarm algorithm (AFSA). AFSA-SVM was applied in two-class pattern classification problem, and common spatial pattern (CSP)algorithm was employed to extract the features of motor imagery electroencephalogram (EEG)and near-infrared spectroscopy (NIRS)signals in this paper. The total classification accuracy of AFSA-SVM classifier was higher than that of traditional SVM classifier about 5 %. The method proposed could improve classification accuracy of SVM classifier and has its advantages.