EEG signal classification with feature selection based on one-dimension real valued particle swarm optimization

Abstract

In this study, a new scheme was presented for the EEG signal classification with feature selection based on one-dimension real valued particle swarm optimization. In the proposed scheme, normal and abnormal EEG signals were decomposed into various frequency bands with one fourth-level wavelet packet decomposition. Approximation entropy value of the wavelet coefficients at all nodes of the decomposition tree were used as a feature set to characterize the predictability of the EEG data within the corresponding frequency bands. Then, the one-dimension real valued particle swarm optimization algorithm was used to find the optimal feature subset by maximizing the classification performance of a support vector machine based EEG signal classifier. Experimental results showed that the proposed method improved the classification performance substantially and got a much less size of optimal feature subset with compared to the other methods.