A novel classification strategy of motor imagery EEG signals utilizing WT-PSR-SVD-based MTSVM

Abstract

In order to control the actions of the devices by the electroencephalogram (EEG) signals accurately, the classification method of motor imagery EEG signals based on wavelet transform (WT)-phase space reconstruction (PSR)-singular value decomposition (SVD)-based multi-layer twin support vector machine (MTSVM) is presented in this study. WT-PSR-SVD which is abbreviated as WPS is presented to obtain the features of the EEG signals, the wavelet decomposed signals (WDSs) of the EEG signals are reconstructed into high-dimensional space by PSR, respectively, and the dynamic characteristics of the WDSs of the EEG signals can be reflected. The traditional multi-layer twin SVM trained by the training samples with the features based on WT-SVD (WS-MTSVM) can be used to compare with the proposed WPS-MTSVM method. The experimental results demonstrate that the classification ability of motor imagery EEG signals of the proposed WPS-MTSVM method is better than that of WS-MTSVM. It can be seen that feature extraction of motor imagery EEG signals of WT-PSR-SVD is superior to that of WT-SVD, and the classification method of motor imagery EEG signals of the proposed WPS-MTSVM method is effective.