Multi-classification for EEG motor imagery signals using data evaluation-based auto-selected regularized FBCSP and convolutional neural network

Abstract In recent years, there has been a renewal of interest in brain–computer interface (BCI). One the BCI tasks is to classify EEG motor imagery (MI). A great deal effort made on MI classification. What seems be lacking, however, multiple This paper develops single-channel-based convolutional neural network tackle multi-classification tasks. For multi-classification, single-channel learning strategy can extract effective information from each independent channel, making between adjacent channels not affect other. data evaluation method and mutual information-based regularization parameters auto-selection algorithm are also proposed generate spatial filters. The used problem an inaccurate mixed covariance matrix caused by fixed invalid training data. To illustrate merits methods, we tenfold cross-validation accuracy kappa as measures test two sets. BCI4-2a BCI3a sets have four mental classes. set, average 79.01%, 0.7202 using evaluation-based auto-selected filter bank regularized common pattern voting (D-ACSP-V) series (SCS-CNN). Compared traditional FBRCSP, improved 7.14% for set. By 9.54% compared with 83.70%, 0.7827.