MI-EEG Recognition Based on Euclidean Alignment and Style Transfer Mapping

To improve the efficiency and universality of transfer learning in the application process of motor imagery (MI) brain-computer interface (BCI), we propose a classification recognition framework for MI-BCI. We use Euclidean alignment (EA) to align electroencephalogram (EEG) signals obtained after preprocessing, then regularized common spatial pattern (RCSP) is used to remove the redundancy information of the signals. We use a transfer learning (TL) algorithm based on feature mapping to carry out specific knowledge transfer so that the classifier on the source domain can be transferred to the target domain for classification recognition. The experiment results show that the average classification accuracy of the framework on data sets 1 and 2a of the BCI competition IV have increased significantly compared with recent literature.