An EEG Transfer Learning Algorithm Based on Mutual Information and Transfer Component Analysis

The Brain -Computer Interface (BCI) is the decoding of EEG signals from different users and conversion into required signal instructions. However, because different subjects produce different EEG signal distributions far the same signal, when taking the EEG data of a single subject into the trained classifier to classify the EEG signals of different subjects, the experimental accuracy is greatly reduced. In recent years, transfer learning (TL) has been applied in the field of brain -computer interfaces, and transfer learning can effectively reduce the difference in distribution between the two fields. In order to reduce the negative migration problem caused by the feature lengthiness of the transfer learning process. In this paper, a manifold spatial domain adaptive algorithm (11-TCA) based on mutual information feature selection is proposed. Firstly, the EEG data is preprocessed, the data is aligned in the manifold space, and the tangent features are obtained on the tangent plane of the SPD manifold after aligned data, and then the features are sorted and selected by mutual information algorithm, and finally the new source domain and target domain features are obtained by reducing the distribution distance between the source domain and the target domain by TC4 algorithm. Experimental validation was performed on the BCI competition IV dataset 1 and compared with existing algorithm results. Experimental results show that the proposed M-TC4 method has an average experimental accuracy of 71.23% of the single source domain on the BCI competition IV datasetl, which Compared with the existing experimental methods, it has certain advantages.