Research on classification algorithm of EEG Signal based on transfer learning

• 1、School of Information and Communication Engineering, Beijing University of Posts and Telecommunications, Beijing 100876

Abstract：Deep learning has been applied in the field of motor imagerybrain-computer interfaces with some success, but its application still faces many challenges. The existing feature extraction and classification methods cannot eliminate the influence of individual variability of different subjects, and it is difficult for a model trained on one dataset to achieve the same performance on other datasets. In this paper, we propose a model transfer algorithm based on federated learningfor motor imagery EEG signals classification.By studying the influence of individual differences on EEG classification, a novel theoretical framework and technical approach are proposed. A deep transfer learning method based on generative adversarial network (GAN) is adopted in the overall framework, and federated aggregation is introduced to train a joint EEG classification model, which solves the problem of data heterogeneity and data leakage during data is fused.The results of cross-subject experiments on BCI Competition IV 2a dataset show that the joint classification model we propose can greatly improve the accuracy of cross-subject classification.It can be seen that the model proposed in this paper improves the influence of individual variability on the classification effect to a certain extent, solves the problems of universality and portability of EEG classification model, and data leakage during data fusion.which is of practical significance for promoting the development of brain-computer interface technology to the direction of practicality and marketization.

Keywords： artificial intelligence brain computer interface motor imagination transfer learning