Motor Imagery EEG Signal Classification Algorithm Based on Riemannian Space

Brain computer interface (BCI) can directly control external devices by providing additional signal pathways, which has a broad application prospect in medical rehabilitation and other aspects. It is a long-term research hotspot in this field to find an efficient classification algorithm and reduce the training cost. Most researchers are currently improving the traditional algorithm in order to achieve better results, but the improvement is limited. Therefore in this paper, we proposed a novel EEG signal classification algorithm based on Riemannian space, not on traditional Euclidean space. In this proposed algorithm, sample covariance matrix is used as the feature of EEG signal, and Riemannian mean and distance are used to classify the matrix, which realizes the application of manifold learning in BCI. Experimental results show that this method can effectively distinguish EEG signals, and the accuracy is better than that of traditional methods.