EEG Motor Imagery Classification Based on Sliding Window and Attention

As one of the most active research directions in the field of neural engineering, the brain-computer interface has developed rapidly in recent years, and its application fields have also been expanding. As one of the brain-computer interface paradigms, motor imagery is mainly used in medical treatment. It can provide a new way of movement for disabled patients with motor disorders, help them get rid of the disease and achieve self-care. However, the limited performance of EEG signal decoding is restricting the extensive development of the brain-computer interface industry. In this paper, a model based on sliding window and attention mechanism(SAIR) is proposed for the classification of EEG motor imagery. SAIR uses various techniques to improve the accuracy of motion image classification. Sliding window and multi-head self-attention methods expand data and extract valuable features. The model was evaluated by experiments, and the accuracy reached 0.8133 on the dataset of BCI Competition IV-2a.