Feature Classification of Motor Imagery EEG Based on Deep Learning Networks

The Brain Computer Interface (BCI) technology refers to not rely on normal brain nerve and musclar tissue, but on computer or other equipment to build a new information transmission circuit between the brain and the external environment, it can directly realize the information exchange between brain and external environment, which is a new way for human to communicate with the outside world. Motor imagery brain-computer interfaces are increasingly being used in practical application scenarios where the degree of accuracy in classification is crucial. However, due to the small amount of data available for MI EEG, overfitting can easily occur when using deep learning network models for classification training, resulting in low recognition rates. To overcome this problem, we have proposed, attention mechanismc convolutional neural network (AMCNN) which is a novel optimized convolution neural network model. A 4-Layer convolutional neural network is built as the classifier and convolutional kernels of different sizes are applied. The performance obtained by the proposed approach is evaluated by Accuracy, Precision, Recall and Kappa value. The accuracy on dataset 2a from competition IV(BCIIV-2a) reaches 86%, and the best Precision, Kappa and Recall on dataset 2a from competition IV is greater than many of other methods.