Research On Classification Method Of Motor Imagery EEG Signal Based On Joint Model

Brain-computer interface,as a new type of information exchange with multidisciplinary integration,has broad application prospects in the fields of medical rehabilitation,military defense,consumer entertainment,etc.The critical technology is to classify motor imagery EEG signals.For the existing traditional methods of classifying motor imagery EEG signals,its temporal and spatial features cannot be well taken into account in the process of feature extraction,and the classification accuracy is difficult to meet the needs of practical applications.In this thesis,we conduct research on the classification method of motor imagery EEG signals based on attention mechanism of gated recurrent unit and cross-channel residual connection network.The main contents of the project are as follows:(1)To study the temporal features of motor imagery EEG signals,in order to propose a method for classifying motor imagery EEG signals based on attention mechanism of gated recurrent unit(AGRU).The method extracts the temporal features of EEG signals and completes the classification by combining attention mechanism and gated recurrent unit.Through comparison experiments with the four existing classification methods,the effectiveness of the motor imagery EEG signal classification method based on AGRU is verified,but the classification accuracy needs further improvement.(2)To study the temporal-spatial features of motor imagery EEG signals,in order to propose a method for classifying motor imagery EEG signals based on attention mechanism of gated recurrent unit and cross-channel residual connection network(AGRU-CRCNet).This method is an improvement on the problem of incomplete extraction of motor imagery EEG signal features of AGRU.Firstly,the temporal features are extracted by the AGRU module;secondly,the temporal features are input into the CRCNet module to further extract the spatial features of the motor imagery EEG signals,and the temporal-spatial features are obtained;finally,the features are input into the classifier to complete the classification.The experimental results show that the classification method based on AGRU-CRCNet proposed in this thesis achieves an average accuracy of 91.94% on the public dataset of EEGMMIDB,which proves that the method can effectively take into account the temporal-spatial features of motor imagery EEG signals and improve the classification accuracy to a certain extent.