Analysis Research Of EEG Signal Based On Motor Imagery

Brain Computer Interface(BCI)is a human-conputer interaction technology,which doesn't depend on the brain's normal output pathway of peripheral nerves and muscles tissue,to achieve the communication and control of external computer or electronic devices by EEG.EEG is a potential activity and generated by the discharge of brain nerve cells electrophysiological on the cerebral cortex,it is possible to identify different conscious states of classification of different EEG.BCI provides a new way to communicate with the outside world for some the patients suffering from nervous system or muscles tissue disease.BCI also has potential applications in other fields,it has important research value and broad prospects for development.In this thesis,the EEG of Motor Imagery is the research object and has the characteristics of Event-related Desynchronization(ERD)and Event-related Synchronization(ERS),aiming at the feature extraction of motion imaginary EEG signals,analyzes wavelet packet energy entropy and AR model.Aiming at the classification algorithms of motion imaginary EEG signals,analyzes Linear discriminant analysis(LDA)and support vector machine(SVM).The feature extraction method of EEG is studied.For the wavelet packet energy entropy,The wavelet packet energy entropy is calculated by decomposing and reconstructing the wavelet packet.The eigenvector is constructed by the differential method and the ratio method.The experimental results show that the method can effectively complete the feature extraction.For the AR model,introduce the basic principle of the model,The final prediction error criterion and the information criterion are used to select the order,according to the Burg algorithm,t then build the eigenvector to achieve feature extraction.The classification algorithms of EEG signal is studied.For LDA,using the Fisher criterion to design the classifier,Through the optimization criterion,the extreme points and the optimal decision value are obtained,and the linear discriminant classification function is obtained to realize the classification and recognition of the signal.For SVM,for linear separable and non-linear separable cases,design of optimal classification surface by using different kernel functions.Combining with the previous feature extraction method,the classification and recognition of the signal are carried out,and the results are analyzed and discussed,The experimental indicates effectiveness of the method,for the application of the motor imagery lay the foundation.