Feature Extraction And Intention Recognition Of Single Trial Motion Imagination Eeg Signals

Because of the outstanding contribution in rehabilitation and other industries,brain computer interface(BCI)has become a research hotspot in the field of international intelligent science.The BCI system based on the electroencephalogram(EEG)signal of single movement imagination takes the EEG signal generated by different movement imaginations as the input,through analysis,a control signal for judging the motion state is generated.The pattern recognition of EEG signal is the focus of the whole system design.In this paper,based on the intention recognition of single motor imagery EEG signal,the process of acquisition,denoising,feature extraction and classification of motor imagery EEG signals are studied and analyzed.(1)In view of the lack of supervision of the spontaneous EEG evoked paradigm,the long-term monotonous tasks are likely to cause subjects' attention to drop and other problems,the paper proposes a new EEG-evoked paradigm called “Bowl-Ball” experiment.On the basis of EEG's intentional imagination,dynamic feedback control is added to study the EEG intention recognition in the task of constrained manipulation objects.(2)EEG signals contain abundant time and space information.In order to filter out the noise artifacts,this paper compares the maximum component analysis method and the independent component analysis method,including the separation performance and real-time processing of removing the artifacts contained in the EEG signal.The results show that both methods have a separation effect on noise signals,and both can guarantee real-time performance.Compared with independent component analysis,the maximum component analysis method has better separation effect,higher correlation index,higher separation similarity,and smaller and more stable similarity value,which has a wider application prospect.(3)Due to the non-stationary and non-Gaussian characteristics of EEG signal,the EEG signal is analyzed from the perspective of phase characteristics.By extracting the intrinsic mode function of the EEG signal,the phase fluctuation function is constructed by the Hilbert transform.The features of the EEG signal are extracted using the detrended fluctuation analysis.The EEG intentions are recognized using regularized discriminant analysis.The BCI competition data set is used for comparative experiments.Based on the proposed method,the accuracy of EEG intent recognition in BCI competition can reach 92% whereas “Bowl-Ball” experiment can reach 96%(p<0.01).The research shows that the phase fluctuation can distinguish the features corresponding to different intentions well,and the neural feedback signal can improve the individual's mission performance by affecting the arousal degree and improve the accuracy of the intent recognition.In this paper,the motor imaging EEG-evoked paradigm with feedback and the analysis and research on the motor imaging EEG signals provide theoretical support for the future recognition of motor imagination intentions,and have a positive significance for improving the classification performance of the BCI system.