Research And Implementation Of EEG Control Method Based On Motor Imagery

The Brain Computer Interface(BCI)achieves the interaction between human brain and external devices by analyzing the EEG signal and decoding the user's intention into control commands to control the output device.The core of brain computer interface technology is the recognition of EEG signal,but EEG signal is non-linear and nonstationary,so how to extract the features of EEG signal effectively becomes the key to recognize EEG signal.Therefore,this thesis focuses on the feature extraction and system realization of motor imagery EEG signal,which has important theoretical research and practical application value.Firstly,the research status of BCI technology at home and abroad is reviewed.A BCI system based on left-handed and right-handed motor imagery EEG signal is designed,and the acquisition method of EEG signal is introduced.At the same time,through the study of analysis methods of EEG signal,the basic methods of feature extraction and classification are determined.Then,to solve the problem of high delay time in the traditional BCI system,a shortterm feature extraction method ST-EMD based on Empirical Mode Decomposition(EMD)is proposed.In this method,the EEG signal is segmented by window function,and the boundary extremes of the intercepted signal is extended.The signal is adaptively decomposed into multiple Intrinsic Mode Functions(IMFs)to obtain high-resolution time and frequency features.Finally,the features are classified by Support Vector Machine(SVM).Experimental results show that compared with EMD method,ST-EMD can effectively reduce the delay time while ensuring the recognition accuracy.Furthermore,to solve the problem of low recognition rate of EEG signal,a multi feature fusion method based on EMD and improved Common Spatial Pattern(CSP)algorithm is proposed.The EMD-CSP method preprocesses EEG signal by ST-EMD algorithm,and then the decomposed IMF components are combined into a new signal matrix.According to the amplitude difference of IMF signal matrix in different states,a spatial filter is constructed to obtain the spatial distribution features of the signal matrix.Finally,it is fused with the time-frequency features extracted by EMD.Offline experiments show that the method has higher average recognition accuracy and smaller standard deviation than single feature.Finally,the human-computer interaction system based on motor imagery is designed and implemented on the intelligent wheelchair platform,ST-EMD and EMD-CSP algorithms are integrated into the system to control the wheelchair movement.The experimental results show that compared with the traditional control method,the improved control method has higher recognition rate and less time-consuming,which verifies the feasibility of the improved algorithm and the effectiveness of the system.