Research And Implementation Of Technologies Of Brain-Computer Interface Based On Multi-Class Motor Imagery

The brain computer interface(BCI)is a revolutionary human-computer interaction technology that allows for direct data transmission between the brain and various external devices.The electroencephalography(EEG)signal is analyzed and processed to identifies the real intention of the subject and then converts it into the corresponding external device control instruction to control the external device.However,the process of obtaining EEG data is easily disrupted by external environment and artifacts.How to effectively remove the artifacts of EEG signals and extract the effective features to achieve the classification that has emerged as the BCI’s most important technology.Therefore,it is of great significance and value to study the artifact removal,feature extraction and classification of EEG signals and the implementation of BCI system.Firstly,the thesis expounds the current state of research on EEG signal recognition algorithm in BCI system at home and abroad and analyzes and summarizes the main difficulties and problems faced by BCI system.Moreover,the basic framework and modules of BCI systems are analyzed and the experimental paradigm for EEG data acquisition is studied and designed which lay a foundation for subsequent research.A method based on KSE-Fast ICA is proposed to remove the EEG artifact in order to solve the problem that the traditional EEG artifact removal method is easy to lose useful EEG signals.Firstly,EEG is decomposed into several independent component by using fast independent component(Fast ICA).Then,the component of ophthalmic artifact is identified by the combination of kurtosis and sample entropy(SE).Following that,complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)and wavelet threshold are used to denoise the detected artifact components.Finally,the pure EEG signal is reconstructed by inverse CEEMDAN and Fast ICA algorithm.Experiments show that KSE-Fast ICA algorithm can remove ocular artifacts more effectively than other ophthalmic artifact removal algorithms.A algorithm based on DT-CSP-SVM is proposed to solve the problem of low recognition accuracy of multi-class motor imagery EEG.Firstly,one versus rest common spatial pattern(OVR-CSP)is used to construct multiple spatial filters.Then,support vector machine(SVM)is used for classification to select the best spatial filter,and the selected filter and SVM are used to construct the first branch of the decision tree(DT).Finally,the decision tree branches are constructed repeatedly using OVR-CSP and SVM until the decision tree that can distinguish all categories.The results of the experiments reveal that the DT-CSP-SVM method extracts more distinctive information from multiclass motor imagery EEG data and has a greater classification accuracy.Finally,an intelligent wheelchair system based on KSE-Fast ICA and DT-CSP-SVM algorithm is designed and implemented.The subject control the wheelchair to move according to the required fixed trajectory.The algorithm and system are verified by analyzing the actual moving path and time of the subject.Compared with the intelligent wheelchair system based on other EEG recognition algorithms shows that the actual moving path of the system based on the algorithm proposed coincides with the fixed track required to move more and takes less time which verifies the effectiveness of the intelligent wheelchair system based on KSE-Fast ICA and DT-CSP-SVM algorithms.