The Research And Application Of Brain-Computer Interface Based On Motor Imagery

With the development of the science and technology, Brain computer interface is received more and more attention and has become a hot issue of biomedical engineering research field in nowadays, for its prospering both in theory research and practical application. Furthermore, the classification of the EEG data recorded during motor imagery, which is the core content of this thesis, is a very important branch.First of all, experiments that how to gain the EEG data were designed in this thesis. Then feature extraction and classification were studied by using the EEG data gained from experiments. Finally, we established a BCI system based on motor imagery and attempted to put the system into real application. The main content of the thesis can be summered as follows:(1) The thesis introduced the basic concept of BCI technology, the components and the classification of BCI system. And it also described the status of research for BCI system and the problems to be solved in current studies in details. Furthermore, the structure of the brain, the EEG production and the ERD/ERS phenomenon of motor imagery were discussed, which provided theoretical basis for the following research.(2) Experimental paradigms about motor imageries were designed to obtain the EEG data. And the thesis introduced the experimental equipments, and several issues which must be paid attention to. Then three subjects participated in experiments, after that, the obtained data were selected based on the subjective and objective evaluation.(3) Two feature extraction methods which are based on AR model and CSP algorithm were analyzed in depth. For the AR model, the coefficient estimation, order selecting and design method of AR model were discussed. For the CSP algorithm, the thesis studied about the basic principle, the selection of best feature and the implementation.(4) Two classifiers, which based on Fisher LDA and SVM, were discussed, including their classification principles and implementations. Then the research on the algorithms of feature extraction and pattern classification were carried out on the data both gained from experiments and the data of2008International BCI Algorithm Contests. After analyzing and discussing, the results showed that the classification accuracy rate can be up to a high value if appropriate feature extraction method and classifier were adopted.(5) On the basis of the EEG data analysis, a BCI system based on motor imagery was designed. The users could send the corresponding command to control the device by imagining the movement of left and right limbs.Then two application systems were developed. One was "Game of virtual car", the other was called "Active lower extremity joint rehabilitation equipment". After training, the best subject was selected to test the two application systems and he could control them freely. The result demonstrated that the system was feasible with great potential value.