Eeg Recognition And Application Of Motor Imagery

Brain-Computer Interface (BCI) is a system which allows direct translation of brain states into actions, bypassing the usual muscular pathways. As a new kind of technology about communication and control, BCI can provide language exchange and environmental control for patients with severe dyskinesia. Besides, it has potential application value in the field of automatic control, military and brain cognitive. Because of its application prospects, international scientific community attaches great importance to BCI. It has become a research hotspot of brain science, rehabilitation engineering, biomedical engineering and human-computer control.Electroencephalogram (EEG) based on motor imagery is a commonly used thinking operation in BCI search because it does not need the structure of the physical environment. The accuracy and speed of EEGs recognition is the points of BCI from laboratory research to practical application. EEGs recognition based on motor imagery was discussed in this paper, and several research results are got as follows:Firstly, In brain-computer interface systems of motor imagery, an improved common special pattern algorithm was proposed to solve the problem that low classification accuracy in condition of EEGs collected with few channels. This method was based on that EEGs were filtered through a frequency band which event-related desynchronization /event-related synchronization physiological phenomenon was obvious. The best corresponding eigenvector of the biggest eigenvalue was selected which could describe the state of motor imagery EEGs. Then a new feature extracting manner was proposed .This improved CSP method was combined with Support Vector Machine(SVM) for the classification of motor imagery EEGs. The experiment results showed that the improved CSP algorithm could reflect the EEGs task state more exactly and avoid the repetitive eigenvector selection.Secondly, a new EEG identification method based on SVM Ensemble was proposed to solve the problem which low classification accuracy and weaker robustness for collecting EEGs during different time in Brain-Computer Interface systems of motor imagery. Bagging algorithm and cross validated committees were adopted in individuals generation of Ensemble SVM. Besides, Bagging method was improved, and factors which can affect the performance of the integration were analyzed. This method was used for the recognition of the EEGs based on motor imagery of finger and tongue during different time which provided by Tubingen University. The experiment results showed the accuracy of Ensemble SVM was better than that of single SVM for different time collecting EEGs, and cross validated committees and improved Bagging performed better than traditional Bagging method.Thirdly, software is designed for EEGs collecting, including three parts. In the first experimental, EEGs were collected when the subject was knocking at the keyboard under visual stimulation. In the second experimental, EEGs were collected when the subject was imaging the movement of left and right hand under visual stimulation.. In the third experimental, EEGs were collected when the subject was knocking at the keyboard under auditory stimulation. EEG amplifier produced by Symtop Instrument Corporation is selected for EEGs collecting. CSP and SVM algorithm were combined for the recognizing and analyzing the EEGs of the three experimental parts. The results showed that CSP algorithm can exclude the same task components and extract the different task components even in interference with electro-oculogram, so the classifier can get excellent classification performance. At last, BCI keyboard-input system was designed. In this system improved CSP algorithm combined with SVM were selected for recognizing actual collecting EEGs. The keyboard-input function based on thinking EEGs was realized. And the feasibility of the experiment and the validity of the EEGs recognition method are proved.In this paper, EEG processing and recognition was discussed and analyzed, and it builds contribution to promote the classification performance and generalization ability of BCI system and has reference value for realizing online BCI.