The Research Of Brain-Computer Interface Based On Motor Imagery

Brain-Computer Interface (BCI) is a technology that used to realize the communication between human thoughts and computer or other electric instruments,which typically operates by EEG signals without depending on the brain's normal output pathway of peripheral nerves and muscles. For those with severe disabilities (e.g., spinal cord injury, amyotrophic lateral sclerosis or brainstem stroke etc.), a BCI may be the only feasible method for communicating with others and for environmental control. BCI technology is highly valued both in theory and application area of rehabilitation.Based on the knowledge of national and international BCI development, the work for this paper systematically studied a BCI that was based on the left-right hand motor imagery. It discussed both offline problems and real-time online problems, and made several significant improvements.Due to subject specificity, a general feature extraction method is not possible to acquire high classification ratio from all subjects. To solve this problem, the paper proposed three feature-analysis methods:"EEG frequency-energy-difference brain topography methods during left-right hand motor imagery based on time-frequency analysis";"EEG complexity-difference brain topography methods during left-right hand motor imagery based on sample entropy analysis";"EEG feature-difference brain topography methods during left-right hand motor imagery based on high-order cumulant analysis". Those methods successfully realized the subject-specific design, which enabled acquiring higher classification accuracy with least number of leads and least number of features, meanwhile simplest feature extraction and classification algorithm.The paper systematically discussed the feasibility of taking the frequency energy, sample entropy, and 4th-order cumulant as the extraction feature for the left-right hand motor imagery classification, then developed fast feature extraction algorithm and optimum parameters selection algorithm for each of the three features. Five classifiers were proposed and discussed, including their classification principle, design method, advantages and limitations, and applicable conditions. The work of this paper provided a foundation for the BCI selection of best classifier and optimum feature parameters both theoretically and experimentally. To further improve the efficiency of the BCI system, three factors that can influence the classification accuracy were studied, which were the imaginary content design, EEG signal acquiring electrodes selection, and training data selection. The experiment results showed that higher-complexity left-right hand motor imagery could enhance classification accuracy. Study also showed that compared with traditional electrodes, tripolar-ring Laplacian electrodes had higher signal to noise ratio, better spatial resolution and sensitivity, thus was more suitable for the regional-functioned motor imagery EEG acquiring and real-time online BCI system. What's more, syncretic decision-making algorithm based training data selection could remove those bad signals caused by subjective factors, benefite the performance of classifiers, and consequently improve the efficiency of the BCI system.The real-time online BCI system based on left-right hand motor imagery was established, and syncretic decision-making algorithm was used in recognition of vacant condition, which made the BCI system could work continuously. Based on this online BCI system, two application systems were also developed. One was"BCI based rehabilitation-assisting system for paralytic", the other was"BCI based environmental control system". Though those two systems were on their early stages, experiments had shown that they are feasible with great potential value.