Based On P300and Motion Image Brain-machineinterface Research

Many people in the world suffer severe motor disorders and lose the ability to communicate with the outside through language expression or body movement, caused by spinal cord injury, muscle atrophic Amyotrophic Lateral Sclerosis and etc. Brain Computer Interface (BCI) is a system that collects electrical signal from the brain, and then uses some machine agorithmns to convert differnent brain states to control commands, thus realize the control over the computer or other external devices. At present, BCI technology may be the only way to communicate for the people with serious movement disorders. However, the research of BCI is still in the laboratory stage and the current BCI systems still have some techincal barriers in speed and recognition rate.The electical signal which has been applied to BCI can be roughly divided into two groups:one is evoked signals, such as Visual Evoked Potential, Event Related Potential (P300), Auditory Evoked Potential, etc; the other one is spontaneous rhythms, such as a rhythm, β rhythm and μ rhythm. The research of the paper focuses on the BCI technologys based on the evoked and the spontaneous signals respctively. That is, the Event Related Potential (P300), β rhythm and μ rhythm are respctively adopted in the BCI system.A P300potential detection algorithm is proposed in the paper, which combines the average superposition theory, Common Spatial Pattern (CSP) and multiple Support vector machine (SVM) classifiers. The proposed algorithm is evaluated on the dataset II for the third BCI competition in2005, and achieves a better experiment results with the average correct recognition rate of98%.In the research of the BCI based on spontaneous rhythms, Bayesian Linear Discriminant Analysis (BLDA) and Boosting classifier are applied to process dataset I of the BCI competition Ⅲ, and the choices of classifier, feature vector and the leads are compared and analysised, which will influence the final classification results. The research results demonstracted that the algorithm based on the Boosting classifier has a better classification performance with the highest recognition rate of93%and is easier to implement.