Electroencephalogram Classification And Brain Computer Interface Research Based On Motor Imagery

Brain computer interface (BCI) is a communication system that does not depend on the brain's normal output pathways of peripheral nerves and muscles. This technology becomes the hot issue of the biomedical engineering research field in nowadays, because it has great value in theory research and practical application. Classification the electroencephalogram (EEG) recorded during different motor imagery is a main branch of BCI research.According to the Natural Science Foundation of Hebei Province under Grant E2006000034, experiments to record the EEG during different motor imageries were designed. Classification based on power and complicity was made by using the EEG data gained from experiments. Then, an on-line BCI system was established and this system was attempted to be put into real application. The main points of this paper are:1. In order to fulfill the demand of the research, experiments to record EEG during imagery left hand movement, right hand movement and foot movement were designed.Based on the study of EEG production and the characters of motor imagery EEG and the consideration of EEG recording system's functions, the experiments were projected. Three different EEG of motor imageries were gained though experiments and these data were used in the following research.2. Power character of EEG was extracted and used in classificationEEG was provided as a stationary signal for the request of on-line BCI research. Fourier transform was used to extract power feature at peculiar frequency band, and the feature was used in classification. The method to extract power character, the measure to find the peculiar frequency band and the results of classification will contribute to the design of an on-line BCI system.According the fact that EEG is non-stationary and non-linear, time-frequency analysis with Hilbert-Huang transform (HHT) was made. Power feature comprehensive considered about temporal-frequency-spatial information was used in classification. Features extracted by the HHT were optimized using filter and wrapper hybrid method. The results of classification proved the validity of the feature extract method and optimize method. 3. Non-linear complexity of EEG was extracted and used in classificationA non-linear dynamic method called Sample Entropy (SampEn) was applied to extract the feature of EEG signals from different motor imagery. Support vector machine used as classifier. The result shows that it can reach a better effect for motor imagery's classification to extract the feature by sample entropy. The novel idea which combined optimize the feature subset and parameter of the classifier was investigated in classify research. It provided a route to resolve the contradiction between the speed and accuracy during classification. A better classification result was gained.4. An on-line BCI system was establishedBased on the analysis of off-line data, an on-line BCI system using motor imagery EEG was designed. Aimed at the problem that present on-line BCI system can not recognize"idle"and"active"state, a method using the fact that close eyes can increase amplitude of alpha wave as a sign to switch different state was provided. A module which function is to monitor the state and switch different state was designed for the system. The testing result shows that, after a period of training, the subject can switch freely among different states by using this on-line BCI system. Control commands were sent out by the subject through this system at an especial high accuracy. An exploration research about the on-line BCI system's real applications was made. A mini electric car which simulates a wheelchair was controlled through this on-line BCI system.