Classifying Human Hands Motor Imagery From EEG Data

Brain-computer interface is a kind of information transfer channels between brain and external devices, this technology does not depend on brain peripheral nerves and muscles. Brain-computer interface technique has huge application prospect, has become a research hotspot in the field of biomedical engineering. The study of left and right hand motor imagery in this paper is one branch of BCI technique. It’s should be noticed that, brain-computer interface is not translating the human mind directly, it use some human physiological phenomena and biological electrical signals.At first, introduce the history of BCI technique in this paper, then state the development of BCI and the experiment scheme. According the scheme I get enough data from the experiment, and analyzed the data by many methodsDuring data processing, three feature extraction algorithm have be used to extract the feature during the task of left and right hand motor imagery is running. The power feature can clear show the area where the mind is activated by the motor imagery. But it take too long time to compute the power feature. Approximate entropy and Hurst exponent can delegate the complicacy feature and chaos feature.Classify the feature signal after the procedure of feature extraction by several classifiers, and the results indicate that power feature can cooperate Fisher classifier well. The approximate entropy algorithm with KNN classifier, Hurst exponent with SVM classifier can play to their strengths.The paper also analyzed the gender differences between men and women on the calculated result, it show that the same algorithm on the average classification accuracy of the male is higher than female. This result can be used as reference by other research project. The differences between individuals have large influence on the results, it shows that a powerful BCI system need fully consider the personalized settings.