Decoding Human Right And Left Hand Motor Imagery From EEG Single Trials

For a long time, humans have been hoping to communicate with outside or direct control external equipment through their own thoughts in the brain, and realize the exchange of information between brain and machine. Brain-machine interface technology has create a new bridge between brain and the outside world, that not dependent on surrounding nerve, exchange of muscle information and control channel.In this paper, on the basis of full understanding of the latest research progress at home and abroad, take left and right hand motor imagery as a starting point, and systematic research brain-computer interface technology of left and right hand imagine movement, use of different methods for offline to analysis the brain signal, and achieved certain results.After data preprocessing we analyze16electrode data at three aspects, including energy, nonlinear characteristics and chaotic characteristics, gained the temporal variations in different brain areas when left and right hand motor imagery. Then, we use three different types of classifiers, including the linear Fisher criterion classifier, k-nearest neighbor classifier and support vector machine (SVM) classifier, to classify the data of left and right hand motor imagery, and achieved satisfying classify results.Through the experiment data calculation and analysis, the following conclusions can be drawn:In a static state as the reference data, calculates the power spectrum and Hurst index of their corresponding time-varying numerical, These values can be a good characterization oi" change rule as left and right hand motor imagery. The right hand motor imagery, the cortex area of the left brain becomes active after a period of active cortical regions decreased, while the right hemisphere cortical regions become active when the left hand motor imagery, but no right-hand imagination motor imagery as obvious, But there is no such obvious as right hand motor imagery. In the classification process, using the combining method, including the power spectrum and the linear Fisher criterion classifier, sample entropy and k-nearest neighbor classifier, can get very good classification results. Its maximum accuracy is over80%. If using personalized design, maximum accuracy can be improved by4%. In addition, relatively speaking, the men correct classification rate is significantly higher than women, while the data they need longer.