Analsis Of Motor Imagery EEG Based On The BP Network

With the implementation of the Human Brain Project and Neuroinformatics gradually rise, while promoting the awareness of people on the brain. Thus it has also gradually in-depth study of the Electroencephalogram (EEG) is the overall reflection of brain nerve cells electrophysiological activity on cerebral cortex or scalp. Because there are correlation between EEG and conscious state of brain, it is possible to identify different conscious states by classification of different EEG models. This paper mainly analyzes EEG of motor imagery. Motor imagery means imagination of limb motor without actual physical action.EEG evoked during motor imagery has the characteristics of event-related desynchronization(ERD) and event-related synchronization(ERS). By EEG of motor imagery feature extraction and classification we can judged the movement intention to realize its value for the brain-computer interface device control.This paper mainly analyzed and discussed feature extraction and classification algorithms of motor imagery EEG, focusing on the extraction method of the characteristics based on the AR model and classification based on the identification algorithm of BP neural network. Using the data is the fourth brain-computer interface (BCI) of an international competition,and experimental processing with Matlab. Details are as follows:First, this paper overviewed the research status of the EEG, and EEG studies the impact on education.Secondly, the paper described the structure of the brain and EEG characteristics and classification of motor imagery EEG based on theoretical preparation for the experimental study of this paper.Thirdly, it introduced the basic theory of Independent Component Analysis, and described the experimental data of motor imagery, and used the FastICA algorithm denoising preprocessing by eeglab platform.Fourth, this paper introduced the basic theory of the AR model, and according to the theoretical basis, its model order and feature extraction for the experimental data are prepared for the next classification. Fifth, this paper described in detail the model structure and learning rules of the BP neural network, used BP neural network to train and test the experimental data of feature extraction in Matlab, and compared with other algorithms to prove that the BP network to get better classification results.