Analysis And Processing Of EEG Signals Evoked During Motor Imagery

Electroencephalogram (EEG) is the overall reflection of brain nerve cells electrophysiological activity on cerebral cortex or scalp, which contains a large number of physiological and disease information. 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 and then forms a communication control system which don't depend on the normal brain peripheral nerve and muscle output channels, which is so-called brain-computer interface (BCI). 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 which we can judge the movement intention and control external device. So motor imagery EEG becomes the most frequently used feature signal of BCI.This thesis analyzes and discusses feature extraction and classification algorithms of motor imagery EEG, chiefly investigates feature extraction algorithm based on wavelet analysis such as local threshold optimizing in time domain, threshold optimize of discrete wavelet transform reconstruct coefficients in special frequency bands, wavelet package decomposition reconstruct coefficients combining with sub band energies, threshold optimizing of wavelet package decomposition reconstruct coefficients combining with sub-band averages and energies. The thesis classifies motor imagery EEG in temporal and frequency interested band respectively by classification function among disparity of energy, Mahalanobis distance discriminate analysis and support vector machine. All methods mentioned above are used to process the data offered by 2003 BCI competition. The results show that the highest right rate can be up to 90%. Thresholding optimize applied to wavelet package decomposition reconstruct coefficients combining with sub band averages and energies performs better temporal sensitivity and higher distinguish ability. The results are helpful for the design and implementation of follow-up BCI system.