The Research Of Motor Imagery EEG Classification Algorithm And BCI Technologies

As a large branch of the BCI(Brain-computer interface)system,the motor imagery-based(MI-based)BCI has attracted more and more attention of people.However,The EEG signals are complicated as well as nonlinear and non-stationary,which make them hard to analyze.The recognition results are dependent on the datasets selected,and the performance is not stable.In order to improve the recognition rate of EEG signals and provide effective feature extraction and classification method,two methods of feature extraction and classification on two-category MI EEG signals are proposed.Beside,a principle on-line BCI prototype is developed.1.The ensemble empirical mode decomposition(EEMD)as a kind of efficiently adaptive signal processing method is firstly used for motor imagery recognition tasks because of the good decomposition resolution.An efficient EEMD-based feature extraction scheme is presented,which combined the Hilbert marginal spectrum(MS)and instantaneous energy spectrum(IES)features with window-based EEMD-based approximate entropy(ApEn)features.The impactful factors of IMFs and frequency bands are selected for the features as well.The linear discriminant analysis(LDA)classifier is designed for classifying.The method is tested on nine subjects.The results shows that feature combination we propose have a competitive performance in recognition rate with other methods on the same dataset.The mean accuracy of nine subjects is 82.74%.2.A new kind of efficiently adaptive signal processing method intrinsic time-scale decomposition(ITD)is applied for motor imagery(MI)recognition tasks in this paper.The whole system consists of the ITD decomposition,feature extraction,feature selection and classification.The features including energy,autoregressive(AR)model,the morphological features and fuzzy approximate entropy(fApEn)are extracted from the high frequency components of ITD.Then one-way analysis of variance(ANOVA)is used for feature selection and finally classified by LDA classifier.The results demonstrate the proposed ITD-based system can obtain competitive classification accuracy and precedes the other existing works and champions of the BCI Competition(BCIC)on mutual information(MI)criterion.The maximal MI of BCIC II dataset can reach 0.75 and the average maximum MI steepness of BCIC III dataset is 0.3699.Furthermore,the ITD is low computational complexity method and real-time implementable,which is more competent for online BCI devices.3.A ThinkGear-based online BCI system is introduced,which could perform the control of arm prosthetics.The system collected the forehead EEG information of subjects to obtain visual blocking information and translate into control signals.The system could control arm perform several actions.We hope the system may serve as a valuable base and reference for the further study.