Research On Feature Extraction And Pattern Classification Algorithm Of Motion Imagination EEG And Its Online Verification

The direct brain-computer interaction (BCI) of brain signals is a new human-computer interaction technology. The system can be controlled directly by the brain signal reconstruction. It can be used for military purposes, and provide auxiliary control for disabilities and normal people, so as to improve their quality of life. At present, the BCI has became a international cutting-edge research and application hotspot. The BCI based on EEG is a kind of important brain-computer interface, it involves many research subjects. But it has not been popular widely because of the technology not yet mature.This paper based on the motor imagination EEG, proposed a new feature combination method, and explores the effectively pattern classification methods of left and right hand movement. And then those methods are verified on the online system in this paper. The mainly work as follows:(1) The feature extraction method based on the characteristics of phase synchronization is studied, and the influence of phase synchronization combined with other characteristics on classification accuracy is researched. We use the feature of phase synchronization combined with wavelet coefficient energy and frequency band energy characteristics to have a classification of the EEG signals. After processing by Hilbert transform for EEG, we use the phase locking value (PLV) to extract the phase synchronization feature, which can be identified by SVM for test set. The classification recognition rate of phase synchronization combined with wavelet coefficient energy is studied. We use the Hibert transform to obtain the instantaneous phase of EEG and the PLV to extract the phase synchronization feature. The collaboration between the contralateral primary motor area and roof area and the correlation between the degree of collaboration and the degree of activation on brain regions.(2)The pattern classification methods based on the left and right hand movement imagination is studied. The classification accuracy of methods are analyzed and contrasted, and we verify the feasibility of support vector machine (SVM) and linear discriminant analysis (LDA) for motor imagination EEG. Experiments show that SVM and LDA both have a higher classification accuracy than other methods, and the correct classification rate and the average correct classification rate have obvious advantages. Pattern classification is a important link in brain-computer interface system. It relates to the performance of the whole system. The research provide a basis for the validation online of feature extraction and pattern classification.(3)The correlated algorithm is verified online. Based on the above content, we verified the feature extraction and pattern classification methods on the online system of brain-computer interface. We hope that this work can be beneficial for improving the performance of online system.