Research On Brain Computer Interface And Brain Network Of Motor Imagery Electroencephalogram

Motor imagery is a very important paradigm for brain-computer interface(BCI)technology.The motor imagery based BCI system can perfectly combine the human imaginary thoughts and the practical control applications.When used in the rehabilitation trainings for patients with brain injuries,the motor imagery based BCI technology is believed to improve patients' capability for limb movement through the neuroplastisity,so it has great application potential in rehabilitation clinics.However,it still has many technical difficulties to be solved in clinical use of motor imagery based BCI system,such as how to solve the problem of individual differences,how to improve the system's transmission rate under the premise of better stability and reliability,how to reduce the subjects' training time,and how to optimize the BCI system from the view of neural mechanism,etc.So in this thesis,a practical and useful motor imagery based BCI technique was explored,and based on both the off-line and on-line analysis,we studied on signal processing methods which were suitable for different individuals,established a real-time online BCI system,and solved the recognition problem of brain's idling state in the online system.what's more,the characteristics of the brain network were studied to reveal the brain mechanism of motor imagery.Firstly,in our thesis,four commonly used feature extraction methods of motor imagery Electroencephalogram(EEG)signal were studied,power spectrum estimation method of AR model in frequency domain analysis,continuous wavelet transform and wavelet packet decomposition methods in time-frequency analysis,and Hjorth parameter method in time domain analysis.Then aiming at the problem of individual differences in EEG features,an algorithm based on mutual information and principal component analysis(PCA)for EEG feature selection was proposed.Based on the features extracted by the above four feature extraction methods,the proposed feature selection algorithm was adopted to select and combine the most useful features for classification.The results showed that compared with the PCA algorithm,our algorithm had better performance in dimensionality reduction and classification accuracy with the assistance of support vector machine(SVM)classifier under thesame dimensionality of principal components,and it provided a better way for determining suitable feature extraction method and effective features for individuals in the online BCI system.Secondly,a real-time online BCI system based on the motor imagery paradigm of left and right hand was established,using the rocker upper-limb rehabilitation robot as the external control device,and the accumulated continuous classification algorithm was proposed to recognize brain states of idling and imagining in this system.Moreover,the feasibility of the system was proved by preliminary experiments,and its potential application value in rehabilitation training was also demonstrated.Finally,based on the above online BCI system,the brain network of motor imagery was established by the directed transfer function method,and the brain effective connectivity among every channel of EEG during motor imagery process was estimated in our thesis.The relationship between the subjects' BCI control ability and their brain network was studied and analyzed from the network parameters such as node degree,clustering coefficient,characteristic path length,local efficiency and global efficiency,and the mechanism of differences between subjects in BCI control ability was revealed.What's more,a new motor imagery paradigm based on joint action was proposed,and from the causality measurements such as causal density and causal flow,the causality among network nodes and its spatial distribution characteristics were finally found to the brain network of joint action motor imagery,which were different from single action motor imagery.on the whole,the above understanding of the neural mechanism of motor imagery provided a foundation for further new designing of paradigm,optimization of BCI performance,application in clinical rehabilitation training and so on.