The Research Of EEG Recognition Based On Coordinated Motor Imagery

If patients with motor dysfunction caused by brain injury can carry out autonomous rehabilitation training in time,the injured nervus can be repaired or reconstructed,and patients' motor function can be improved mostly.However,it's usually very difficult for patients to complete rehabiliation training with their affected limbs independently.Neuroelectrophysiological phenomena induced by motion imagery in the cerebral cortex is similar to actual motion,and motion imagery can activate motor-related cortex as effectively as actual motion,so that patients can participate in rehabilitation training through motor imagery.Moreover,the brain-computer interface system based on motor imagery can be used to assist the affected limb to motion,so as to achieve the goal of autonomous rehabilitation training.It's of scientific value and practical significance to study motor imagery and pattern recognition of motor imagery electroencephalogram(EEG).According to the research status of motor imagery,in this paper,motor imagery was divided into simple motor imagery and coordinated motor imagery,and to analyse the cortical activity and the separability of EEG signals of this two types of motor imagery.The focus of this study is:to analyse the cortical activity of coordinated motor imagery and simple motor imagery,and to recognize motor imagery EEG signals of different motion modes.The main research contents and innovations are as follows:(1)Motor imagery is divided into simple motor imagery with single-limb participated and coordinated motor imagery with multi-limb collaborative particpated,enhancing the number of motion modes of motor imagery in analysis.According to the correspondence between limb motion position and cerebral cortex,three simple motion modes and three coordinated motion modes are designed.Three simgle motion modes are left-hand lifting,right-hand lifting and left-foot lifting.Three coordinated motion modes are two-hands cooperative lifting,left-hand and left-foot cooperative lifting and right-hand and left-foot cooperative lifting.The silent state is as contrast The experimental paradigm of EEG signals acquisition for motor imagery is designed,and EEG signals acquisition for seven kinds of motion modes(including silent state)is completed.(2)Event related desynchronization(ERD)value and brain functional network measures are used to analyse the cerebral cortical activity.Coordinated motor imagery with multi-limb collaborative participated involves cooperative work of multiple brain regions of corresponding limb positions.ERD value can describe the activity of local area of the cerebral cortex,but it can't reflect the functional integration between different brain regions.Brain functional network can describe functional connectivity between brain regions,so it's suitable to analyse the global activity of cerebral cortex of coordinated motor imagery.The experimental results based on ERD value analysis show that:the local activity of coordinated motor imagery at C3,C4 and Cz leads is higher than that of simple motor imagery;the experimental results based on brain functional network measures analysis show that:the global activity of cerebral cortex of coordinated motor imagery is higher than that of simple motor imagery,validating the hypothesis that coordinated motor imagery can increase the number of motion modes to be recognized.(3)A feature extraction method based on functional partition brain functional network is proposed The sum of connection coefficients between any two network nodes in the weighted brain functional network is defined as the network functional connectivity value,and the absolute value of the ratio of network functional connectivity value during imagery period to that during rest period before imagery is defined as network functional connectivity increase rate.According to the clustering differences of network topology near C3,C4 and Cz leads of different motion modes,the cerebral cortex is divided into three functional regions centered on C3,C4 and Cz leads,respectively.Computing the network functional connectivity increase rate of each region,form a three-dimensional feature vector.Using the three-dimensional feature vector as input vector,seven kinds of motion modes are recognized by multi-classification support vector machine.The result shows that the average correct recognition rate of each mtion mode is higher than 83%,indicating that the feature vector of network functional connectivity increase rate can better represent motion imagery EEG signals.The research work of this paper has achieved the expected results:1.By adding the experimental paradigm of coordinated motion imagery,the number of recognizable motion modes has increased;2.By defining brain functional network functional connectivity value and network functional connectivity increase rate feature,the problem of multiclass motion modes recognition was well solved.