Research On Online Motor Imagery Brain-computer Interface System For Lower Limb Rehabilitation Device

In recent years,as our country enters an aging society,the number of stroke patients is growing,which increases the burden on society.Therefore,it is necessary to study how to help stroke patients better perform rehabilitation training.At present,the rehabilitation training for the stroke patients is still dominated by the traditional methods.Which are not only boring,but also lack active participation of patients,resulting in poor rehabilitation results.As an emerging human-computer interaction method,the brain-computer interface(BCI)can activate the neuronal cells in the corresponding brain area of the patient when it is applied to the field of rehabilitation training,which improves the rehabilitation training effect of the patient through the neural plasticity mechanism.Therefore,an online motor imagery brain-computer interface system for lower limb rehabilitation devices is built in this thesis,and the following researches are conducted:First of all,three experimental paradigms of motor imagery are designed in this thesis,which are left and right hand motor imagery,left and right foot motor imagery,and sitting and standing motor imagery,to explore the classifiability of motor imagery EEG(MI-EEG)data under different mental tasks.Then the corresponding offline training experiments of motor imagery are carried out with 11 subjects,and the offline MI-EEG signal data is collected and preprocessed.The results verify the event-related desynchronization(ERD)phenomenon of left and right hand motor imagery and prove the good classifiability of the collected MI-EEG data.Secondly,the online MI-BCI rehabilitation system is introduced in detail from three aspects:hardware selection,software interface development,and algorithm implementation.The system can complete the acquisition and classification of real-time EEG signals,and then control the movement of the lower limb rehabilitation device according to the classification results,so as to assist the subjects in motor rehabilitation training.The feasibility of the system has been proved through the experimental test.Finally,a CNN algorithm fusing spatial information based on EEGNet is implemented.Combined with the data enhancement method proposed in this thesis,for the off-line MI-EEG data collected by each subject under three experimental paradigms,the algorithm achieves an average accuracy of over 94% and an average kappa coefficient of over 0.89 on the test set.For two subjects in the online test experiment,an average online accuracy of 71.3% and an average Kappa coefficient of 0.43 are also achieved.In addition,the average response time of the MIBCI system for online tests is 0.957 s,which verifies the real-time performance of the system.