Research On MI-EEG Analysis And Combined With VR Training For Rehabilitation

In recent years,the incidence of diseases such as Stroke,Muscular Dystrophy(MD),Amyotrophic Lateral Sclerosis(ALS),and Motor Neuron Disease(MND)has increased year by year.The brain consciousness of these patients is normal,however,the nerve or muscle pathways of the brain controlling the movement of the limbs are damaged to varying degrees,resulting in the dysfunction of the limbs' movement and the loss of the ability to live independently,which brings heavy burden to the family and society.Therefore,the rehabilitation of motor function of patients related to the above diseases is worth studying.Motor Imagery(MI)therapy refers to repetitive motor imagery in the brain without actual motor output,and to achieve motor function rehabilitation by activating specific areas of the brain based on motor memory.The Motor Imagery Brain Computer Interface(MI-BCI)technology is a type of Brain Computer Interface(BCI)technology,which guides patients to perform MI to identify patients' motor imagery intent by MI therapy,and effectively used to promote the remodeling of the central nervous system in patients with motor function.Virtual Reality(VR)technology can provide patients with a strong sense of immersion training environment,help patients perform MI better,and to produce more easily recognized Electroencephalograph(EEG).Aiming at the rehabilitation of stroke patients as the main population,this paper studies the analysis of motor imagery EEG in MI-BCI,the influence of VR technology guidance and feedback on EEG recognition performance,and the design and development of limb rehabilitation training system based on MI-BCI + VR.The research contents are as follows:(1)Analysis of motor imagery EEG based on Filter Bank Common Spatial Pattern: The core of MI-BCI is EEG analysis and recognition.An improved-FBCSP is proposed for specific subjects' motor imagery EEG.The EEG analysis process includes five phases: Time segment optimization of motor imagery,band filtering based on filter bank,common spatial filtering,feature selection based on mutual information,and feature classification based on SVM.The segment optimization phase selects the best MI time period for a specific subject.The two phases of frequency band filtering and CSP spatial filtering complete feature extraction.The feature selection phase selects the best individual features.The classification phase establishes a classification model based on the best individual characteristics and gives the final classification result of related EEG.The improvedFBCSP introduces time segment optimization,and the step of personalized parameter optimization is added in the band filtering phase at the same time,aiming at better application of FBCSP in personalized EEG analysis.Based on this method,the data of the training data set in the B-list of the MI task in the 3rd Brain-Computer Interface Contest was analyzed and processed.The results show that this method has achieved good results,which lays a foundation for better application of FBCSP algorithm in personalized rehabilitation treatment for specific patients.(2)Research on the influence of VR guidance and feedback on the performance of EEG recognition: Rehabilitation training based on MI-BCI and VR technology can be divided into two phases,offline and online.Participants need to generate EEG signals through active MI in both phases.The VR guidance and feedback will have an impact on the subjects' MI process,and then affect their EEG signal recognition performance.In order to explore the effect of VR guidance and feedback on the performance of EEG recognition in MI,four groups of experiments were designed with normal subjects: None VR guidance during offline phase,VR guidance during offline phase,none VR feedback during online phase,and VR feedback during online phase,and EEG signals in four cases were collected.Using the improved FBCSP proposed in Chapter 2 to analyze the collected EEG,it is concluded that both the training scenes with VR guidance in the offline training phase and VR feedback in the online phase can help the subjects to better exercise imagery in varying degrees,which made the EEG signals generated by the subjects much easier to recognize.Furthermore,the effects of different VR feedback scenes on EEG signals were studied.The result shows that the effect varies from person to person and is related to the physical and psychological factors of the subject.Therefore,a personalized training environment could be designed according to the needs of the patient when designing a VR rehabilitation training system.It laid a foundation for better application of VR technology in the field of rehabilitation combined with MI-BCI.(3)Design of limbs rehabilitation training system based on MI-BCI+VR: A set of limb rehabilitation training system is designed and developed by combining MI-BCI and VR technology based on the research of(1)and(2).The system divides the rehabilitation process into three phases:(1)Offline training phase: This phase consists of a simple training scene based on VR technology,which prompts and guides the patient to perform MI by the movements of hands and feet in the scene,and exercises their ability to perform MI and completes initial rehabilitation;(2)Online training phase based on prompts and feedback: During this phase,the patient is asked to perform the corresponding MI during the training and give feedback to further strengthen the patient's MI ability on the basis of the first phase,and complete the mid-term rehabilitation to restore the patient's simple limb movements in the process;(3)Game-based online training phase: This phase is composed of interesting puzzle games designed and developed based on VR technology.The patients might control the game objects proactively to complete some more complex actions through independent MI,and realize the later rehabilitation of more complex limb actions.The three training phases are progressive,guiding patients to gradually strengthen their MI ability,and promote remodeling of the central nervous system through continuous MI,and gradually restore limb motor function in order to obtain good rehabilitation training results.By experimenting with 8 healthy people,it shows that the system can operate normally.