Recognition And Real-time Control Application Of EEG Signal Based On Motor Imagery

Interpreting the real behavior intention of patients with language loss and limb paralysis is very important for patients to interact with the external environment.Brain computer interface(BCI)technology can realize thinking decoding and greatly improve the quality of life of patients.In addition,due to the plasticity of brain nerve,the effect of rehabilitation training with the participation of motor imagery EEG(MI-EEG)signal is better.Therefore,it is important and significant to research on MI-EEG signal processing,classification and recognition and real-time application.The classification algorithm of MI-EEG signals of bilateral lower limbs based on Open Vi BE is studied.Aiming at the shortage of noise sensitivity and easy over fitting in the common space pattern(CSP)feature extraction algorithm,Tikhonov regularization(TR)is used to optimize and improve to construct Tikhonov regularized common space pattern(TRCSP)filter.Support vector machine(SVM)and linear discriminant analysis(LDA)algorithms are used to complete classification and recognition.The performance of TRCSP+SVM,TRCSP+LDA,CSP+SVM and CSP+LDA were tested and compared based on statistical methods.Finally,the classification performance of TRCSP+SVM method has achieved the best effect and is significantly better than other combinations.The average accuracy is as high as97.28%±0.97%,the average kappa coefficient is as high as 0.91±0.06,and the average AUC value is as high as 0.98±0.01.Compared with CSP+SVM method,TRCSP+SVM method has 10% improvement in accuracy.It is fully proved that the classification method combining the improved TRCSP feature extraction method and SVM algorithm has advantages in improving the classification and recognition rate of lower limb MI-EEG signal.According to the regulation mechanism that different polarity pulse stimulation signals acting on the central pattern generator(CPG)site can induce the reversal of gait patterns of both lower limbs.A gait controller with fewer stimulation electrodes and relatively simple timing,combined with the motor imagery brain-computer interface(MI-BCI)and functional electrical stimulation(FES)technology,was designed and a“What you think is what you move” limb rehabilitation training mode is constructed.TRCSP+SVM recognition method is selected as the control strategy of the whole online rehabilitation training system.The result data of classification and recognition will be stably transmitted to the gait controller through the virtual reality peripheral network(VRPN)protocol to generate the corresponding instructions to trigger the stimulation signals of different modes.When the subjects performed the task of imagining “Moving the left leg”,the lower limbs of rats was actived gait movement of left flexion and right extension;When the subjects performed the task of imagining “Moving the right leg”,the lower limbs of the rats was actived gait movement of right flexion and left extension.Successfully achieve the desired movement of lower limbs corresponding to the subject’s intention under the action of FES.In order to test the function and effect of motor imagery(MI)based paralyzed double lower limb real-time motion control system in rehabilitation training,Three-week rats’ experiments were carried out.Compared with SCI group,the score of BBB scale in SCI+MI-ES group was about 4 points higher,the maximum grasping force of hind limb was about 5N higher,and the recovery of lower limb motor function was more significantly improved.The effectiveness of the gait control system of paralyzed lower limbs based on MI in the application scene of rehabilitation training is verified.It provides new ideas and methods for hemiplegic patients with SCI to reconstruct motor function,and pushes forward the practical process of MI-BCI system in the field of rehabilitation medicine.