Intelligent Rehabilitation System Of Lower Limbs Based On SEMG And SEMG Compensation Of Functional Electrical Stimulation

Stroke is the first cerebrovascular disease that puzzles the daily life of middle-aged and elderly people.With the development of medical technology,the cure rate of stroke is getting higher and higher.However,the physical and mental health of many patients is still destroyed by the physical disability caused by its sequelae.Therefore,treating the sequelae of stroke and helping to restore the limb function of patients,especially the lower limb function,is still an important research content in the field of rehabilitation.The lower limb rehabilitation system based on surface electromyography and functional electrical stimulation electromyography compensation has inestimable research value in the rehabilitation field.The lower limb rehabilitation system can use surface electromyography to identify the lower limb movements of patients,and implement targeted electrical stimulation to specific muscle groups according to the patient’s movement state in the process of limb rehabilitation training,so as to help patients recover their limb functions faster.Among them,there is a mature and flexible system for lower limb movement recognition,but the research on functional electrical stimulation is still lagging behind,lacking the correct and scientific electrical stimulation feedback method for limb movement.This paper mainly studies the lower limb rehabilitation system based on surface electromyography and functional electrical stimulation electromyography compensation.The main work includes:(1)Research on lower limb motion recognition algorithm based on surface electromyography.Ten healthy subjects were selected to collect surface EMG data.After pretreatment,EMG features were extracted from time domain and frequency domain respectively.After optimization in recognition time,features,channels and classifier parameters,it was found that binomial kernel support vector machine had the highest recognition rate,with the recognition accuracy of 93.2% and 97.2% under the data length of 0.12 s and 0.2s and 8 channels respectively.In addition,two experimental paradigms are designed to identify and classify the musle contraction actions before the real motion starting,in which,the recognition accuracy of voluntary muscle pre-contraction and obstructive muscle contraction can reach 84.7%and 99.6% respectively.Finally,the algorithm is applied to online motion recognition,and the comprehensive recognition rate is 91.6%.(2)Knee joint motion modeling of lower limbs based on fusion biodynamics and non-negative matrix factorization.The surface EMG of four healthy subjects’ active muscles(quadriceps femoris and biceps femoris)and six auxiliary muscles in the process of knee joint movement were collected,and the biomechanical model and non-negative matrix factorization,which were formed by the fusion of intention-EMG,muscle strength-EMG and muscle length-EMG,were used to fit the active EMG and auxiliary EMG respectively.Finally,the results show that when using the fusion biodynamic model and non-negative matrix to simulate the EMG of the main and auxiliary muscles during the knee joint movement,the fitting error is low and the fitting effect is good,in which the average Root Mean Square Error(RMSE)of the active muscle is 1.76% and the average RMSE of the auxiliary muscle is 4.63%.Finally,the small-amplitude motion method and the cooperative pattern feature of the average template are proposed to build the knee joint motion model of lower limbs of specific targets.The final results show that it is feasible to build the knee joint motion model of specific targets with this scheme,in which the average RMSE of the active muscles is 7.68% and that of the auxiliary muscles is 12.16%.(3)Research on the mechanism of electrical stimulation compensation for EMG based on PID control algorithm.According to the time delay of the electrical stimulation system,an electrical stimulation simulation experiment is designed to verify the ability of PID control algorithm to suppress time delay and noise.The final result shows that PID algorithm has the ability to suppress time delay and strong noise.Then,the zero EMG experiment was designed as the zero template of EMG compensation experiment,and under the guidance of the specific target knee joint model of lower limbs,the EMG compensation based on PID control algorithm was carried out for the four channels of the experimenter’s biceps femoris,quadriceps femoris,tibialis anterior and gastrocnemius.The final result showed that the scheme could compensate the lower limb muscle groups by EMG,in which the quadriceps femoris muscle supplement was 70.1%,the biceps femoris muscle supplement was72.2%,the tibialis anterior supplement was 69.3%,and gastrocnemius supplement was 71.8%.(4)According to the obtained research results,the lower limb rehabilitation system was designed.A lower extremity rehabilitation platform was built using relevant hardware,and pure passive mode,active mode,and electrical stimulation compensation mode,and the related interactive pages were designed for these three modes and rehabilitation machine control.Good application results have been achieved in motor recognition and electromyographic compensation.