Research On The Control Of Desktop Upper Limb Rehabilitation Robot Based On SEMG Motion Recognition

With the aging of the population,there is a sharp rise in the number of patients with motor dysfunction of the upper limbs caused by cerebral apoplexy.At present,using rehabilitation robots to carry out rehabilitation training for hemiplegia patients has become a research hotspot at home and abroad.However,the existing functional design of the upper limb rehabilitation robot mostly focused on mechanical and passive training which can not reflect the movement intention of patients.Also it lacks the research on active and passive control strategies for different stages of rehabilitation,which will decrease the active participation of patients and the general applicability of the rehabilitation system.To solve the problems,the dissertation starts from the system design of the upper limb rehabilitee robot,and the sEMG signal and electrical signal which can reflect the human movement intention are introduced.The passive position control strategy based on sEMG action recognition and the active softness control strategy based on contact force are studied respectively and finally to achieve the experimental verification of this method based on the construction of upper limb rehabilitation robot.The main contents of the thesis are presented as follows:(1)The design of the desktop rehabilitation robot system is introduced as follows.The mechanical body and hardware control system of the desktop upper limb rehabilitation robot are designed based on the physiological structure characteristics of the human upper limb and the actual application requirements of the robot.And according to the different rehabilitation stages of the affected limb,the advanced rehabilitation training scheme is proposed.The passive rehabilitation training program based on sEMG motion recognition is used in the period of soft paralysis and the active rehabilitation training program based on contact force is adopted in the convalescence period.(2)The research of motion recognition technology based on healthy upper limb sEMG in the soft paralysis period is introduced as follows.Firstly,this dissertation determines the 4-layer wavelet threshold denoising method based on the db4 wavelet base by comparing and analyzing the denoising effect of sEMG by wavelet and filter and notch filter.Secondly,to solve the problem that the features of lower-order statistics such as time-domain and frequency-domain can not represent the non-Gaussian information of signal,the higher-order statistics based bispectral feature extraction method is studied,and the best separable fusion feature DBS+MAV+MPF(DMM)is obtained by analyzing DBI index of the joint feature space;Finally,a PSO-LM-BPNN classifier is designed to achieve the accurate classification of the optimal fusion feature DMM.This classifier can overcome the problem of traditional LM-BPNN long training time and easy to fall into the local extremes.And it can realize the effective recognition of four action patterns with an average recognition rate of 95%.(3)A study on active and passive control strategies for rehabilitation training programs in soft paralysis and convalescence is introduced as follows.In this dissertation,passive position control based on sEMG motion recognition is studied for the period of soft paralysis and establishes the mathematical model of target motion trajectory.Besides,by improving the P_c and P_m adjustment formulas in the traditional genetic algorithm,a position controller based on the improved adaptive genetic algorithm is designed to optimize PI,which improves the accuracy,stability and speed of the control system and can drive the affected limb along the target trajectory for autonomous and passive rehabilitation training.The active compliance control based on the contact force is studied during the recovery period.To realize the rapid and accurate tracking of the force signal of the affected limb,a new anti-integral saturated PI controller is designed based on the impedance control theory,and the damping coefficient is adjusted to apply to patients at different stages of recovery.(4)To verify the application effect of the advanced rehabilitation training scheme and the corresponding control strategy,the passive and active rehabilitation training verification experiments are designed based on the rehabilitation robot platform.The experimental results show that the desktop-type upper limb rehabilitation robot system can realize the advanced rehabilitation training program.The control effect is ideal,which can effectively improve the active participation of patients.