Proportional And Simultaneous EMG Control For Multi-DOF Prosthesis

The surface electromyography(sEMG)signals are superposition of action potentials on the surface of the skin generated from muscular contraction,related to physical movement,and can be used to reflect human intention.Due to the advantages of rich information and convenient collecting,it's used widely in the areas of motion control and rehabilitation medicine.Aiming to proportional and simultaneous control of myoelectric prostheses,experiments are designed to acquire synchronously sEMG and kinematics parameters.Human coordination control rule from the central nervous system(CNS)was explained by analysis of relationships between synergies,activity coefficients,joint angles of upper limb based on muscle synergy theory.Muscle synergy model was used to solve the problem of proportional and simultaneous control of myoelectric prostheses in this paper.(1)Compared with pattern classification,continuous estimation can realize proportional and simultaneous control which is more accord with human control rule.Based on human forearm movement patterns in this paper,independent and combined motion of three degrees-of-freedom(DOF)experiments of upper limb were designed,which applies three dimensional motion capture system to synchronously collect forearm sEMG signal and joint angles for further research.(2)Muscle synergy theory was introduced to further explain and solve the problem of multi-DOF motion coordination in human movements.Six synergies of independent movement corresponding to upper limb are determined in divide and conquer approach,synergies which means low-dimensional control parameters of the CNS were extracted from root mean square feature based on non-negative matrix factorization(NMF).And then the activity coefficients of synergies were calculated using non-negative constrained least-squares algorithm.By comparing the activation coefficients and the joint trajectory,the activation coefficients can reflect actions type and the muscle activation.However,activation coefficients are not completely identical to the true trajectory of joint motion,and activation coefficients of actions are not fully decoupled.(3)In order to further solve the problem of the deviation between the activation coefficients and the motion trajectory,and to decompose coupling of each action,.Support vector regression(SVR)algorithm based on artificial bee colony and BP networks were used to construct an activation model which maps the activity coefficients into corresponding joint angles.Continuous estimation for joint movements from collected sEMG could be acquired by the above activation model.The experiment of estimation for independent and combined motion of two joint show that SVR activation model can achieve better estimation performance than BP activation model.For analyzing the relationship between the performance of SVR activation model and the speed of actions,the movements of the palm and the wrist are designed in the high,middle and low speed mode respectively.According to the correlation coefficient,root mean square error and t-test results,we found that SVR activation model can obtain good estimation results under high speed and medium speed.(4)An online proportional and simultaneous control simulation experiment of the myoelectric prostheses was carried out.Independent and combined motion online experiments were used to test proportional and simultaneous control performance of single DOF and multi-DOF by SVR activation model.This study provides a new control method for myoelectric prostheses.