An Assistive Full Arm Robotic Exoskeleton Based On EMG Control

Brachial plexus injury(BPI)is one of the most serious peripheral nerve injury,and the recovery of upper limb movement function for patients after surgery is very limited.So,the patients who suffer from BPI often accompanied with a permanent disability of upper limb.In order to help BPI patients rebuild their upper limb function,this study propose an assistive full-arm robotic exoskeleton based on electromyographic(EMG)for BPI patients in the first time around China.The full-arm robotic exoskeleton is mainly composed of mechanical structure,control module and algorithm.In the design of structure,the exoskeleton provides 5 degrees of freedom(DOFs),including 2 DOFs for hand exoskeleton,1 DOF for elbow exoskeleton and 2 DOFs for shoulder exoskeleton.Wearing the designed full-arm exoskeleton,patients are able to complete 9 kinds of upper limb movements.The system adopts the tendondriven method,which separates the body part of exoskeleton from the dive part and reduces the burden of patients' upper limb when using the exoskeleton.The control module consists of hardware circuit and software program.The hardware circuit includes a main controller module,a power module,an electromyogram signal acquisition module,and a motor drive module.The exoskeleton control system is based on two control algorithms: PID control based on reference trajectory and pattern recognition control based on myoelectric signals.In the pattern recognition control,by extracting the patient's electromyographic signal after the nerve transposition and using the linear discriminant analysis method based on Bayesian decision to decode its motion intention,the patient can rebuild upper limb function without any help.PID control experiments based on reference trajectories verified the feasibility and accuracy of the exoskeleton system of the upper limb exoskeleton.By the pattern recognition control experiments based on the EMG signals,the results of the myoelectric signal recognition of 9 types of motions in 4 healthy subjects were verified.The results showed that the average online recognition rate of 9 types of motion reached more than 97%.Through sampling analysis and human-computer interaction experiments,it is shown that the on-line recognition process for each type of action takes less than 85 ms,which fully satisfies the delay requirements in the use of the exoskeleton,and subjects are able to control the designed exoskeleton system smoothly.