Research On Exoskeleton System For Postoperative Wrist Rehabilitation Based On The Control Of Electromyography

Long term fixation after wrist joint surgery often causes skeletal muscle system rigidity and motor neurasthenia,which eventually results in decline of motor ability.In comparison with the traditional treatments,such as second operations and ultrasound therapy and other methods,wrist rehabilitation exercise is less costly,effective and safe.In traditional rehabilitation exercises,physical therapists are the main role and patients the support roles,so the sense of participation for patients is weak,which make the rehabilitation results ineffective and costly.Therefore,it is of great significance to develop a wrist rehabilitation training robot as an alternative.This paper focuses on four aspects: mechanical design of wrist exoskeleton,the algorithm for sEMG-angle estimation,control methods based on electromyography and related experiment.Based on the study of both wrist motion characteristics and anatomical structure,the requirements for mechanical design of exoskeleton are proposed.The mechanical design of the 3-DOF exoskeleton contains the overall structure deign and the motion transmission.This section also presents the joints misalignments between exoskeleton and wearer in the exercise of wrist flexion/extension and wrist ulnar-deviation/radial-deviation and give analysis to significance of joint misalignment compensation.In the end,a joint misalignment compensation mechanism based on cable actuation was designed.The experimental platform for the angle and sEMG data collection is build and the obtained data are used for training and validation sets making.Then on the basis on extracted time-domain sEMG features,three feature processing methods,namely,Relief F feature selection,feature resampling and feature time-lag compensation,are proposed,and two models are also introduced to verify the effective of the feature processing methods.Finally,two sEMG-angle estimation algorithms are designed,which are wavelet neural networks using Relief F selected features and least squares support vector machines using resampled and time-lag compensated features.On the basis of the wrist motion recognized by the EMG signals,the rehabilitation trajectory is selected and a PID controller is designed to assist human-machine to follow the trajectory and perform the the passive rehabilitation.Then,based on the sEMG-angle prediction algorithm,an embedded PID controller with feed-forward compensation is designed for active rehabilitation training.Finally,an adjustable elastic impedance control based on the sEMG feedback is also designed for impedance rehabilitation.The hardware and software systems of rehabilitative robot are also introduced,and the experiment for mechanical transmission verification and joint misalignment compensation experiments are also conducted.Then,three rehabilitation trainings,namely,the passive rehabilitation training experiment,the active rehabilitation,are carried out.The experimental results verify the reliability of the mechanical structure and the feasibility of the control method,which can meet the needs of stage rehabilitation for patients after wrist surgery.