Research And Design Of Surface Electromyography Control System Of Upper Limb Exoskeleton

Stroke is a disease that causes motor cortex damage due to the rupture of cerebral blood vessels and compression of nerves,which can lead to hemiplegia,that is,impaired motor function of unilateral limbs.Manual treatment is often used in clinics,which causes a lot of waste of medical resources and increases the burden on medical staff.Current researches mostly use exoskeleton instead of artificial auxiliary treatment.Its control system mainly uses the signals collected by force and angle sensors as the core for passive or impedance control.This process requires the installation of multiple sensors,and the overall mechanism is complicated.Moreover,the patient's intention to exercise cannot be accurately obtained,causing problems such as loss of confidence in the training process.Therefore,this article researched and designed the upper extremity exoskeleton surface electromyography control system to decode the human movement intention and provide accurate movement commands for the exoskeleton,so that the patient can actively participate in and complete the exercise training efficiently.The main work content is as follows:(1)Firstly,the overall plan of the upper extremity exoskeleton surface electromyography control system was designed.Aiming at the problems of single degree of freedom of the electromyography(EMG)exoskeleton and insufficient flexibility,this project imitates the joint structure of the human arm to develop an upper limb exoskeleton with 5 degrees of freedom,which can complete 7 upper limb movements of the shoulder,elbow,and wrist.Mechanical and programmed limits are designed for the exoskeleton to avoid causing a second-injury to the patient.The length of the exoskeleton is adjustable,and the left and right sides can be flipped.The driving device adopts an integrated direct current(DC)servo motor with large torque and small volume.The hardware circuit design of the system mainly includes the main controller module,human-computer interaction module,motor drive module and electromyography acquisition module.The software design includes electromyography acquisition,interface display,data processing and motor control programs.(2)Secondly,conduct in-depth analysis of the surface EMG signal.After familiarizing with its generation mechanism,start the collection work.Determine the position of the EMG electrode by analyzing the upper limb muscle group,and select four muscles for collection based on the rehabilitation movement target action.The EMG signal denoising adopts a fourth-order Butterworth filter,the window segmentation adopts the overlap method,and the active segment detection adopts the moving window waveform length method.Through comparative analysis,five time domains and two frequency domains are finally selected for feature extraction.The interval scaling method normalizes the characteristic data.(3)Then,in view of the limitations of EMG interface technology and processing methods,the traditional EMG control method cannot control the problem of multi-degree-of-freedom exoskeleton.This paper studies the pattern recognition algorithm based on surface EMG control to realize the multifunctional upper extremity exoskeleton.control.The characteristics and performance of three common pattern recognition classifiers are studied,and decision Trees-Support vector machine pattern recognition classifier based on greedy algorithm grouping optimization is designed to complete the construction of training upper limb movement recognition classifiers,and finally obtain the recognition of human upper limb movements model.(4)Finally,after the construction of the upper extremity exoskeleton surface electromyography control system is completed,the experiment will be verified.5 subjects completed 7 online experiments of upper limb movements in sequence according to the training indicator on the control panel.The experiment was carried out 350 times.The number of successful and correct driving of the upper limb exoskeleton was 340,the success rate was 97.1%,and the online recognition time of upper limb movements in the 7 experiments is less than 85ms.The system has a high success rate and short recognition time.The experiments show that the designed greedy algorithm for optimal decision-making binary tree classifier performs well.