Research On Intelligent Prosthetic Hand On Machine Learning Algorithm

For people with hand disabilities,the loss of a limb took a toll on them both physically and psychologically.As a supplement to the residual limbs of patients with physical disabilities,prosthetic hands were a tool to help them to improve their lives and adjusted their psychological conditions,whether they had practical functions or not.In the recovery process of amputee patients,intelligent prosthetic hands could not only help patients reintegrate into society,but also help disabled patients heal psychological trauma,became an important tool in patients’ daily life,replaced their original limbs to play a control function,and became a daily.It was an important tool in life and had very important social and practical significance.For this reason,this paper started from the intelligent development of prosthetic hands,used machine vision technology,sensor technology,machine learning technology and circuit design as technical means,and used self-made myoelectric signal data sets and visual image data sets as experimental objects.Designing and manufacturing a visual-myoelectric prosthetic hand structure that was controlled by machine vision-assisted myoelectric prosthetic hand,and exploring the potential of machine vision and machine learning in the intelligent design of prosthetic hand.The main work of this paper was as follows:(1)Designed a prosthetic hand structure for machine vision-assisted myoelectric prosthetic hand controlAccording to the different distribution of functions,the design scheme of distributed prosthetic hand was proposed,using K210&STM32 as the main and auxiliary dual microprocessors,and the two exchange signals through wireless communication.The structure of the prosthetic hand was made by 3D printing,and the various parts of the prosthetic hand were designed through modeling software,including the design of the palm,the design of the distribution point of the finger drive motor,the design of the fingers and knuckles,and the control of the prosthetic hand in this paper was determined.The "rope-tendon" model was used as the finger movement mode,and the pulley was used to conduct force transmission to the "rope-tendon",so that three knuckles could be driven by a metal rope.Using MATLAB on the PC to simulate the kinematics of the fingers,completed the selection of visual sensors,drive motors,and pressure sensors,and build a visual-myoelectric prosthetic hand experimental platform.(2)Real-time gesture recognition for EMG signalsIn order to make the designed prosthetic hand had preliminary intelligent functions,the idea of using myoelectric signals to control the prosthetic hand to realize gesture movements was designed.Firstly,the surface electromyography signals of stretching,clenching,"good" hand,"ok" hand and pinching five gestures on the specific muscles of normal adult arms were obtained through the electromyography acquisition device,and the corresponding data sets were produced,and principal component analysis was used,decision tree and support vector machine to classify and recognize the data,and select the optimal model for transplanting the microprocessor through the comprehensive recognition rate of the test set on the PC side.The experimental results showed that the comprehensive recognition rate performance of the support vector machine machine learning model on the PC-side test set was better,reaching 96%,and the recognition rate in the actual machine test was also above 85%.At the same time,the recognition time was monitored.,all within200 ms.It showed that the prosthetic hand designed in this paper could realize the basic requirement of intelligent prosthetic hand.(3)Machine vision and pressure sensors were used to control the grip force of myoelectric prostheticsAt present,the intelligent prosthetic hand with EMG signal as the control source could only meet the multi-gesture function or the single-gesture multi-level grip function,and the design of a single control source largely restricts the intelligentization of the prosthetic hand.Therefore,by increasing the collaborative control of data sources,the intelligence of the prosthetic hand could be promoted.In this paper,machine vision and pressure sensors were introduced,and commonly used items were used as experimental objects.The pressure thresholds corresponding to different items were designed and used to realize the grip control of the prosthetic hand..The results showed that the visual-myoelectric prosthetic hand designed in this paper could not only hold objects stably without destroying them,but also used different gestures when facing different objects,so that the prosthetic hand could achieve dual control of grip and posture,allowing the prosthetic smarter hands.