Research On Elimination Control Of Motion Compensatory After Stroke Based On Bioinformatics Feedback

With the increasing number of stroke patients,the problem of insufficient rehabilitation support caused by the shortage of rehabilitation medical resources is becoming more and more serious.Traditional rehabilitation,which relies on the experience of health care providers and complex scales,is a critical test for the scarce medical resources in developing countries.Therefore,it is very necessary to study and apply sensor-based biological information feedback.Currently,a variety of sensing devices,including infrared cameras,Surface Electromyography(s EMG),inertial sensors,have been widely used in the field of rehabilitation.However,merely collecting macroscopic motion information will ignore the root cause of the abnormal motion state.At the same time,one-to-one guidance based on medical staff is also very inefficient for abnormal rehabilitation conditions such as compensation.Aiming at the above problems,this paper proposed to use the pressure distribution cushion and s EMG signal to ensure that the compensation of the trunk and upper limb could be detected in real time,and to improve the efficiency and quality of rehabilitation through the compensatory elimination control loop assisted by VR and the robotic arm and passive combination.The main work of this paper is as follows:Firstly,by collecting the pressure distribution data of patients in the sitting state,a realtime compensation category detection model of trunk based on Support Vector Machine(SVM)was established,and high recognition rate was achieved in offline and online states,with an average detection accuracy of 98.5%.Then,an active and passive compensatory intervention method using virtual reality environment and robotic arm is proposed.Then,aiming at the s EMG which can reflect the motion intention,this paper proposes a real-time detection method of upper limb compensation.A method for predicting the peak of maximum voluntary contraction based on body weight information.In this method,a variety of s EMG indexes correlated with muscle strength were selected as eigenvalues,and then a series of "Force-s EMG" sample points were fitted bySupport Vector Regression(SVR).Finally,K-means clustering algorithm was introduced to dynamically filter the real-time data.The experimental results show that the Mean Absolute Value(MAV)is the SEMG characteristic Value with the highest prediction accuracy,which can maintain a high prediction accuracy(about 92%)in the case of a small number of sample points in the calibration link,and is an ideal SEMG feature for visualization of muscle activation degree.Finally,this paper combined the above two subsystems through TCP communication and MySQL database,and built a compensatory elimination control system based on UR5 manipulator.The system can access the patient’s medical record information,select the rehabilitation mode,detect and intervene the abnormal behavior and other functions.Based on this system,we conducted experiments on the elimination of stroke compensation for patients.In the three kinds of forward leaning,rotation and shoulder lifting compensation,the average compensation Angle under man-machine cooperative control was reduced by 47,32 and 22%,respectively,which has a very high clinical value.