| Stroke has become a very common and harmful disease with the increasing of aging population.Clinical trials have shown that timely and appropriate rehabilitation training can enhance patients’ motor function and promote their rehabilitation.Traditional rehabilitation training and evaluation are mainly assisted by professional doctors,which not only takes time,but also greatly affects the accuracy of evaluation results.As a new technology,the rehabilitation robot provides a new method for rehabilitation training and exercise evaluation of stroke patients.This study is aiming to develop a low-cost and simple-construction upper limb rehabilitation robot,and establish a quantitative evaluation model based on machine learning algorithm to automatically evaluate the upper limb motor function of stroke patients.The main work of this study is as follows:(1)A desktop upper limb rehabilitation robot with two degrees of freedom was designed,and the hardware control system for the robot was built.Combined with ergonomic knowledge,the dimensions of the rehabilitation robot were determined,and the 3D model of the rehabilitation robot was established by using SOLIDWORKS software.The components were selected to meet different functional requirements in the hardware control system,and the signal processing circuit was designed to facilitate data acquisition.(2)A motion control system equipped with brushless DC motor was designed for the desktop upper limb rehabilitation robot.PID controller,fuzzy PID controller and fuzzy controller were adopted to control the rotation angle,speed and thrust of the motor respectively.In addition,brushless DC motor motion control software and upper limb rehabilitation training and motion evaluation software were developed,and the function and application method of upper computer software were explained.(3)Through the desktop upper limb rehabilitation robot,the patients’ kinematics and dynamics data in the process of active training were collected.Through filtering,normalization and other pre-processing operations,the four evaluation indicators,the trajectory offset,average velocity,maximum instantaneous acceleration,and the maximum thrust of the upper limb were obtained.Finally,based on BPNN,KNN and SVR algorithm,quantitative evaluation models of upper limb motor function in stroke patients was established.(4)Through the step response experiment of the motion control system,the reliability of the control system was verified.Step response experiments show that the time from excitation to stabilization of the motor output thrust,rotation Angle and speed control system is 614ms,672ms and 325ms,respectively.There is no overshoot in the motor output thrust,rotation Angle and speed control system,and the steady-state errors are all zero.The designed motor motion control system has the characteristics of fast response speed,good accuracy and stability.(5)With the help of physicians,the clinical quantitative evaluation experiment of upper limb motor function in stroke patients was carried out to analyze the scoring performance of the three proposed quantitative evaluation models.It was found that the accuracy of BPNN model is 87.10%,which is higher than that of KNN model(83.87%)and that of SVR model(74.19%).The root mean square errors of BPNN model,KNN model and SVR model were 2.824,3.222 and 10.838,and the coefficients of determination were 0.973,0.966 and 0.612,respectively.The correlation coefficient between BPNN model’s scores and physicians’ scores was as high as 0.986,higher than that of KNN model(0.983)and that of SVR model(0.945).Compared with KNN model and SVR model,BPNN model has the best scoring performance. |
PDF Full Text Request