A Novel SEMG-Controlled Hand Function Exoskeleton Robot For Rehabilitation In Post-Stroke Individuals

Stroke is the first cause of death and disability in adults in China,and it leads to a high burden on families and society.Functional hand motor disability is one of the symptoms that is common and difficult to recover after stroke.Rehabilitation for post-stroke individuals are performed manually by physiotherapists.Although they can alleviate the symptoms of hemiplegia after stroke,the cumbersome process and high price limits the application of rehabilitation.The hand function rehabilitation robot is an effective means of rehabilitation for improving the rehabilitation effect and doing the repeated work instead of the rehabilitation therapist.However,most of the rehabilitation robots can't realize precise control of thumb and motivate the patient to actively participate and engage in rehabilitation tasks.In the sight of this,this paper aims to design a new generation of hand function rehabilitation robot that realizes fine control of thumb,and designs real-time recognition system of gesture and real-time control method of hand rehabilitation robot.The main work of the thesis includes:Firstly,the control system of hand function rehabilitation exoskeleton robot is designed,realizing the bilateral rehabilitation that the surface electromyography(sEMG)signals collected from the non-paretic side controls the movement of hand function rehabilitation exoskeleton robot wore upon the paretic side.The hardware equipment needed and overall line connection was analyzed.Moreover,the software for signal collection and motor control were programed in Labview.Secondly,a novel hand function rehabilitation exoskeleton robot was designed using Solidworks to realize five-finger independent movement,especially to realize thumb rotation.Simulation experiments and prototype test was carried out to evaluate the performance of the hand function rehabilitation exoskeleton robot and optimize its mechanical structure.Finally,two experiments were designed and conducted to collect sEMG signals from healthy subjects.The first experiment:The extracted 24 characteristic parameters were introduced,and differences between the left and right hand were statistically analyzed through root mean square(RMS).The co-contraction index(CI)valuesbetween the muscle pairs were calculated for each gesture to analyze muscle synergy.The statistical significance of most of the muscles between right hand and left hand was found in six hand motions,but the difference between the average accuracy of gesture classification using data of left hand and that of right hand(the difference is 1.1%)is small.Meanwhile,offline gesture classification research was carried out with sEMG signals of the right-hand data using machine learning algorithm,and support vector machine(SVM)showed a higher classification accuracy(0.907±0.004).The second experiment:SVM was applied to program the real-time gesture classification software.realizing the real-time classification of ten gestures,with an average accuracy of 0.6450±0.1534 for the final classification.