Study On Prediction Of Elbow Joint Angle Based On Surface EMG Signals

As China's aging problem gets worse,strokes with high morbidity and disability can cause heavy emotional and economic burdens on individuals,families and society.Based on the principle of brain plasticity,if a stroke patient receives appropriate amount of appropriate rehabilitation training within three months after the onset of the disease,there is a high probability that they will return to their original physical function level.However,the traditional rehabilitation training was not satisfactory because of the small number of rehabilitation doctors,high labor intensity and excessive dependence on experience,and rehabilitation robots came into being.In order to increase the patient's awareness of subjective participation in the rehabilitation process,the use of surface EMG signals as control signals for rehabilitation robots has become a research hotspot.Surface EMG signals are electrical signals generated when muscles contract,and contain the intention of human movement,and can be used as quantitative control signals for rehabilitation robots.In order to meet the needs of upper limb function recovery after stroke,the movement of the elbow joint was selected as the research focus.Aiming at the shortcomings of the existing acquisition equipment,a synchronous acquisition system of surface electromyographic signals and elbow joint angle signals was established in this paper,and the signal acquisition experiments of 6 healthy subjects were completed using this system.For the collected original signal,by analyzing the frequency band and range of mixed noise,the Fourier transform and wavelet transform are used to denoise the EMG signal and angle signal,respectively.Using linear correlation coefficient and mean square error as the parameters of the evaluation model,compare the performance of 14 different eigenvalues,select the 5th-order AR model coefficients as the eigenvalues of the EMG signal,compare the performance of 5 regression algorithms,and select the BP neural network as the regression The algorithm finally established a quantitative relationship model between the surface EMG signal and the joint angle.The average error of the model's predicted angle was about 4 °,and the linear correlation coefficient between the predicted value and the ideal value was 94.4%.Surface EMG signals predict the angle of flexion and extension of the elbow joint.