Prediction of Human Dynamic Ankle Moment Based on Surface Electromyography Signals

Abstract

In order to achieve the dynamic ankle moment decoding from surface electromyography(sEMG),a prediction method based on Elman neural network is proposed for the output of ankle dynamic moment. By using smoothed filtering normalized sEMG signal, thigh angle and shank angle as the input of the network, and the ankle moment obtained by the Newton-Euler inverse dynamics method as the expected output of the network, a network prediction model based on Elman training algorithm is established. Based on the trained neural network model, the dynamic prediction experiment of ankle moment of lower extremity during normal walking is carried out. The results show that Elman neural network forecasting method using surface electromyography signal and thigh/shank angle as input variables is an effective way to realize accurate estimation of ankle moment during natural walking. The average correlation(CORR) between the predicted value and the expected value is 0.979,and the average normalized root mean square error (NRMSE) is 0.116. The correctness and validity of this method are verified and it can provide an interface for the control research of lower extremity exoskeleton robots.