Prediction of Continuous Motion for Lower Limb Joints Based on SEMG Signal

in order to help the patients with lower extremity dyskinesia to carry out rehabilitation training task through the exoskeleton, a prediction model of lower extremity joint continuous motion based on the combination of GA-BP neural network and limited amplitude filtering is proposed in this paper. The relationship between surface electromyography and lower-limb joints' motion using genetic algorithm to optimize the parameters of BP neural network, is studied mainly. Based on optimizing the surface electromyography filtering method, the relationship between the surface electromyography (SEMG) and the knee joint angle mapping were constructed by GA-BP neural network. The model error can be reduced by the optimized filtering of the output results and the continuous prediction of knee joint changes can be achieved too. The root-mean-square error of the predicted knee joint angle before and after optimization is reduced by 24% which is proved by experiments. In addition, the effect of different muscles on joint angle and the robustness of the prediction model are analyzed by reducing the input value of surface electromyography.