Abstract:In order to capture the motion state of human lower limb joints under different motion modes in real time,a method of continuous motion prediction of lower limb joints using electromyography (EMG) was proposed.The EMG and motion data of human body in squat motion,knee flexion and extension motion and up and down stairs motion were collected and analyzed,and the simulation model of human skeleton and muscle was established by using musculoskeletal geometric modeling software Opensim,the inverse kinematics analysis was carried out to extract the movement curve of human lower limb joints.The mapping relation between the motion of human lower limbs in sagittal plane and EMG was established,and the sparrow search algorithm was used to optimize the Elman neural network to predict the continuous changes of ankle,knee and hip joint angles,the prediction results were compared with those of traditional back propagation neural network,support vector machine regression and Elman neural network.The results show that Elman neural network optimized by sparrow search algorithm has higher accuracy in predicting the change of lower limb joint angle;moreover,the predicted value and measured value of joint motion of the prediction model show certain correlation under different motion modes,and the correlation coefficient is greater than 0.89.It is proved that it is feasible to use EMG signal to predict the continuous motion of multiple joints of lower limbs.