基于NARX和sEMG的肘关节连续运动预测

Continuous Kinematics Prediction of Elbow Joint Driven Based on NARX and sEMG

为了建立表面肌电信号(Surface Electromyography,sEMG)与人体肘关节连续运动量的精确预测模型,通过传感器记录肘关节屈伸角并采集与上肢运动相关联的肌肉表面肌电信号,经滤波处理后从中提取时域特征;在此基础上将非线性自回归(non-linear autoregressive,NARX)神经网络用于肘关节连续运动角度的预测,最终根据sEMG信号识别出的人体意图所对应的估计肘角。大量的实验结果验证了利用本文建立的模型可以精确估计人体肘关节连续运动角度,该模型可以有效用于人体假肢和辅助装置的控制,且本文方法的估计性能优于反向传播(back propagation,BP)神经网络。

In order to build an accurate prediction model of the surface electromyography(sEMG)and continuous movement variables of the upper limb elbow joint,the elbow joint flexion and extension angle is recorded by the sensor and the surface lectromyography of the muscle associated with the upper limb movement is collected and filtered.The time domain features are extracted from the processing.On this basis,a non-linear autoregressive(NARX)neural network is used to predict the angle of elbow joint continuous motion,and the estimated elbow angle corresponding to the human intention can be finally identified according to the sEMG signal.Extensive experiments are conducted to verify that continuous joint angles of upper limb motion can be accurately estimated by using the proposed model.The model can be effectively used for the control of human prostheses and auxiliary devices,Moreover,the proposed method is superior to BP neural network in estimation performance.