Upper Limb Multi-Joint Angle Estimation Based on Multichannel sEMG Signals Using Elman Neural Network

Continuous motion angle estimation based on surface electromyography (sEMG) signals is a significant part of human active motion intention recognition, which plays an crucial effect in the aspect of natural human-robot interaction and rehabilitation therapy. In this paper, to predict the upper limb multi-joint angle based on multichannel sEMG signals, the Elman neural network model (ELNN) is applied and investigated to estimate upper limb multi-joint motion angle from multichannel sEMG signals under different motion modes of the upper limbs. Specifically, the sEMG signals of anterior deltoid (AD), posterior deltoid (PD), biceps brachii (BB), triceps brachii (TB), extensor carpi radialis (ECR) and flexor carpi radialis (FCR) will be collected and preprocessed, then, the ELNN model based on multichannel sEMG signals is employed to predict the multi-joint motion angles of the upper limbs including shoulder, elbow and wrist. Theoretical analysis, experimental results and root-mean-square error (RMSE) analysis indicate that the presented ELNN model has better prediction accuracy and dynamic characteristics than BP network in continuous estimation of upper limb multi-joint motion angle.