Estimation of elbow motion intension under varing weight in lifting movement using an EMG-Angle neural network model

An EMG (electromyogram)-angle neural network was built to estimate elbow movement in a lifting task from sEMG signals in this study. The movement conditions vary with 6 different weights and no load. Subsequently the predicted angle could be utilized as the control signal of a 2DOF powered exoskeleton. Sixteen-channel EMG raw data sets were acquired from two commercial wearable MYO gesture armbands. The signals were then segmented, rectified and filtered with FIR filter to extract features of EMG signals. Afterward a three-layered BPNN (back propagation neural network) was trained to predict the elbow joint angle. Finally experiments were conducted to evaluate the feasibility of the developed architecture.