Force Estimation Based on sEMG using Wavelet Analysis and Neural Network

for further improving the action of sEMG in human-robot interaction, wavelet and neural network is utilized in prediction of grip force. Firstly, it is described based on the introduction platform how to gain sEMG as well as its traditional features. Then, the wavelet decomposition and reconstruction algorithm is used to analyze the sEMG signals and extract the corresponding energy characteristics. Different grasp force of and sEMG signals are collected simultaneously whose feature matrix is used in training model. Those are evaluated by root mean square error, whose results show that RMSE = 1.0 +/- 0.4 of BP network and RMSE = 1.8 +/- 0.5 of LSTM model.