Research on prediction of human lower limb joint angle based on sEMG signal

The lower limb exoskeleton robot is a human-machine collaborative intelligence system that is human-centered. Accurate real-time recognition of human motion intention is a critical technology that must be overcome in developing lower-limb exoskeleton robots. Many studies analyse human motion intention through physiological electrical signals, such as human real-time joint angle, angular velocity, motion acceleration, etc. Surface electromyography (sEMG), as an easy-to-collect human physiological signal, has the advantage of giving full play to its time advancement and behavioral perception globality. In summary, this paper focuses on the mapping relationship between the sEMG signal and human lower limb joint angle to realize the efficient identification of human motion intention and improve the human-machine coordination of lower limb exoskeleton robot. Finally, the model constructed in this paper has the following three advantages. A data set with low redundancy is used as a training set and a test set. Compare the data feature extraction effects of different feature extraction methods for extracting surface electromyography signals, including mean absolute values (MAV), variance (VAR), empirical mode decomposition (EMD), wavelet packet decomposition (WPD). Based on the time series characteristics of surface EMG signals, a time domain neural network is selected to construct a prediction model.