A Novel Method for Predicting Action Switching in Continuous Motion based on sEMG Signals

摘要

As one of the main signals of human movement intention recognition, surface electromyography (sEMG) has received more and more attention. However, due to the non-stationary, non-periodic and chaotic characteristics of the sEMG signal itself, it is difficult to recognize intentions in continuous actions. In practical applications, if the state switching position in continuous motion can be accurately captured, the accuracy of motion intention recognition will be greatly improved. Therefore, this paper proposes a predictive classification method based on multi-channel sEMG signals, and conducts preliminary exploratory experiments, aiming to provide a new idea for the study of action switching in continuous motion. Combining the predictive model with the diagnostic model, and testing the data of 8 objects through online simulation, the average switching delay is 145.5ms. The experiment confirmed the feasibility of the research idea and laid the foundation for the future research of continuous multi-action switching.