Research On Real-time Online Muscle Fatigue Estimation Method Based On SEMG Signal

Surface electromyography(sEMG)is one of the three major physiological electrical signals(EMG,EEG and ECG)of human body.It comes from the potential changes of central nervous system when it reflects external stimuli.Under the innervation of nerve,the membrane potential is changed by initiating polarization reaction on the surface of cell membrane,and then the potential is superimposed and transmitted on the muscle fiber,thus generating detectable physiological electrical signal on the surface of skin.As the external comprehensive representation of the potential activity of human muscle cells,it can reflect the current physiological and conscious state characteristics of human body,such as muscle fiber strength,muscle state,human psychological activity,brain movement intention,etc.In addition,the changes of the physical characteristics of sEMG are closely related to the biochemical and physiological changes of skeletal muscle during fatigue contraction.The most intuitive feeling of muscle fatigue is muscle weakness,which caused by long-term exertion or exercise.Physiologically speaking,muscle fatigue is the internal cause of the temporary decline of the ability to work or contract of the muscle movement system,which leads to the body's failure to maintain the expected intensity action.There are two main theories about the mechanism of muscle fatigue,central and peripheral.The central reason is that fatigue is the result of protective inhibition of cerebral cortex.In order to avoid body injury,active protection is taken.The peripheral reason is analyzed from the biochemical aspect.It is believed that the main reasons for the change of action conduction potential and the decrease of muscle fiber contraction force are the disturbance of calcium ion movement,the accumulation of phosphate and the decrease of ATP.Based on the relationship between sEMG and muscle fatigue,this paper focuses on the healthy human body and the detection of local muscle fatigue under static fatigue.The main research contents are as follows:(1)Exploratory classification of muscle fatigue level.In the study of local muscle fatigue in static environment,based on the experimental process of muscle isometric contraction to fatigue,the muscle state is divided into five categories: relaxation state,load state,fatigue transition state,deep fatigue state and fatigue state,and the muscle state is classified by supervised learning method.(2)To explore and optimize the characteristic indexes of surface electromyography during muscle fatigue.On the basis of the above research,through the research and analysis,the characteristic parameters which can be used to detect the muscle fatigue state in real time are found out.(3)Research on real-time online evaluation method and real-time online detection system of muscle fatigue.Through the research of fatigue grade and fatigue index,a real-time on-line analysis system is developed to detect the state of target muscle group.In this paper,a real-time on-line sEMG signal analysis system is developed,which can detect the current state of local muscle groups on-line.The experimental results show that under the off-line state,the above five muscle states are divided by ANN,KNN,LDA and SVM.Under the RMS and MNF features,the average recognition effect achieved by ANN and LDA is the best,over 95% and 90%.Under the RMS and Ar6 features,the average recognition accuracy of ANN,KNN and LDA classifier is higher.Considering that ANN has more time cost,the recognition rate of KNN is slightly worse than LDA,and the time-domain signal features have faster calculation speed,so in the real-time online muscle fatigue estimation system,LDA classification discriminator based on RMS and Ar6 features is adopted.The system can achieve a recognition effect of 92% at most when recognizing whether the muscle state reaches the deep fatigue state,but the recognition results of three states,relaxation state,load state and fatigue transition state,need to be improved.By considering the above three states as one state,the recognition accuracy of muscle state before the deep fatigue state and fatigue state can be significantly improved.In addition,in the real-time online recognition analysis,it is found that the combination of autoregressive coefficient and root mean square value can achieve better recognition effect,while the better differentiated average frequency will have a lower recognition accuracy in muscle state recognition,considering that RMS and MNF features have certain feature index information redundancy.The implementation of this study has important practical significance and application value in rehabilitation medicine,sports kinematics,human-computer interaction,sports injury,auxiliary training and other fields.