Research On The Detection Method Of Muscle Fatigue Based On SEMG

With the continuous improvement of sensor technology.The biological signal collected by the sensor gradually enters the researcher's field of vision.Because the EMG signal is easy to observe and has a high real-time nature,it has attracted the attention of a large number of researchers.EMG signals are commonly used to monitor and evaluate the state of target muscle groups,of which muscle fatigue detection is widely used in the field of rehabilitation medicine.Its main purpose is to collect EMG signal data for analysis through sensors,which helps to elucidate the pathophysiological mechanism of muscle fatigue and tailor treatment methods for patients with different diseases.The surface electromyography signal(s EMG)is favored by the majority of researchers for its easy collection.However,the existing fatigue detection methods usually only collect a single active muscle signal for detection,and the detection accuracy and real-time performance are poor,and most studies use traditional feature extraction methods for fatigue detection,waveform processing and feature selection for detection The results are very influential.In response to this problem,this paper analyzes the surface EMG signal characteristics in detail,extracts the effective frequency band for fatigue detection through preprocessing,and performs fusion processing on the signals of different muscle groups,thereby improving the significance of muscle state changes and reducing the non-stationary and non-stationary The effect of linearity on the accuracy of fatigue detection.Based on the collected data sets,different neural network frameworks are used to use the neural network fatigue detection algorithm,and the neural network-based Teacher-Student network framework is used to reduce the interference between the differences between the subjects and the fatigue detection.The main contents of this article are as follows:(1)By analyzing the cooperative working principle of muscles,a surface EMG signal data set containing active muscles and cooperative muscles was collected.By comparing and analyzing the surface electromyography(SEMG)signal characteristics of male and female dominant biceps femoris and semitendinosus during fatigue,the similarities and differences between male and female,active muscle and synergistic muscle during exercise fatigue were explored.Sports training and therapy provide theoretical basis and evaluation methods.Since surface EMG signals are susceptible to interference from the environment,it is particularly important to filter the data set.In this paper,from the perspective of signal analysis,for the first time,an adjustable Q factor wavelet transform is used to denoise the surface EMG signal,and the frequency band in which the signal is concentrated is selected to reduce the impact of environmental noise on fatigue detection.The discrete wavelet and the adjustable Q-factor wavelet are used to filter the signal,and the median frequency(MF)and the average power frequency(MPF)are extracted as evaluation indicators.2.Since the surface EMG signal is susceptible to interference from the environment,it is particularly important to filter the data set.In this paper,from the perspective of signal analysis,for the first time,tunable Q-factor wavelet transform is used to denoise the surface EMG signal,and the frequency band of signal concentration is selected to reduce the impact of environmental noise on fatigue detection.The discrete wavelet transform and the adjustable Q-factor wavelet transform are used for comparative experiments,and the median frequency(MF)and the average power frequency(MPF)are extracted as indicators.Experimental results show that the noise reduction effect of adjustable Q-factor wavelet transform is far better than discrete wavelet transform.(2)This article is based on a variety of basic neural network frameworks.The experimental results show that the GRU neural network obtains the highest fatigue detection accuracy.Furthermore,this paper proposes a Multichannel Fusion Recurrent Attention Network(MFRANet).First,MFRANet enhances the local anti-interference ability of the signal by fusing multi-channel EMG signals,and reduces the impact of single channel signal noise on the overall detection performance.Secondly,MFRANet analyzes the signal from two dimensions,time domain and space domain.Gating mechanism is used to enhance the complex time correlation between channels.Attention mechanism is used to reconstruct the nonlinear relationship between channels,which improves the generalization.Experiments show that the signal fusion method of agonist and synergistic muscle proposed in this paper has significantly improved the accuracy of muscle fatigue detection.And the time has been significantly reduced compared to traditional machine learning methods.(3)In order to further weaken the impact of surface electromyography signal individual differences on fatigue detection,this paper uses the Teacher-Student network framework method based on MFRANet network to further improve the accuracy of fatigue detection by improving the neural network training mode.Experimental results show that the Teacher-Student network framework method based on MFRANet network proposed in this paper improves the accuracy of fatigue detection to 95.7%,and the detection time is significantly reduced compared with traditional machine learning methods.