Muscle Force Estimation Study Fused Surface Electromyography And Ultrasound Radio Frequency Signals

Muscle strength refers to the contractile force produced by one or more muscles in the process of human movement.The estimation of muscle strength has always been an important subject in biomedical research and has important applications in clinical rehabilitation,prosthetic control,sports,military training,and other fields.For non-invasive muscle force estimation,surface electromyogram signal(sEMG)has been used for a long time,which can reflect the activity and functional state of the neuromuscular.However,it is difficult for EMG signals to carry the functional characteristics of deep muscle tissue,while the radio-frequency(RF)ultrasonic signals transmitted through muscle tissue carry better muscle motion characteristic information,which can reflect the morphological characteristics of muscle tissue from the other hand.In this paper,surface EMG and ultrasonic radio-frequency were fused to study muscle strength,considering not only the physiological indexes of muscle but also the morphology of muscle,and studying not only the electrical signals on the surface of muscle but also the ultrasonic radiofrequency signals transmitted through muscle tissue,so as to evaluate the contractile force of muscle more comprehensively and reliably.The purpose of this study is to improve the pattern recognition of muscle tissue force characteristics combined with EMG and ultrasound(such as the stable output of a specific force by the prosthetic control system)and to analyze the muscle force state of the subjects under different loads by using surface EMG and ultrasound techniques.The main research contents are as follows:The model is combined with a Convolutional Neural Network(CNN)and a Support Vector Machine(SVM)to estimate the muscle force.The CNN convolution and pooling operation are used to automatically extract the effective features of the signal CNNFeat.The extracted features are taken as the input of the SVM classifier,and the SVM algorithm is used for processing and classification to further improve the generalization ability of the model and the identification accuracy of muscle force.The complementarity of ultrasound and EMG fusion is verified in three situational modes(user-dependent,multi-user,and userindependent).The experiments show that the recognition rate of ultrasonic RF signal is higher than that of surface EMG in either scenario,and the recognition rate of the fusion of ultrasound and EMG is higher than that of single EMG or ultrasound.In addition,the network feature(CNNFeat)is compared with the traditional EMG and ultrasonic features.The experiment shows that the network feature(CNNFeat)could reflect the change process of signals to a certain extent,improve the recognition effect of the classifier,and has strong robustness.The combined network of CNN-SVM not only overcomes the defect of traditional classifier manual feature extraction but also improves the performance of classification.The experiment in this paper includes multi-step preprocessing,such as wavelet threshold denoising and data window processing,to extract the commonly used surface EMG and ultrasonic RF characteristics of subjects under the state of biceps brachii static contraction.EMG features include three time-domain features:Root Mean Square(RMS),wavelength Length(WL),and Zero Crossing(ZC).One frequency-domain feature:Discrete Wavelet Transform(DWT);And AR coefficients of order 6(AR6)of the autoregressive model.Ultrasonic features include linear fitting slope k and intercept b.