Feature Analysis Of SEMG Signal Based On Time-Frequency Compression Transform

The surface electromyography(sEMG)is an electrical signal generated during the physiological activity of the muscle,which carries the brain movement intention and muscle action information.The sEMG has the advantages of convenient monitoring,rich information,non-invasive,and harmless.Time-frequency analysis is an effective tool to deal with the complexity and uncertainty of sEMG signals.However,traditional time-frequency analysis methods are often limited by Heisenberg uncertainty principle,which may result in spectrum energy divergence problem.It is often difficult to provide high-resolution time-frequency analysis results,which leads to low accuracy and poor reliability of the analysis system.Therefore,the development of time-frequency analysis can greatly enrich the theory of sEMG signal feature extraction and promote the field of disease diagnosis.This paper mainly focuses on the problem of the traditional time-frequency analysis methods.Several novel time-frequency analysis methods are proposed,which include the high-order multisynchrosqueezing transform,second-order time-reassigned multisynchrosqueezing transform and weak component enhencement algorithms.In the chapter 2,a high-order multisynchrosqueezing transform is proposed to handle strong time-varying signals.First,the theoretical basis of the proposed method is established with the proof of convergence.Second,a high-order instantaneous frequency estimator is proposed for accerating the speed of the convergence.Finally,the performance of the HMSST is validated by the numerical signals.In the Chapter 3,a nonlinear synchrosqueezing transform which can deal with strong background noise is proposed to extract key time-varying features from low signal-to-noise ratio signals.Firstly,the theoretical basis of energy distribution of LFM signal processed by short-time Fourier transform is deduced;Then,based on the energy concentration index,the nonlinear frequency modulation operator is introduced to improve the time-frequency spectrum energy concentration,and the nonlinear time-frequency transformation method is proposed;At the same time,the instantaneous frequency estimation algorithm and iterative time-frequency transformation processing strategy are proposed to gradually improve the feature expression problem of low signal-to-noise ratio.Through the simulation and experimental signal processing,it is verified that the proposed algorithm can still effectively capture the nonlinear characteristics and key information of the signal under the background of strong noise.In the chapter 4,a second-order time-reassigned multisynchrosqueezing transform is proposed to handle the non-stationary signals with transient characteristics.First,the theoretical basis of the proposed method is established using frequency-varying signal model.Second,second-order group delay estimator is proposed for characterizing the transient features of the signal.Finally,the effectiveness of the proposed method is validated via several numerical examples.In the Chapter 5,the study foucses on the enhancement of the weak characteristics of the signal combined with the de-noising strategy.The weak characteristics of time-varying and frequency-varying signal models are simulatiously addressed using instantaneous frequency and group delay parameters.Several numerical exmples are employed to validate the effectiveness of the proposed methods.The proposed time-frequency analysis methods in the Chapter 6 are used in upper limb sEMG signal processing and motion pattern recognition applications.The time-varying and transient characteristics are analyzed using concentrated time-frequency spectrum,which can mining more dynamic information on the state of the motion unit and better understand generation mechanism and transmission law of sEMG signals.