Muscle fatigue quantification method based on surface electromyogram signals and application thereof

Abstract

The invention provides a muscle fatigue quantification method based on surface electromyogram signals and application thereof, and particularly relates to the technical field of signal processing, and the method comprises the steps: collecting and preprocessing sEMG signals, extracting signal modal components with a slowly-varying bandwidth characteristic through a TVBCMD algorithm, and recognizing dynamic boundary parameters through an FBD-FST method so as to determine parameters of the TVBCMD algorithm. According to the method, a recursive framework of an ABCMD algorithm is constructed, signal modal components are extracted one by one according to spectrum trend energy, an average power frequency MPF and a median frequency MF are extracted from the extracted components, quantitative analysis is carried out, the muscle fatigue degree is accurately evaluated, and the method has important kinematics and medical significance for preventing human body muscle diseases and sport injuries.

Description

Technical Field

[0001] The present invention relates to the technical field of signal processing, and in particular to a muscle fatigue quantification method based on surface electromyography signals and an application thereof. Background Art

[0002] With the deepening of modern sports science and medical research, the accurate quantitative analysis of muscle fatigue status has become an important research field for preventing sports injuries and muscle diseases. In recent years, surface electromyography (sEMG) has received widespread attention and application as a non-invasive, convenient and efficient means of monitoring muscle activity. With the advancement of technology, sEMG signal analysis technology has continued to develop, providing the possibility for real-time monitoring of muscle fatigue status. However, in practical applications, the industry development faces many challenges. On the one hand, the complexity of the human muscle fatigue state makes the sEMG signal highly nonlinear and time-varying, which brings difficulties to the accurate extraction and analysis of signal features. On the other hand, the existing muscle fatigue detection methods are often difficult to accurately extract the real modal components inside the sEMG signal under the condition of strong noise background, resulting in inaccurate quantitative analysis of muscle fatigue status. In addition, most of the traditional instantaneous frequency (IF) initialization methods select the highest energy value as IF based on the overall time-frequency distribution (TFD), which cannot fully reflect the change trend of the modal components of the sEMG signal in different frequency bands, further limiting the accuracy of muscle fatigue status assessment. Therefore, how to accurately extract the main modal components of sEMG signals in a complex noise background and achieve precise quantitative analysis of muscle fatigue status is a key issue that needs to be urgently solved in the current industry.

[0003] In response to the above challenges, traditional existing technologies mainly try to solve them through a series of signal processing technologies. On the one hand, in order to extract modal components from sEMG signals, researchers have developed a variety of methods, such as matching pursuit method and its variants (such as adaptive chirp mode decomposition ACMD). These methods convert broadband signals into narrowband signals through modulation and demodulation technology, and recursively extract multi-component AM-FM signals using variational optimization models. The ACMD method is particularly suitable for analyzing broadband frequency modulation signals and has been widely used in the field of physiological signal monitoring. However, the traditional ACMD method requires the initial instantaneous frequency (IF) and bandwidth penalty parameters to be set in advance, which to a certain extent limits its adaptability and accuracy. In order to overcome this problem, researchers proposed bandwidth-aware adaptive chirp mode decomposition (BA-ACMD), and through multiple experimental analyses, the relationship between the bandwidth and of the extracted components was established, and a mathematical expression was established. The weighted spectral trend was used to determine the boundaries of different modal components, thereby adaptively determining the parameters required for ACMD. On the other hand, in order to improve the IF initialization method, traditional technologies also try to use the time-frequency ridge extraction method and the instantaneous frequency initialization method based on Hilbert transform to obtain the IF. These methods can achieve relatively ideal results when the energy extracted from TFD has an obvious change trend and the noise interference is small. However, for signals with high energy concentration such as sEMG signals, these methods still find it difficult to fully reflect the change trend of the modal components of the signal within different frequency bands. Therefore, although the traditional existing technologies have solved some problems in the assessment of muscle fatigue status to a certain extent, there are still many limitations and challenges, which require further innovation and improvement.

[0004] In summary, how to accurately quantify and analyze the degree of muscle fatigue, so as to accurately determine the degree of muscle fatigue, is a difficult problem that needs to be solved urgently in this technical field. Summary of the invention

[0005] The main purpose of the present invention is to provide a muscle fatigue quantification method based on surface electromyography signals and its application, so as to solve the technical problem that the prior art cannot accurately quantify and analyze the degree of muscle fatigue, and thus cannot accurately judge the degree of muscle fatigue. It has important kinematic and medical significance for preventing human muscle diseases and sports injuries.

[0006] In order to achieve the above-mentioned purpose, the present invention provides a muscle fatigue quantification method based on surface electromyography signal and its application, which has important kinematic and medical significance for preventing human muscle diseases and sports injuries.

[0007] In a first aspect, the present invention provides a method for quantifying muscle fatigue based on surface electromyographic signals, the method comprising:

[0008] Collecting sEMG signals and preprocessing the sEMG signals;

[0009] The preprocessed sEMG signal is processed using the TVBCMD algorithm to extract multiple signal modal components with slowly varying bandwidth characteristics;

[0010] Using the FBD-FST method to identify the dynamic boundary parameters of each of the signal modal components to determine the required parameters of the TVBCMD algorithm;

[0011] According to the required parameters of the TVBCMD algorithm and the TVBCMD algorithm, a recursive extraction framework of the ABCMD algorithm is constructed, and the signal modal components are extracted one by one according to the spectral trend energy corresponding to each of the signal modal components until a preset stop condition is met;

[0012] A muscle fatigue feature is extracted from the extracted signal modal component, wherein the muscle fatigue feature includes but is not limited to the mean power frequency MPF and the median frequency MF, and the muscle fatigue degree is quantitatively analyzed based on the extracted muscle fatigue feature.

[0013] Optionally, the TVBCMD algorithm dynamically extracts a plurality of the signal modal components based on the ACMD technology by introducing a time-varying bandwidth penalty parameter.

[0014] Optionally, the using of the FBD-FST method to identify the dynamic boundary parameters of each of the signal modal components to determine the required parameters of the TVBCMD algorithm includes:

[0015] Determine the dynamic boundaries of each of the signal modal components by detecting the inflection points of the frequency spectrum trend of the sEMG signal using the FBD-FST method;

[0016] Determining the dynamic boundary parameter corresponding to the dynamic boundary according to the dynamic boundary of the signal modal component;

[0017] According to the dynamic boundary parameters, the required parameters of the TVBCMD algorithm are determined, wherein the required parameters are used to indicate the time-varying bandwidth penalty parameter and the initial instantaneous frequency.

[0018] Optionally, the determining the dynamic boundary of each of the signal modal components by detecting the inflection point of the frequency spectrum trend of the sEMG signal by the FBD-FST method; and determining the dynamic boundary parameter corresponding to the dynamic boundary according to the dynamic boundary of the signal modal component comprises:

[0019] Performing Fourier transform on the sEMG signal to obtain a frequency spectrum of the sEMG signal;

[0020] Extracting a spectrum trend of the spectrum, identifying an inflection point of the spectrum trend, and using the inflection point as a boundary point of different modal components of the signal;

[0021] The variation curve of the inflection point in the entire time period of collecting the sEMG signal is obtained by a sliding window technique, and the variation curve is used as a dynamic boundary parameter for dividing different modal components of the signal.

[0022] Optionally, the recursive extraction framework of the ABCMD algorithm is constructed according to the required parameters of the TVBCMD algorithm and the TVBCMD algorithm, and the signal modal components are extracted one by one according to the spectral trend energy corresponding to each of the signal modal components until a preset stop condition is met, including:

[0023] Step 1: extracting the signal modal component with the strongest energy according to the required parameters of the TVBCMD algorithm and the TVBCMD algorithm;

[0024] Step 2: subtracting the extracted signal modal component from the sEMG signal to obtain a residual signal;

[0025] Step 3: Repeat steps 1 and 2 until the energy of the remaining signal is lower than a preset threshold.

[0026] Optionally, extracting muscle fatigue features from the extracted signal modal components, the muscle fatigue features including but not limited to mean power frequency MPF and median frequency MF, and quantitatively analyzing the degree of muscle fatigue according to the extracted features, including:

[0027] Perform statistical analysis on the extracted signal modal components and calculate the mean power frequency MPF and median frequency MF;

[0028] The muscle fatigue degree is quantitatively evaluated based on the calculated average power frequency MPF and median frequency MF, combined with the known relationship between the muscle fatigue degree and the muscle fatigue characteristics.

[0029] In a second aspect, the present invention provides an application of the muscle fatigue quantification method based on surface electromyography signals as described in the first aspect in preventing human muscle diseases and sports injuries.

[0030] The present application provides a muscle fatigue quantification method based on surface electromyographic signals and its application. The method collects sEMG signals and performs preprocessing to improve signal quality, uses TVBCMD algorithm to process preprocessed sEMG signals, and effectively extracts multiple signal modal components with slowly varying bandwidth characteristics. In order to optimize the parameters of the TVBCMD algorithm, the present invention introduces the FBD-FST method to accurately identify the dynamic boundary parameters of each signal modal component. On this basis, a recursive extraction framework of the ABCMD algorithm is constructed, and the signal modal components are extracted one by one according to the spectral trend energy size of the signal modal component until the preset stop condition is met. This process ensures the comprehensive and accurate extraction of the signal modal components. Finally, from the extracted signal modal components, the present invention further extracts key muscle fatigue features such as mean power frequency MPF and median frequency MF, and quantitatively analyzes the degree of muscle fatigue based on this, providing a scientific basis for muscle fatigue monitoring. The method solves the technical problem that the prior art cannot accurately quantify and analyze the degree of muscle fatigue, thereby accurately judging the degree of muscle fatigue, and has important kinematic and medical significance for preventing human muscle diseases and sports injuries. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The drawings constituting a part of the present application are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0032] Figure 1 A flowchart of a method for quantifying muscle fatigue based on surface electromyography signals provided in this application.

[0033] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0034] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings in this application. Obviously, the described embodiments are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0035] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in sequences other than those illustrated or described herein.

[0036] In the present invention, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in this application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a specific way.

[0037] The present application provides a muscle fatigue quantification method based on surface electromyography signals and its application. The method collects and preprocesses sEMG signals, extracts signal modal components using the TVBCMD algorithm, and determines the algorithm parameters using the FBD-FST method. An ABCMD recursive framework is constructed to extract components one by one according to the spectral trend energy. Finally, muscle fatigue features such as MF and MPF are extracted from the components to achieve quantitative analysis of muscle fatigue, providing a new means for muscle health monitoring.

[0038] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0039] Figure 1 The present invention provides a flow chart of a method for quantifying muscle fatigue based on surface electromyography signals. The method aims to quantify muscle fatigue based on surface electromyography signals, such as Figure 1 As shown in FIG. 1 , a muscle fatigue quantification method based on surface electromyography signals is provided in this embodiment, and the method includes:

[0040] S101: collecting sEMG signals and preprocessing the sEMG signals.

[0041] The collecting of sEMG signals and preprocessing of the sEMG signals may include the following steps:

[0042] Step 1: Prepare the acquisition equipment:

[0043] Use a sEMG acquisition device containing multiple electrodes, such as the MyoWare sensor or similar commercially available devices.

[0044] The electrode array should be placed over the target muscle group, such as the biceps in the arm or the thigh muscles in the leg.

[0045] Use a conductive gel or patch to ensure good contact between the electrodes and the skin.

[0046] Step 2: Signal acquisition:

[0047] Connect the sEMG acquisition device to a data acquisition system, such as an electromyograph or a dedicated data acquisition card.

[0048] Set the sampling frequency between 1000Hz and 5000Hz to capture enough signal details.

[0049] Make sure the subject is in a relaxed state and then perform specific muscle activities (such as flexion or extension) to record sEMG signals.

[0050] Step 3: Filtering:

[0051] The acquired sEMG signals were filtered using a bandpass filter to remove low-frequency noise (such as baseline drift) and high-frequency noise (such as electromagnetic interference).

[0052] The frequency range of the filter was set to 10 Hz to 500 Hz, which is the main frequency component range of the sEMG signal.

[0053] Step 4: Denoising:

[0054] Apply algorithms such as Independent Component Analysis (ICA) or Principal Component Analysis (PCA) to separate the noise component from the signal and remove it.

[0055] Adaptive filters are used to adjust the filter parameters according to the dynamic characteristics of the signal to further reduce noise.

[0056] Step 5: Standardization:

[0057] The denoised sEMG signals were normalized to eliminate the differences in signals between different subjects or between the same subject at different times.

[0058] Normalization methods include converting the signal to a standardized signal with a mean of 0 and a standard deviation of 1, or normalizing the signal based on its peak value or root mean square value.

[0059] Step 6: Signal verification:

[0060] The preprocessing process is verified using known standard signals or simulated signals to ensure the accuracy and reliability of the preprocessing method.

[0061] Compare the signal quality before and after preprocessing to evaluate the effectiveness of filtering, denoising, and normalization steps.

[0062] It can be understood that through the above steps, the quality of sEMG signals has been significantly improved, and noise and interference have been effectively removed. The pre-processed signals are clearer and more stable, providing a reliable basis for subsequent signal analysis and application.

[0063] S102: Processing the preprocessed sEMG signal using the TVBCMD algorithm to extract multiple signal modal components with slowly varying bandwidth characteristics.

[0064] Among them, TVBCMD algorithm refers to time varying bandwidth chirp mode decomposition algorithm (time varying bandwidth chirp mode decomposition, TVBCMD), and this application uses TVBCMD algorithm as the abbreviation of time varying bandwidth chirp mode decomposition algorithm. Based on the ACMD algorithm, the present invention proposes a time varying bandwidth chirp mode decomposition algorithm (time varying bandwidth chirp mode decomposition, TVBCMD) to extract the internal sub-modal components of slowly varying bandwidth frequency modulated signals. ACMD algorithm refers to Adaptive Chirp Mode Decomposition (ACMD for short), which is an advanced signal processing technology. It has a wide range of applications in signal processing, fault diagnosis, blind source separation and other fields.

[0065] Specifically, the TVBCMD algorithm dynamically extracts multiple signal modal components based on the ACMD technology by introducing a time-varying bandwidth penalty parameter. The specific principle of the TVBCMD algorithm is:

[0066] The original ACMD method is extended to process arbitrary time-varying bandwidth FM signals. First, the generalized FM signal model with M signal modal components can be expressed as:

[0067]

[0068] Where: x(t) is the signal to be decomposed, x m (t) The mth signal to be decomposed, A m >0 and f m >0 represents the instantaneous amplitude and instantaneous frequency of the modal component, A m (t) is the amplitude variation function, is a frequency variation function, θ m is the initial phase. So we only need to take A m (t) and f m (s) is enough, A m (t) is the amplitude function that changes with time t, f m(s) is the instantaneous frequency of the original signal component at time point s, is a frequency function that changes with time t. It can be seen that the IF (English abbreviation of initial instantaneous frequency) of FM (FM is the English abbreviation of amplitude modulation and frequency modulation signal, and x(t) in the above formula is used to indicate FM) has time-varying characteristics and occupies a wider frequency band in the frequency domain. From the idea of ​​generalized Fourier transform, it can be known that frequency demodulation can convert broadband modal signals into narrowband forms. The demodulation result of the above formula is:

[0069]

[0070] in:

[0071]

[0072] Where: u m (t) and v m (t) are the real and imaginary parts of the corresponding demodulated signal, is the frequency value of the frequency function of the mth demodulation operator at time point s, when When , the demodulated signal is transferred to the baseband, where the modal component has the narrowest bandwidth.

[0073] TVBCMD estimates the optimal signal component by minimizing the energy of the remaining signal. Specifically, for the mth signal whose bandwidth varies with time, it can be estimated by the following variational model:

[0074]

[0075] Where: ||·|| 2 Represents L 2 norm, and The energy of the second-order derivative of the demodulated signal is used to measure the smoothness of the signal, thereby evaluating the signal bandwidth. Here we introduce a time-varying bandwidth penalty parameter α(t) to control the filtering bandwidth of the signal component at different times. The smaller α(t) is, the smoother the output signal mode is, that is, the narrower the filtering bandwidth at a certain moment. The TVBCMD algorithm is essentially a TF bandpass filter, in which the TF bandpass filter refers to a filter that combines time-frequency (TF) analysis technology to allow signals within a specific frequency range to pass, while suppressing signals below and above the predetermined frequency range.

[0076] Assume that the signal sampling time is t = t 0 ,…,t N-1 , substituting equation 2 into equation 5, the matrix form of the objective function is:

[0077]

[0078] Where: x = [x(t 0 )…x(t N-1 )] T ,y m =[(u m ) T (v m ) T ] T ,u m =[u m (t 0 )…u m (t N-1 )] T , v m =[v m (t 0 )…v m (t N-1 )] T ;

[0079] α m =[β m β m ] T #(7)

[0080] in:

[0081] β m =[β(t 0 )…β(t N-1 )] T #(8)

[0082] Where β(t) is the bandwidth penalty parameter at a certain moment.

[0083] K m =[A m B m ]#(9)

[0084] in:

[0085] A m =diag[cos(θ m (t 0 ))…cos(θ m (t N-1 ))]#(10)

[0086] B m =diag[sin(θ m (t 0 ))…sin(θ m (t N-1 ))]#(11)

[0087] in Ω is a second-order difference matrix, and 0 represents the zero matrix.

[0088] In order to minimize the objective function of equation 5, the ADMM alternating multiplier method is used to update the demodulated signal and instantaneous frequency. First, the demodulated signal vector x is updated by setting the objective function gradient to zero. m , that is,

[0089] get:

[0090]

[0091] Where n represents the number of iterations, γ m =diag[α m ]. Directly use the gradient to be zero to update the IF f m It is more difficult. From equations 3 and 4, we can know that the instantaneous frequency of the demodulated signal is It can be used as a frequency increment to update the demodulated frequency. First, use inverse tangent demodulation to extract the frequency increment. The expression is as follows:

[0092]

[0093] Where: (·)' represents the first-order derivative. Generally speaking, the instantaneous frequency of the signal should be a smooth, slowly varying function, so the frequency increment should satisfy the following relationship:

[0094]

[0095] Here Ω is a second-order difference matrix, The frequency increment is calculated by equation 13, is the expected frequency increment, μ>0 is the penalty parameter, and solving the above equation yields the final frequency increment:

[0096]

[0097] Where e is the identity matrix. Then the final IF vector can be updated as:

[0098]

[0099] Then, the analysis matrix is ​​updated according to equations 7, 8 and 9 The final target signal component can be updated according to the following formula:

[0100]

[0101] Repeat the above steps until the convergence condition is reached, that is, the signal components extracted from two adjacent iterations are not much different. At this time, the residual component is obtained by subtracting the extracted signal component from the original signal Then to rm (t) Perform the above steps to recursively extract the signal modal components until the last extracted signal component reaches the stopping threshold δ and then stops decomposition. δ is generally set to 1e-9 to 1e-7.

[0102] It can be deduced from the above formula that the matching IF and α(t) need to be specified in advance. In the original ACMD algorithm, α is a fixed value. As the number of iterations increases, α gradually decreases, and the output signal mode becomes smoother. The mathematical expression between the signal frequency bandwidth and α is given in BA-ACMD:

[0103]

[0104] in is the bandwidth of the signal component, A c =0.5 means the amplitude of the normalized bandwidth is 0.5.

[0105] At this time, if the frequency bandwidth of a certain modal component at a certain moment is obtained in advance, α(t) at this time can be calculated by this expression. Algorithm 1 gives the pseudo code of TVBCMD, as shown in Table 1 below. In addition, the IF of the input ACMD is commonly used to extract the instantaneous frequency ridge on the TFD, but this method can only extract the local extreme point of energy in the full frequency band as the IF of the main mode, and the amount of calculation is large. For the sEMG signal under muscle fatigue, its TFD is complex and the main energy is concentrated within 20 to 300 Hz. It is difficult to directly observe the obvious trend of modal component changes. The signal components extracted by this method are very prone to frequency aliasing. Therefore, in order to solve the above problems, the next step will propose the FBD-FST method to determine the bandwidth of different modal components and the corresponding IF.

[0106] Table 1

[0107]

[0108]

[0109] S103: Using the FBD-FST method to identify the dynamic boundary parameters of each of the signal modal components to determine the required parameters of the TVBCMD algorithm.

[0110] Among them, the FBD-FST method refers to the frequency boundary determination method-Fourier spectrum trend (FBD-FST), which is used to identify the instantaneous bandwidth parameters of different modal components.

[0111] Furthermore, the use of the FBD-FST method to identify the dynamic boundary parameters of each of the signal modal components to determine the required parameters of the TVBCMD algorithm includes:

[0112] The FBD-FST method is used to detect the inflection point of the spectrum trend of the sEMG signal to determine the dynamic boundary of each of the signal modal components; according to the dynamic boundary of the signal modal component, the dynamic boundary parameter corresponding to the dynamic boundary is determined; specifically, it includes: performing Fourier transform on the sEMG signal to obtain the spectrum of the sEMG signal; extracting the spectrum trend of the spectrum, and identifying the inflection point of the spectrum trend, and using the inflection point as the boundary point of different signal modal components; obtaining the change curve of the inflection point in the entire time period of collecting the sEMG signal through the sliding window technology, and using the change curve as the dynamic boundary parameter for dividing different signal modal components.

[0113] According to the dynamic boundary parameters, the required parameters of the TVBCMD algorithm are determined, wherein the required parameters are used to indicate the time-varying bandwidth penalty parameter and the initial instantaneous frequency.

[0114] When implementing it, you can refer to the following principles:

[0115] It can be seen from the above step S102 that when the frequency bandwidth of the signal mode is a priori knowledge, we can obtain α(t) according to formula 18. During the human muscle fatigue process, the spectrum of the sEMG signal will shift to the left. Relatively speaking, the amplitude of the high-frequency component decreases, the amplitude of the low-frequency component increases, and the bandwidth of its spectrum slowly decreases. Inspired by this, this step proposes the FBD-FST method, which divides the spectrum of the sEMG signal into multiple frequency bands and extracts the dynamic boundaries of these frequency bands that change over time. This is used as the bandwidth parameter for extracting the main modal components in different frequency bands, and then determining IF and α(t).

[0116] The spectrum trend can be used to detect the change of certain frequencies in the signal over time, as well as the change law of the frequency components. The signal modal components in this step are the performance of the signal in different frequency bands, and are the various components after the signal is decomposed. In signal processing, the original signal can be decomposed into multiple modal components through specific decomposition methods (such as empirical mode decomposition EMD, variational mode decomposition VMD, etc.), each modal component corresponds to a specific frequency band, and the existence of the signal modal components makes the signal present different frequency band characteristics in the frequency domain. By decomposing the signal, the modal components in different frequency bands can be obtained, and these modal components each occupy a certain bandwidth. This step is based on the FBD-FST method to determine the boundaries of different frequency bands in the sEMG signal.

[0117] First, the spectrum can be divided into a spectrum trend and a high-frequency component. The trend can be understood as a linear combination of many Fourier basis functions. Therefore, the spectrum trend can be modeled as:

[0118] s N×1 =K N×(2M+1) c (2M+1)×1 +h N×1 #(19)

[0119] Where s = [S (f 0 ),…,S(f N-1 )] T represents discrete Fourier spectrum, N is the number of spectrum samples, K is the Fourier basis matrix, c is the coefficient vector, M is the order of the Fourier basis, and h is the high-frequency component.

[0120] The coefficient c can be solved using the Tikhonov regularization method:

[0121]

[0122] Where λ is the regularization parameter, usually set to 0.01, I is the identity matrix, (·) -1 is the inverse of the matrix. Using the coefficient c estimated by the above formula, the spectrum trend can be obtained as:

[0123]

[0124] In order to avoid the occurrence of more unnecessary local minima in the spectrum trend caused by a larger M, M should satisfy 3N×10 -3 ≤M≤3N×10 -2 .

[0125] The traditional method uses the distance between two consecutive local minima of the spectrum trend as the frequency band boundary to obtain stable boundary parameters. However, the purpose of this step is to extract the dynamic frequency band boundary of the sEMG signal, and then determine the bandwidth change curve of the signal components in different frequency bands over time. Therefore, a reasonable rule is needed to divide the frequency band boundary. The position of the inflection point can reflect the change in the concavity of the curve. Inspired by this, this step proposes to use the inflection point of the spectrum trend as the boundary point of different frequency bands. For the relatively slowly changing time-varying bandwidth frequency modulation signal, the inflection point of its spectrum trend curve also changes with time. It is only necessary to determine the inflection point position of the spectrum trend at the current moment, and then use the sliding window technology to obtain the inflection point change curve of the entire time period. The curve is used as the dynamic boundary parameter for dividing different frequency bands, so as to more effectively capture the changes in the bandwidth of the main modal components in different frequency bands.

[0126] The spectrum trend of a single window signal obtained by Fourier model fitting is s t , the inflection point of the spectrum trend can be obtained by the following formula:

[0127]

[0128] where f d >0 is the defined boundary point. It should be noted that due to the recursive extraction nature of the TVBCMD algorithm, this method only needs to determine the left and right boundaries of a frequency band each time, and determine the dynamic boundaries of different frequency bands in order from high to low energy. Its essence is to divide the frequency bands of the sEMG signal in the TF domain, so as to obtain the dynamic boundary information of different frequency band components.

[0129] S104: constructing a recursive extraction framework of the ABCMD algorithm according to the required parameters of the TVBCMD algorithm and the TVBCMD algorithm, extracting the signal modal components one by one according to the spectral trend energy corresponding to each of the signal modal components until a preset stop condition is met.

[0130] Among them, the ABCMD algorithm refers to the adaptive bandwidth chirp mode decomposition algorithm (adaptive bandwidth chirp mode decomposition, ABCMD), referred to as the ABCMD algorithm in this application. The ABCMD algorithm obtains the recursive extraction framework of the ABCMD algorithm based on the FBD-FST method and the TVBCMD algorithm, first extracts the modal component with the strongest energy, and then subtracts the mode from the original signal.

[0131] Further, the recursive extraction framework of the ABCMD algorithm is constructed according to the required parameters of the TVBCMD algorithm and the TVBCMD algorithm, and the signal modal components are extracted one by one according to the spectral trend energy corresponding to each of the signal modal components until a preset stop condition is met, including:

[0132] Step 1: extracting the signal modal component with the strongest energy according to the required parameters of the TVBCMD algorithm and the TVBCMD algorithm;

[0133] Step 2: subtracting the extracted signal modal component from the sEMG signal to obtain a residual signal;

[0134] Step 3: Repeat steps 1 and 2 until the energy of the remaining signal is lower than a preset threshold.

[0135] When implementing it, you can refer to the following principles:

[0136] As described in the above steps, the FBD-FST method can give the dynamic boundaries f of different frequency bands. d, and then the IF and α(t) of the TVBCMD algorithm can be automatically determined, so the algorithm is named ABCMD algorithm. The specific process of the algorithm is introduced below.

[0137] First, the window size is set to 2000 signal points and the step size is 50. The spectrum trend of each window signal is calculated respectively, and the inflection point curves of these spectrum trends are extracted as the boundaries of the frequency band.

[0138] The detected frequency band bandwidth can be expressed as:

[0139]

[0140] Among them, I m , m=1,…,M is the bandwidth variation curve of the mth frequency band, and Represent the left and right boundaries of the mth frequency band respectively. It should be noted that this paper assumes that the frequency band change of sEMG signal should be slow and smooth, so the obtained data is expanded to the same length as the original signal by interpolation method and mean filtering is performed to obtain the bandwidth BW of the corresponding frequency band. m , and record this process as:

[0141] BW m =SS(I m )#(twenty four)

[0142] According to the bandwidth expression of formula 18, the corresponding time-varying bandwidth parameter α can be calculated m for:

[0143]

[0144] Its corresponding IF m for:

[0145]

[0146] Then, using the two parameters obtained, TVBCMD can effectively extract the signal components within the frequency band. The remaining signal component is denoted as r m (t), where r 0 (t) = x(t) is the original signal, and this process is recorded as:

[0147]

[0148] After each modal component is extracted, the original signal needs to be subtracted from the obtained modal component. This process can be recorded as:

[0149]

[0150] Then, let m=m+1 and continue to perform the above steps until the energy of the remaining components reaches the stop threshold, and then the modal component extraction is terminated.

[0151] It can be understood that the proposed FBD-FST method can be used in this step to divide the sEMG signal spectrum into frequency bands, and then determine the control parameters required for TVBCMD. Then, the main modal components in each frequency band are extracted from low frequency to high frequency in sequence through the ABCMD algorithm, so as to further study the change rules of signal components in different frequency bands during muscle fatigue.

[0152] S105: extracting muscle fatigue features from the extracted signal modal components, wherein the muscle fatigue features include but are not limited to mean power frequency MPF and median frequency MF, and performing quantitative analysis on the degree of muscle fatigue based on the extracted muscle fatigue features.

[0153] The muscle fatigue feature is extracted from the extracted signal modal component, and the muscle fatigue feature includes but is not limited to the mean power frequency MPF (mean power frequency, MPF) and the median frequency MF (media frequency, MF), and the muscle fatigue degree is quantitatively analyzed according to the extracted muscle fatigue feature. The specific implementation includes the following steps:

[0154] Step 1: Calculate the median frequency MF of each selected signal modal component, that is, the center of gravity of the component spectrum, which reflects the overall frequency level of muscle activity.

[0155] Step 2: Determine the mean power frequency (MPF), which is the frequency point with the highest energy in the spectrum. It is directly related to the recruitment pattern and fatigue state of muscle fibers.

[0156] Step 3: Based on the extracted MF and MPF features, a muscle fatigue index (MFI) is constructed. The index can be a weighted sum of these features or synthesized through a nonlinear model (such as a neural network) to reflect the comprehensive degree of muscle fatigue.

[0157] Step 4: Set a threshold or reference range, compare the calculated MFI with the standard value, and achieve quantitative grading of muscle fatigue degree, such as mild, moderate, severe fatigue, etc.

[0158] Step 5: Output the analysis results, including the MFI value and the corresponding fatigue level, to provide a scientific basis for athlete training monitoring, rehabilitation plan formulation, or human-computer interaction system design.

[0159] It can be understood that this step not only improves the accuracy and sensitivity of muscle fatigue assessment, but also provides a powerful tool for in-depth understanding of the mechanism of muscle fatigue. This method is suitable for a variety of application scenarios and has broad market prospects and social value.

[0160] The present application also provides a method of using Figure 1 The application of the muscle fatigue quantification method based on surface electromyography signal described in the embodiment in preventing human muscle diseases and sports injuries.

[0161] Specifically, this embodiment describes in detail how to use Figure 1 The muscle fatigue quantification method based on surface electromyographic signals described in the embodiment is actually applied in preventing human muscle diseases and sports injuries, aiming to demonstrate the application effect, advantages and specific application scenarios of the method in a practical environment. Specifically, it includes the following steps:

[0162] Step 1: sEMG signal acquisition and preprocessing

[0163] In the embodiments, an experimental example is provided to demonstrate the effectiveness of ABCMD in actual muscle fatigue detection. 15 subjects aged 22 to 27 years old, in good physical condition, without any neurological and musculoskeletal diseases, were selected. The physical characteristics of these subjects are shown in Table 2. Before signal acquisition, all subjects were given more than 24 hours of rest time, and were informed of the entire experimental process and precautions, and signed informed consent in accordance with the Declaration of Helsinki.

[0164] Table 2 Physical characteristics of the subjects

[0165] Variables Mean±SD Minimum Maximum Age 25.0±1.8 21 28 Body height(cm) 174.0±8.3 155 185 Bodymass(kg) 71.4±10.6 56.8 92.5 <![CDATA[BMI(kg / m 2 )]]> 23.5±2.6 17.9 27.9

[0166] The acquisition equipment used is a distributed wireless electromyography acquisition system produced by China Runyi Taiyi Company, including a wireless acquisition module and a software platform RTLAB 2021. The minimum and maximum sampling rates of the system are 500Hz and 4000Hz, respectively, which can realize real-time display and storage of data. The sampling frequency is set to 2000Hz. During the experiment, a 4kg dumbbell is used as a load weight, and 75% medical alcohol is used to wipe the surface of the muscle skin of the subject to remove grease and dead skin on the skin surface, etc., to reduce noise interference during the acquisition process and improve the quality of the collected sEMG signal. The entire experimental process is carried out on a Windows 11 system laptop, and the simulation program is coded on MATLAB R2020b.

[0167] The sEMG signals of the middle bundle of the deltoid muscle of the subject's shoulder were collected, and the wireless electromyographic sensor module was attached to the corresponding muscle using medical double-sided adhesive. The subjects were provided with dumbbells of different weights (3kg, 4kg, and 8kg), and were asked to choose the appropriate dumbbell weight according to their training conditions. When collecting the sEMG signals of the middle bundle of shoulder muscles, the subjects sat in front of the laboratory table, kept their backs straight, raised their arms like flying birds, and kept the middle bundle of shoulder exerting force continuously. When the subject's arm shaking amplitude increased significantly or verbally reported that he could not persist due to muscle fatigue during the force, the signal collection was stopped.

[0168] sEMG signal is a weak bioelectric signal, which is easily interfered. And because sEMG signal is collected by muscle surface electrodes, it will inevitably introduce the same band environmental noise and power frequency interference. Therefore, this experiment uses 10-500Hz bandpass filter and 50Hz notch filter to process sEMG signal to reduce noise interference.

[0169] Step 2: sEMG signal processing and analysis

[0170] A group of preprocessed sEMG signals were randomly selected, and their time-frequency energy was mostly concentrated in 0-200Hz, and it was difficult to observe obvious modal component change trends. According to the ABCMD algorithm, the frequency band boundaries of the sEMG signal were divided using the inflection points of the spectrum trend, and the modal component with the largest spectrum energy was first extracted, and then the subsequent components were extracted recursively.

[0171] During the experiment, the changes in IF and bandwidth of the four modal components extracted by the ABCMD algorithm showed that all IFs had a significant downward trend. The bandwidth of the first modal component decreased most significantly, but its IF decrease was small, indicating that as the degree of muscle fatigue deepened, the frequency of the components with higher energy in the sEMG signal decreased less. The IF and bandwidth of the remaining components all had a trend of shrinking to varying degrees. It further proved that the overall frequency of the sEMG signal showed a downward trend, but the frequency decrease degree of the components in each frequency band was different. The time domain information of the four modal components changed more significantly in AMF1 and AMF4, which showed that the amplitude increased with the deepening of muscle fatigue. The change trend of IF can reflect the frequency decrease amplitude of the final extracted modal component. The extracted IF was subjected to linear regression analysis to obtain the corresponding fitting curve. Then, the component with the largest absolute value of the IF fitting slope Slope was selected as the most representative fatigue modal component extracted by the ABCMD method for subsequent quantitative analysis of muscle fatigue. Generally speaking, the first IF extracted by the ABCMD method is between 50 and 200 Hz, and its energy accounts for a large proportion, but its Slope value is not the highest. Relatively speaking, the high-frequency modal component between 300 Hz and 500 Hz, although its energy accounts for a small proportion, its frequency decreases faster and more obviously. In addition, due to the differences in individual muscle mass, different subjects show different frequency decrease amplitudes in modal components in different frequency bands. Therefore, high-frequency modal components are usually selected as the key indicator source for evaluating muscle fatigue.

[0172] Step 3: Muscle fatigue quantification experiment

[0173] In order to verify that the modal components extracted by the ABCMD method have a better effect in quantifying muscle fatigue, this study uses MPF and MF, which are suitable for characterizing the frequency change trend of non-stationary signals, as fatigue characteristic indicators of sEMG signals to quantify the degree of muscle fatigue.

[0174] The MPF calculation formula is:

[0175]

[0176] The calculation formula of MF is:

[0177]

[0178] Where: f is the frequency of the sEMG signal, and PSD(f) is the power spectral density function of the sEMG signal. The selected modal component AMF and the MPF and MF features of each frame of the original sEMG signal are calculated, and a linear fit is used to obtain the corresponding fitting line. The fitting line can accurately describe the muscle fatigue process. The absolute value of the slope of the fitting line Slope and R 2The value of Slope is used to evaluate the effectiveness of each modal component in quantifying the degree of muscle fatigue. Slope reflects the changing trend of the feature with the development of muscle fatigue. The larger the Slope value, the stronger the ability of the modal to reflect muscle fatigue. 2 It reflects the anti-interference ability of the mode, R 2 The larger the value, the better the anti-interference performance of the mode. The experimental results show the quantization results and fitting lines of the AMF component and sEMG signal. It can be seen that the Slope and R of the MPF and MF feature fitting lines of the AMF component 2 The values ​​are optimal.

[0179] Table 3 shows the Slope and R obtained from the AMF components of 15 signal groups and the original sEMG signal. 2 It can be found that the average Slope and R of the MPF and MF fitting lines of the AMF component 2 The Slope (0.0162) and R 2 (0.765) performs best. Meanwhile, the MPF fitting effect of the original sEMG signal is the worst, with average Slope and R 2 The values ​​of are 0.0062 and 0.511 respectively. The results show that the AMF components extracted using the ABCMD method have significant advantages in distinguishing the degree of muscle fatigue and anti-interference performance.

[0180] Table 3 Slope and R obtained from the AMF component of the 15-group signal and the original sEMG signal 2

[0181]

[0182]

[0183] In order to further verify the effectiveness of this method, the Mann-Whitney U test was used to statistically verify the difference between the AMF component and the original sEMG signal in the ability to represent muscle fatigue. 2 The Mann-Whitney U test was performed, as shown in Table 4. The results showed that there was a significant difference between the performance of the AMF component and the original sEMG signal (p<0.05). This further proves the superiority of the proposed method in distinguishing muscle fatigue degree and anti-interference ability.

[0184] Table 4 Two independent sample U test fatigue detection

[0185]

[0186] This embodiment applies the ABCMD algorithm to the quantitative analysis of muscle fatigue. First, based on the ACMD algorithm, a TVBCMD algorithm is proposed to effectively extract the modal components of the slowly varying bandwidth frequency modulation signal. Then, the FBD-FST method is used to output the spectrum trend of the signal, and the effective boundaries of different modal components are determined by identifying the inflection point of the spectrum trend. Finally, a recursive extraction framework of ABCMD is established to extract the signal modal components in order from strong to weak energy. The FBD-FST method can effectively determine the decomposition parameters of random pulse and repetitive pulse signal components under strong noise. The experiment of analyzing the sEMG signal of the human shoulder muscle fatigue state shows that the ABCMD algorithm can effectively extract the AMF component inside the sEMG signal, and the best results are achieved in the quantitative analysis of MPF and MF, verifying that the AMF component has a stronger ability to distinguish the degree of muscle fatigue and anti-interference.

[0187] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary techniques in the art that are not disclosed in the present application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the following claims.

[0188] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A muscle fatigue quantification method based on surface electromyography signal, characterized in that: include: Collecting sEMG signals and preprocessing the sEMG signals; The preprocessed sEMG signal is processed using a TVBCMD algorithm to extract a plurality of signal modal components having a slowly varying bandwidth characteristic; Using the FBD-FST method to identify the dynamic boundary parameters of each of the signal modal components to determine the required parameters of the TVBCMD algorithm; According to the required parameters of the TVBCMD algorithm and the TVBCMD algorithm, a recursive extraction framework of the ABCMD algorithm is constructed, and the signal modal components are extracted one by one according to the spectral trend energy corresponding to each of the signal modal components until a preset stop condition is met; A muscle fatigue feature is extracted from the extracted signal modal component, wherein the muscle fatigue feature includes but is not limited to the mean power frequency MPF and the median frequency MF, and the muscle fatigue degree is quantitatively analyzed based on the extracted muscle fatigue feature.

2. The method according to claim 1, characterized in that The TVBCMD algorithm dynamically extracts a plurality of signal modal components based on the ACMD technology by introducing a time-varying bandwidth penalty parameter.

3. The method according to claim 2, characterized in that The method of using the FBD-FST method to identify the dynamic boundary parameters of each of the signal modal components to determine the required parameters of the TVBCMD algorithm includes: Determine the dynamic boundaries of each of the signal modal components by detecting the inflection points of the frequency spectrum trend of the sEMG signal using the FBD-FST method; Determining the dynamic boundary parameter corresponding to the dynamic boundary according to the dynamic boundary of the signal modal component; According to the dynamic boundary parameters, the required parameters of the TVBCMD algorithm are determined, wherein the required parameters are used to indicate the time-varying bandwidth penalty parameter and the initial instantaneous frequency.

4. The method according to claim 3, characterized in that The FBD-FST method is used to detect the inflection point of the frequency spectrum trend of the sEMG signal to determine the dynamic boundary of each of the signal modal components; Determining the dynamic boundary parameter corresponding to the dynamic boundary according to the dynamic boundary of the signal modal component includes: Performing Fourier transform on the sEMG signal to obtain a frequency spectrum of the sEMG signal; Extracting a spectrum trend of the spectrum, identifying an inflection point of the spectrum trend, and using the inflection point as a boundary point of different modal components of the signal; The variation curve of the inflection point in the entire time period of collecting the sEMG signal is obtained by a sliding window technique, and the variation curve is used as a dynamic boundary parameter for dividing different modal components of the signal.

5. The method according to claim 4, characterized in that The recursive extraction framework of the ABCMD algorithm is constructed according to the required parameters of the TVBCMD algorithm and the TVBCMD algorithm, and the signal modal components are extracted one by one according to the spectral trend energy corresponding to each of the signal modal components until a preset stop condition is met, including: Step 1: extracting the signal modal component with the strongest energy according to the required parameters of the TVBCMD algorithm and the TVBCMD algorithm; Step 2: subtracting the extracted signal modal component from the sEMG signal to obtain a residual signal; Step 3: Repeat steps 1 and 2 until the energy of the remaining signal is lower than a preset threshold.

6. The method according to claim 5, characterized in that Extracting muscle fatigue features from the extracted signal modal components, the muscle fatigue features including but not limited to mean power frequency MPF and median frequency MF, and quantitatively analyzing the degree of muscle fatigue according to the extracted features, including: Perform statistical analysis on the extracted signal modal components and calculate the mean power frequency MPF and median frequency MF; The muscle fatigue degree is quantitatively evaluated based on the calculated average power frequency MPF and median frequency MF, combined with the known relationship between the muscle fatigue degree and the muscle fatigue characteristics.

7. Use of the method according to any one of claims 1 to 6 in preventing human muscle diseases and sports injuries.

Images