Muscle state analysis method for analyzing muscle strength based on electromyographic signals

Abstract

The invention discloses a muscle state analysis method for analyzing muscle strength based on electromyographic signals, belongs to the technical field of neuromuscular physiology and data processing, and solves the problem that traditional electromyographic analysis lacks activation degree and fatigue degree multi-dimensional dynamic evaluation ability. Through multi-channel electromyographic signal collection and feature extraction and in combination with a calculation model, the muscle activation degree and the fatigue state are accurately evaluated. The functional state of the target muscle under the specific action can be dynamically monitored, and a reliable scientific basis is provided for movement safety; and moreover, the defects in the aspect of muscle state analysis in the prior art are overcome, and the method has important clinical application value in the aspect of improving exercise rehabilitation and training effects.

Description

Technical Field

[0001] The present invention relates to the technical fields of neuromuscular physiology and data processing, and particularly to a method for analyzing muscle state based on electromyogram signal parsing of muscle strength. Background Art

[0002] The activity of skeletal muscle relies on muscle contraction to generate power. In the fields of sports science, rehabilitation medicine, etc., accurately evaluating the functional state of muscles is of great significance. The degree of muscle activation reflects the use and mobilization of muscles under specific actions, while the degree of muscle fatigue reflects the state attenuation of muscles during continuous exercise. Especially after long-term high-intensity activities, it can reveal the recovery needs of muscles and potential injury risks.

[0003] Currently, the existing technologies at home and abroad mainly estimate muscle strength by decomposing the firing sequences of motor units with the help of surface electromyogram signals. However, muscle state analysis should not be limited to muscle strength estimation only. By deeply exploring the characteristic changes of surface electromyogram signals and quantifying the degree of muscle activation and fatigue, it plays a key role in optimizing training effects and preventing sports injuries. Unfortunately, although the current technology can provide muscle strength estimation, there is a lack of research on deriving muscle activation and fatigue degrees from changes in muscle strength, making it difficult to meet the needs of accurate assessment of muscle state in fields such as sports science and rehabilitation medicine. Summary of the Invention

[0004] The present invention provides a method for analyzing muscle state based on electromyogram signal parsing of muscle strength. Through multi-channel electromyogram signal acquisition and feature extraction, combined with a calculation model, it can accurately evaluate the degree of muscle activation and fatigue, and can dynamically evaluate the functional state of the target muscle under specified actions, providing a reliable scientific basis for sports safety, and solving the problem that the prior art rarely deeply explores how to derive muscle activation and fatigue degrees from changes in muscle strength.

[0005] A method for analyzing muscle state based on electromyogram signal parsing of muscle strength, the method comprising the following steps:

[0006] S1. Collect the original electromyogram signals of the target muscle through a multi-channel electromyogram signal acquisition device, and preprocess the original electromyogram signals to obtain processed electromyogram signals;

[0007] S2. Based on the processed electromyogram signals, parse to obtain the firing sequences and action potential waveforms of multiple motor units in the target muscle;

[0008] S3. According to the firing sequences and action potential waveforms, establish a muscle strength estimation model and calculate the real-time muscle strength of the target muscle;

[0009] S4. Extract the frequency features of the EMG signal, and calculate the muscle activation level and fatigue level according to the changes of the frequency features during the movement state;

[0010] S5. Generate a comprehensive evaluation index of the target muscle state based on the real-time muscle strength, muscle activation level and fatigue level.

[0011] Further, in S1, the following steps are included:

[0012] S11. Configure multi-channel electrodes to collect the original EMG signal of the target muscle in differential mode;

[0013] S12. Enhance the amplitude of the original EMG signal through an amplifier, and convert the analog signal into a digital signal using a high-sampling-rate analog-to-digital converter. Among them, the sampling frequency needs to be higher than the Nyquist frequency to avoid signal distortion, and a digital signal is obtained:

[0014]

[0015] where Digital Signal is the digital signal, Analog Signal is the analog signal, ADC Range is the range of the analog-to-digital converter, and Gain is the amplification factor;

[0016] S13. Apply the Kalman filtering algorithm to perform real-time noise reduction and filtering processing on the digital signal, and extract effective EMG information:

[0017] The Kalman filtering algorithm is based on the state estimation of a linear system. At each time step k, according to the state x at the previous moment k-1 perform state prediction,

[0018] Its state equation is:

[0019] x k = Ax k-1 + Bu k + w k

[0020] where x k is the state at the current moment; A is the state transition matrix, describing the state change from the (k - 1)th moment to the kth moment; B is the control input matrix, describing the influence of the control input on the state; u k is the control input, that is, the external influence exerted by the system; w k is the process noise;

[0021] After performing state prediction, calculate the observation prediction value z k , and the observation equation is:

[0022] z k = Hx k + vk

[0023] where z k is the observed value, H is the observation matrix, and vk is the observation noise;

[0024] Calculate the Kalman gain K by comparing the predicted observed value and the actual observed value k ,

[0025]

[0026] where P k|k-1 is the predicted covariance matrix, R is the observation noise covariance matrix,

[0027] Update the state estimate using the Kalman gain value:

[0028] x k = x k|k-1 + K k (z k - Hx k|k-1 )

[0029] Update the covariance matrix to reflect the new uncertainty:

[0030] P k|k = (I - K k H)P k|k-1 .

[0031] Furthermore, in S11, during the acquisition process, the signal electrodes are placed parallel to the muscle fibers at the center of the muscle, and the reference electrodes are set in the area without muscle activity.

[0032] Furthermore, in S2, the following steps are included:

[0033] S21. Separate the independent components in the EMG signal by combining independent component analysis with the negative entropy maximization optimization method, and determine the peak positions of the independent component waveforms, corresponding to the firing moments of the motor units:

[0034] Decompose the surface EMG signal to obtain N motor units. Among them, the surface EMG signal of the i-th channel can be expressed as:

[0035]

[0036] s j (t) = ∑ k δ(t - T j (k))

[0037] where a ij represents the action waveform of the j-th motor unit in the i-th channel; L is the waveform length; s j (t) is the firing sequence of the motor unit; Tj (k) is the k-th firing time of the j-th motor unit; δ is the delta function; n(t) is Gaussian white noise with a mean of 0 in the i-th channel;

[0038] The extended surface electromyogram signal is:

[0039]

[0040] where Q is the delay factor, and the surface electromyogram signal is represented using a convolutional mixture model:

[0041]

[0042] where is the waveform coefficient matrix formed by arranging all action potential waveform coefficients a ij in a certain order;

[0043] Use a non-polynomial function to approximate a random variable with zero mean and unit variance, and optimize the negentropy problem to obtain the independent component y = W T x, as follows:

[0044] maxJ G (W) = [E{G(W T x)} - E{G(v)}] 2

[0045] G(x) = log(cosh(x))

[0046] s.t. h(W) = E{y 2} - 1 = ||W|| 2 - 1 = 0

[0047] where J G (W) is the negentropy; G is a non-polynomial function; v is a random variable with a standard normal distribution;

[0048] After updating and solving based on the fixed-point theory, use the symmetric orthogonal method to simultaneously estimate multiple independent vectors, and obtain:

[0049]

[0050] W = [W1, W2,..., W n T

[0051] According to the waveforms of the independent vectors, determine the firing times of each motor unit in the target muscle; where the number of the firing times is the same as the number of the independent vectors;

[0052] S22. Using the least squares method, according to the firing sequence obtained by decomposition and the original signal, fit the action potential waveforms of each motor unit, construct a signal model to verify the waveform accuracy, and screen the optimal waveform:

[0053] Based on the said firing sequence, determine the action potential waveform, and use the least squares method to determine the action potential waveform of each motor unit in each channel of the original signal.

[0054] The signal Y is expressed as:

[0055]

[0056] where A i is the action potential waveform vector; S i is the firing time of the motor unit; N is the number of motor units; E is the error term.

[0057] Minimize the following loss function:

[0058]

[0059] Use the least squares method to solve for A i , and use the principle of linear regression to calculate the waveform of each motor unit.

[0060]

[0061] Based on the firing sequence, select the optimal A i for model verification and analysis, evaluate the effect of the waveform, and judge the model fitting situation:

[0062]

[0063] Furthermore, in S3, it includes the following steps:

[0064] S31. Define the correlation between the twitch force of the motor unit and the action potential waveform:

[0065] For the twitch force waveform of the i-th motor unit, there is the following formula:

[0066]

[0067] where P i and T i are respectively the twitch force amplitude and contraction time of the i-th motor unit;

[0068] S32. Establish the relationship between the action potential waveform amplitude and the twitch force amplitude through a quadratic function, and define the relationship between the contraction time and the twitch force amplitude based on the inverse power function:

[0069] The maximum amplitude muMaxi of the action potential waveform of the ith motor unit obtained by decomposition in all channels and P i Establish a quadratic function relationship as follows:

[0070] P i = x(1)·muMax i 2 + x(2)·muMax i + x(3)

[0071] P i and T i There is an inverse power function relationship:

[0072]

[0073] where T L is the maximum contraction time and C is a constant;

[0074] S33. Construct a gain change model according to the time interval of the firing sequence:

[0075] Denote the adjacent time firing interval and the gain change of the jth firing of the ith motor unit as ISI j and g i，j , respectively. Then the two satisfy the relationship:

[0076] ISI j = t i，j+1 - t i，j

[0077]

[0078] S34. Superimpose the twitch forces of each motor unit to generate a total muscle force model:

[0079] The finally generated muscle contraction force is as follows:

[0080]

[0081] S35. Use the nonlinear least squares method to optimize the model parameters and complete the establishment of the muscle force estimation model:

[0082] The unknown parameters x(1), x(2), x(3) are obtained by using the nonlinear least squares method, and the objective function is:

[0083]

[0084] where the unknown parameter is the estimated force normalized by the maximum and minimum values, F is the measured force normalized by the maximum and minimum values. After obtaining the unknown parameters, the target muscle force estimation model is obtained.

[0085] Further, in S35, the non-linear least squares method adjusts the model parameters through an iterative optimization algorithm to minimize the normalized error between the estimated force and the measured force; during the iteration process, the step size is dynamically adjusted based on parameter sensitivity to improve the convergence speed.

[0086] Further, in S4, the muscle activation degree activation is calculated by the following formula:

[0087]

[0088] The median frequency MF and the mean frequency MPF of the surface electromyogram signal are calculated as follows:

[0089]

[0090] where D(f) is the power spectrum of the muscle surface electromyogram signal, and f represents the surface electromyogram signal frequency;

[0091] The frequency characteristics in the force exertion state and the relaxation state are compared to infer the degree of muscle fatigue. As muscle fatigue increases, the median frequency and the mean frequency will decrease, reflecting the decline in muscle performance after high-intensity activities. The muscle fatigue degree fatigue is calculated by the following formula:

[0092]

[0093] where MF max and MPF max represent the median frequency and the mean frequency at maximum force exertion respectively, and MF init and MPF init represent the median frequency and the mean frequency during relaxation respectively.

[0094] Further, in S5, by combining the muscle activation degree activation and the fatigue degree fatigue, the muscle state is evaluated more comprehensively, and the muscle activation degree activation and the fatigue degree fatigue are integrated into a comprehensive evaluation index G:

[0095] G = activation × (1 - fatigue).

[0096] A storage medium stores a computer program thereon. When the computer program is executed by a processor, the above method is implemented.

[0097] A terminal includes: a memory, a processor, and a computer program stored on the memory and executable on the processor. The processor executes the program to implement the above method.

[0098] Advantages of the present invention: A method for analyzing muscle state based on electromyogram signal to resolve muscle strength according to the present invention realizes precise evaluation of muscle activation degree and fatigue state through multi-channel electromyogram signal acquisition and feature extraction, combined with a calculation model. This method can not only dynamically monitor the functional state of the target muscle under specific actions, providing a reliable scientific basis for sports safety, but also make up for the deficiencies of the existing technology in muscle state analysis, and has important clinical application value in improving the effects of sports rehabilitation and training. Description of the Drawings

[0099] Figure 1 It is a structural diagram of the muscle state analysis method according to the present invention;

[0100] Figure 2 It is a flow chart of Kalman filtering in the present invention;

[0101] Figure 3 It is a flow chart for determining the action potential waveform in the present invention;

[0102] Figure 4 It is a flow chart for establishing a muscle strength model. Detailed Embodiments

[0103] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without making creative efforts shall fall within the protection scope of the present invention.

[0104] Referring to Figure 1 as shown, a method for analyzing muscle state based on electromyogram signal to resolve muscle strength includes the following steps:

[0105] S1. Collect the original electromyogram signal of the target muscle through a multi-channel electromyogram signal acquisition device, and preprocess the original electromyogram signal to obtain the processed electromyogram signal;

[0106] S2. Based on the processed electromyogram signal, analyze and obtain the firing sequences and action potential waveforms of multiple motor units in the target muscle;

[0107] S3. According to the firing sequences and action potential waveforms, establish a muscle strength estimation model and calculate the real-time muscle strength of the target muscle;

[0108] S4. Extract the frequency characteristics of the electromyogram signal, and calculate the muscle activation degree and fatigue degree according to the changes of the frequency characteristics under the exercise state;

[0109] S5. Based on the real-time muscle strength, muscle activation degree and fatigue degree, generate a comprehensive evaluation index of the target muscle state.

[0110] Specifically, the present invention collects and preprocesses the original myoelectric signals through a multi-channel myoelectric signal acquisition device, which can obtain more accurate myoelectric information; analyzes the firing sequence and action potential waveform of motor units based on the processed signals, laying a foundation for subsequent precise analysis; establishes a muscle force estimation model to calculate the real-time muscle force, enabling real-time understanding of muscle force changes; extracts frequency features to calculate the muscle activation degree and fatigue degree, and can deeply grasp the muscle function state; generates comprehensive evaluation indicators to comprehensively reflect the state of the target muscle. This method of the present invention realizes the accurate assessment of the muscle activation degree and fatigue state, can dynamically monitor the function state of the target muscle under specific actions, provides a scientific basis for sports safety, makes up for the deficiencies of the existing technology in muscle state analysis, and has important clinical application value in the fields of sports rehabilitation and training.

[0111] Further, in S1, in the myoelectric signal acquisition stage, a multi-channel myoelectric acquisition device is used to synchronously collect signals from different muscle parts. Multiple electrodes are configured and measured in a differential mode, including a pair of positive and negative signal electrodes and a reference electrode. The differential mode can effectively suppress common-mode interference and enhance the signal quality. Among them, the signal electrodes should be parallel to the direction of muscle fibers and select the central position of the muscle to maximize the signal capture efficiency. The reference electrode should be placed in an area without muscle activity, such as near the joint, to reduce the influence of background noise on the signal.

[0112] The initially collected myoelectric signals have a low amplitude and need to be amplified by a signal amplifier and then converted into digital signals by an analog-to-digital converter for further processing. The amplifier is set between 10x and 1000x to adapt to different muscle activity levels. An analog-to-digital converter with a high sampling rate is selected to ensure that fast-changing signals can be captured. Among them, the sampling frequency needs to be higher than the Nyquist frequency to avoid signal distortion. The digital signal is obtained as follows:

[0113]

[0114] Among them, Digital Signal is the digital signal, Analog Signal is the analog signal, ADC Range is the range of the analog-to-digital converter, and Gain is the amplification factor.

[0115] The Kalman filter algorithm is further used for real-time noise reduction and filtering processing to effectively extract the useful information in the myoelectric signals. The Kalman filter algorithm is based on the state estimation of a linear system. At each time step k, the state prediction is carried out according to the state x k-1 as follows Figure 2 as shown.

[0116] Its state equation is:

[0117] x k = Ax k-1 + Bu k + w k

[0118] where x k is the state at the current time; A is the state transition matrix, describing the state change from time k-1 to time k; B is the control input matrix, describing the influence of the control input on the state.; u k is the control input, i.e., the external influence applied to the system; w k is the process noise.

[0119] After making the state prediction, calculate the predicted observation value z k , and there is an observation equation:

[0120] z k = Hx k + v k

[0121] where z k is the observed value, H is the observation matrix, and vk is the observation noise.

[0122] By comparing the predicted observation value and the actual observation value, calculate the Kalman gain K k .

[0123]

[0124] where P k|k-1 is the predicted covariance matrix and R is the observation noise covariance matrix.

[0125] Use the Kalman gain value to update the state estimate:

[0126] x k = x k|k-1 + K k (z k - Hx k|k-1 )

[0127] Update the covariance matrix to reflect the new uncertainty:

[0128] P k|k = (I - K k H)P k|k-1 .

[0129] Specifically, the present invention configures multi-channel electrodes to collect signals in a differential mode, which can effectively suppress common-mode interference, improve the quality and stability of the collected signals, and reduce the influence of external noise on the original EMG signals. An amplifier is used to enhance the signal amplitude, and a high-sampling-rate analog-to-digital converter is used to convert the signals, and the sampling frequency is higher than the Nyquist frequency to ensure that the analog signals can be accurately converted into digital signals, avoid signal distortion, and retain the effective information in the original EMG signals to the greatest extent. The Kalman filtering algorithm is applied to perform real-time noise reduction and filtering on the digital signals. Based on the principle of linear system state estimation, the Kalman gain is calculated using the state equation and the observation equation to update the state estimation and the covariance matrix, which can accurately extract the effective EMG information and provide reliable data support for subsequent accurate analysis of muscle strength and evaluation of muscle status.

[0130] Further, in S11, during the collection process, the signal electrodes are placed parallel to the muscle fibers at the center of the muscle, and the reference electrodes are set in the area without muscle activity.

[0131] Specifically, the present invention stipulates that during the collection process, the signal electrodes are placed parallel to the muscle fibers at the center of the muscle, which can obtain the electrical signals generated during muscle contraction to the greatest extent, ensuring that the collected EMG signals are representative and accurate, because the electrical signals at the center of the muscle can better reflect the overall activity of the muscle; the reference electrodes are set in the area without muscle activity, which can effectively avoid the interference signals generated by other muscle activities from being mixed into the collected data, ensuring that the collected original EMG signals are true and reliable, thereby improving the accuracy of subsequent analysis and processing of EMG signals, making the results of muscle strength calculation, muscle activation degree and fatigue degree evaluation based on these signals more credible.

[0132] Further, in S2, the firing sequences of each motor unit in the surface EMG signal are calculated to further determine the action potential waveforms for subsequent applications. The specific steps are as follows Figure 3 as shown.

[0133] The surface EMG signal is decomposed into N motor units. Among them, the surface EMG signal of the i-th channel can be expressed as:

[0134]

[0135] s j (t) = ∑ k δ(t - T j (k))

[0136] where a ij represents the action waveform of the j-th motor unit in the i-th channel; L is the waveform length; s j (t) is the firing sequence of the motor unit; T j(k) is the k-th firing time of the j-th motor unit; δ is the delta function; n(t) is the Gaussian white noise with a mean of 0 in the i-th channel.

[0137] The extended surface electromyogram signal is:

[0138]

[0139]

[0140] where Q is the delay factor. The surface electromyogram signal is represented using a convolutional mixture model:

[0141]

[0142] where, is the waveform coefficient matrix formed by arranging all action potential waveform coefficients a ij in a certain order.

[0143] Use a non-polynomial function to approximate a random variable with zero mean and unit variance, and optimize the negentropy problem to obtain the independent component y = W T x, as follows:

[0144] maxJ G (W) = [E{G(W T x)} - E{G(v)}] 2

[0145] G(x) = log(cosh(x))

[0146] s.t. h(W) = E{y 2} - 1 = ||W|| 2 - 1 = 0

[0147] where, J G (W) is the negentropy; G is the non-polynomial function; v is a random variable with a standard normal distribution.

[0148] After updating and solving based on the fixed-point theory, use the symmetric orthogonal method to simultaneously estimate multiple independent vectors, and obtain:

[0149]

[0150] W = [W1, W2,..., W n T

[0151] The peak positions of the waveforms of each independent vector output after the above steps can be considered to correspond to the firing times of the motor units, and the number of firing sequences of different motor units is the same as the number of independent components.

[0152] Determine the action potential waveform according to the firing sequence. Use the least squares method to determine the action potential waveform of each motor unit in each channel of the original signal.

[0153] The signal Y can be expressed as:

[0154]

[0155] where A i is the action potential waveform vector; S i is the firing time of the motor unit; N is the number of motor units; and E is the error term.

[0156] Minimize the following loss function:

[0157]

[0158] Solve for A i using the least squares method, and calculate the waveform of each motor unit using the principle of linear regression.

[0159]

[0160] Select the optimal A i according to the firing sequence for model verification and analysis, evaluate the effect of the waveform, and judge the model fitting situation:

[0161]

[0162] Specifically, the present invention uses independent component analysis combined with the negative entropy maximization optimization method to effectively separate the independent components in the EMG signal, accurately determine the firing time of the motor unit, and provide a key time node basis for subsequent analysis. By fitting the action potential waveforms of each motor unit using the least squares method, constructing a signal model and screening the optimal waveform, it can ensure that the waveform adopted can more accurately reflect the characteristics of the motor unit, and effectively improve the accuracy of the input data of the muscle force estimation model. These operations cooperate with each other to lay a solid foundation for establishing an accurate muscle force estimation model, accurately calculating the degree of muscle activation and fatigue.

[0163] Furthermore, in S3, estimate the muscle force of the position to be tested according to the action potential waveform and the firing sequence, and perform the calculation process as Figure 4 follows.

[0164] For the twitch force waveform of the i-th motor unit, there is the following formula:

[0165]

[0166] where P i and T i are respectively the twitch force amplitude and the contraction time of the i-th motor unit.

[0167] The maximum amplitude muMaxi of the action potential waveform of the ith motor unit obtained by decomposition in all channels and P i Establish a quadratic function relationship as follows:

[0168] P i = x(1)·muMax i 2 + x(2)·muMax i + x(3)

[0169] P i and T i There is an inverse power function relationship:

[0170]

[0171] where T L is the maximum contraction time and C is a constant.

[0172] Denote the adjacent time firing interval and gain change of the jth firing of the ith motor unit as ISI j and g i，j , then the two satisfy the relationship:

[0173] ISI j = t i，j+1 - t i，j

[0174]

[0175] The finally generated muscle contraction force is as follows:

[0176]

[0177] The unknown parameters x(1), x(2), x(3) are obtained using the nonlinear least squares method, and the objective function is:

[0178]

[0179] where the unknown parameter is the estimated force normalized using the maximum and minimum values, and F is the measured force normalized using the maximum and minimum values. After obtaining the unknown parameters, the target muscle force estimation model is obtained.

[0180] Specifically, the present invention defines the correlation between the twitch force of a motor unit and the action potential waveform, as well as the relationship between the amplitude of the action potential waveform, the contraction time, and the amplitude of the twitch force, providing a theoretical basis for muscle force calculation; constructs a gain change model, considering the influence of the time interval of the firing sequence on muscle force, making the model more in line with the actual muscle movement situation; superimposes the twitch forces of each motor unit to generate a total muscle force model, and uses the nonlinear least squares method to optimize the model parameters. Through the iterative optimization algorithm, the parameters are adjusted to minimize the normalized error between the estimated force and the measured force, and the step size is dynamically adjusted based on parameter sensitivity to improve the convergence speed, effectively enhancing the accuracy and practicality of the model, and being able to more accurately reflect the real-time muscle force of the target muscle.

[0181] Further, in S35, the nonlinear least squares method adjusts the model parameters through the iterative optimization algorithm to minimize the normalized error between the estimated force and the measured force; during the iteration process, the step size is dynamically adjusted based on parameter sensitivity to improve the convergence speed.

[0182] Specifically, by adjusting the model parameters through the iterative optimization algorithm to minimize the normalized error between the estimated force and the measured force, it can continuously approach the real muscle force, improve the accuracy of the muscle force estimation model, and make the calculated real-time muscle force closer to the actual situation. During the iteration process, the step size is dynamically adjusted based on parameter sensitivity, avoiding problems such as slow convergence speed or getting stuck in local optimal solutions caused by improper step size selection, greatly accelerating the convergence speed of the model, reducing the calculation time, and enhancing the operation efficiency of the model.

[0183] Further, in S4, the muscle activation degree activation is calculated by the following formula:

[0184]

[0185] Calculate the median frequency MF and the mean frequency MPF of the surface electromyogram signal as follows:

[0186]

[0187] Where D(f) is the power spectrum of the muscle surface electromyogram signal, and f represents the surface electromyogram signal frequency.

[0188] By comparing the frequency characteristics in the force generation state and the relaxation state, the fatigue degree of the muscle can be inferred. As muscle fatigue increases, the median frequency and the mean frequency will decrease, reflecting the decline in muscle performance after high-intensity activities. Calculate the muscle fatigue degree fatigue by the following formula:

[0189]

[0190] Where, MF max and MPF maxRepresent the median frequency and average frequency at maximum exertion, MF init and MPF init Represent the median frequency and average frequency during relaxation respectively.

[0191] Specifically, the present invention calculates the muscle activation degree and fatigue degree through specific formulas, and can comprehensively and accurately evaluate the muscle state. By using the formula for calculating muscle activation degree, the real-time muscle strength is associated with the maximum muscle strength, intuitively quantifying the mobilization degree of the muscle under the current movement, and providing a key index for understanding the working state of the muscle. Calculate the median frequency MF and average frequency MPF of the surface electromyogram signal, and calculate the muscle fatigue degree based on the change of the frequency characteristics in the exertion and relaxation states, grasping the key that the frequency characteristics change during muscle fatigue, and can effectively reflect the fatigue condition of the muscle during exercise. These quantitative indexes complement each other, providing strong data support for in-depth analysis of muscle function, formulation of scientific exercise plans and prevention of sports injuries.

[0192] Furthermore, in S5, by combining the muscle activation degree activation and the fatigue degree fatigue, the muscle state is evaluated more comprehensively, and the muscle activation degree activation and the fatigue degree fatigue are integrated into a comprehensive evaluation index G:

[0193] G = activation × (1 - fatigue).

[0194] Specifically, the present invention integrates the muscle activation degree and the fatigue degree into a comprehensive evaluation index G, which can more comprehensively and accurately evaluate the muscle state. The muscle activation degree reflects the use and mobilization degree of the muscle under specific movements, and the fatigue degree reflects the attenuation of the muscle during continuous exercise. The combination of the two can present the real-time state of the muscle from multiple dimensions. The comprehensive evaluation index G = activation × (1 - fatigue), which connects these two key indexes in a mathematical form, can clearly show the balance relationship between muscle activation and fatigue, enabling the user to quickly understand the current overall condition of the muscle, providing an intuitive and comprehensive reference basis for scenarios such as sports training and rehabilitation treatment, and helping to formulate more scientific and reasonable exercise plans or rehabilitation plans.

[0195] A storage medium stores a computer program thereon, and when the computer program is executed by a processor, the above method is implemented.

[0196] Specifically, the storage medium disclosed in the present invention can store a series of operation processes and algorithms such as complex electromyogram signal processing, muscle strength calculation, and muscle state assessment in the form of software, facilitating deployment and use on different computing devices. This greatly improves the versatility and portability of the muscle state analysis method of the present invention, enabling personnel in different fields such as scientific researchers, medical workers, and sports training experts to easily apply this method for relevant research and practice as long as they have the corresponding hardware devices. At the same time, this storage medium is also convenient for updating and maintenance, and can integrate the latest research results and optimized algorithms at any time, continuously improving the accuracy and efficiency of muscle state analysis to meet the ever-developing scientific research and application requirements.

[0197] A terminal includes: a memory, a processor, and a computer program stored on the memory and executable on the processor. The processor executes the program to implement the above method.

[0198] Specifically, the terminal disclosed in the present invention provides an efficient and reliable implementation approach for muscle state analysis by integrating a memory, a processor, and related computer programs. The memory is used to store programs and data, ensuring the stable storage and invocation of data during the analysis process; the processor executes the computer program and can automatically execute a series of complex operation processes from electromyogram signal acquisition, preprocessing, to muscle strength calculation, muscle activation degree and fatigue degree assessment, and then to generating comprehensive evaluation indicators, significantly improving the accuracy and repeatability of the analysis. In addition, its easy operation and maintenance characteristics reduce the dependence on professional operators, enabling more people to conveniently use this terminal for muscle state analysis. Moreover, the design of this terminal enables it to be flexibly deployed in various environments. Whether in a scientific research laboratory, a sports training venue, or a medical institution, it can provide strong support for sports safety guarantee, sports rehabilitation guidance, and training effect optimization.

[0199] A muscle state analysis method for parsing muscle strength based on electromyogram signals of the present invention realizes the accurate assessment of muscle activation degree and fatigue state through multi-channel electromyogram signal acquisition and feature extraction, combined with a calculation model. This method can not only dynamically monitor the functional state of the target muscle under specific actions, providing a reliable scientific basis for sports safety; but also makes up for the deficiencies of the existing technology in muscle state analysis and has important clinical application value in improving sports rehabilitation and training effects.

[0200] The above is only the specific implementation manner of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art within the technical scope disclosed in this application can easily think of various equivalent modifications or substitutions, and these modifications or substitutions should all be covered within the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims

1. A muscle state analysis method based on electromyographic signal analysis of muscle strength, characterized in that: The method comprises the following steps: S1, collecting original electromyographic signals of target muscles through a multi-channel electromyographic signal acquisition device, preprocessing the original electromyographic signals, and obtaining processed electromyographic signals; S2. Based on the processed electromyographic signal, analyzing and obtaining the firing sequence and action potential waveform of multiple motor units in the target muscle; S3, establishing a muscle force estimation model according to the firing sequence and action potential waveform, and calculating the real-time muscle force of the target muscle; S4, extracting the frequency characteristics of the electromyographic signal, and calculating the muscle activation degree and fatigue degree according to the changes of the frequency characteristics under the motion state; S5. Generate a comprehensive evaluation index of the target muscle state based on the real-time muscle strength, muscle activation degree and fatigue degree.

2. A muscle state analysis method based on electromyographic signal analysis of muscle strength according to claim 1, characterized in that: In S1, the following steps are included: S11, configuring multi-channel electrodes to collect the original electromyographic signals of the target muscles in differential mode; S12. The amplitude of the original electromyographic signal is enhanced by an amplifier, and a high sampling rate analog-to-digital converter is used to convert the analog signal into a digital signal, wherein the sampling frequency needs to be higher than the Nyquist frequency to avoid signal distortion, and a digital signal is obtained: Among them, DigitalSignal is the digital signal, AnalogSignal is the analog signal, ADCRange is the range of the analog-to-digital converter, and Gain is the amplification factor; S13, apply Kalman filter algorithm to perform real-time noise reduction and filtering on digital signals to extract effective electromyographic information: The Kalman filter algorithm is based on the state estimation of the linear system. At each time step k, according to the state x at the previous moment k-1 To predict the status, Its state equation is: x k =Ax k-1 +Bu k +w k Among them, x k is the state at the current moment; A is the state transfer matrix, describing the state change from moment k-1 to moment k; B is the control input matrix, describing the influence of the control input on the state; u k is the control input, i.e., the external influence exerted by the system; w k is the process noise; After the state prediction is performed, the observation prediction value z is calculated k , the observation equation is: z k =Hx k +v k Among them, z k is the observation value, H is the observation matrix, and vk is the observation noise; By comparing the predicted observations with the actual observations, the Kalman gain K is calculated. k , Among them, P k|k-1 is the prediction covariance matrix, R is the observation noise covariance matrix, Update the state estimate using the Kalman gain value: x k =x k|k-1 +K k (z k -Hx k|k-1 ) Update the covariance matrix to reflect the new uncertainty: P k|k =(I-K k H)P k|k-1 。 3. A muscle state analysis method based on electromyographic signal analysis of muscle strength according to claim 2, characterized in that: In S11, during acquisition, the signal electrode was placed in the center of the muscle parallel to the muscle fibers, and the reference electrode was set in an area without muscle activity.

4. A muscle state analysis method based on electromyographic signal analysis of muscle strength according to claim 3, characterized in that: In S2, the following steps are included: S21. Separate the independent components in the electromyographic signal through independent component analysis combined with the negative entropy maximization optimization method, determine the peak position of the independent component waveform, and the corresponding motor unit release time: The surface electromyographic signal is decomposed into N motor units, where the surface electromyographic signal of the i-th channel can be expressed as: s j (t)=∑ k δ(t-T j (k)) Among them, a ij represents the action waveform of the jth motor unit in the ith channel; L is the waveform length; s j (t) is the firing sequence of the motor unit; T j (k) is the kth firing time of the jth motor unit; δ is the delta function; n(t) is the Gaussian white noise with a mean of 0 in the i-th channel; The extended surface electromyographic signal is: Among them, Q is the delay factor, and the convolution mixture model is used to represent the surface electromyographic signal: in, is the waveform coefficient of all action potentials a ij A waveform coefficient matrix formed by arranging in a certain order; Use non-polynomial functions to approximate random variables with zero mean and unit variance and optimize the negentropy problem to obtain independent components y = W T x, as follows: maxJ G (W)[E{G(W T x)}-E{G(v)}] 2 G(x)=log(cosh(x)) s.t.h(W)=E{y 2 }-1=||W|| 2 -1=0 Among them, J G (W) is the negative entropy; G is a non-polynomial function; v is a random variable of standard normal distribution; After updating the solution based on the fixed point theory, the symmetric orthogonal method is used to simultaneously estimate multiple independent vectors, and we get: W=[W1,W2,...,W n ] T Determining the release time of each motor unit in the target muscle according to the waveform of each independent vector; wherein the number of the release time is the same as the number of the independent vectors; S22. Using the least squares method, according to the decomposed firing sequence and the original signal, fit the action potential waveform of each motor unit, build a signal model to verify the waveform accuracy, and select the optimal waveform: According to the firing sequence, the action potential waveform is determined, and the action potential waveform of each motor unit in each channel in the original signal is determined using the least squares method. Signal Y is expressed as: Among them, A i is the action potential waveform vector; S i is the release time of the motor unit; N is the number of motor units; E is the error term, Minimize the following loss function: Solve A using the least squares method i , using the principle of linear regression, calculate the waveform of each motor unit, According to the release sequence, select the best A i Perform model verification and analysis to evaluate the effect of the waveform and determine the model fit:

5. A muscle state analysis method based on electromyographic signal analysis of muscle strength according to claim 4, characterized in that: In S3, the following steps are included: S31. Define the relationship between the twitch force of the motor unit and the action potential waveform: For the twitch force waveform of the i-th motor unit, there is the following formula: Among them, P i With T i are the twitch force amplitude and contraction time of the ith motor unit, respectively; S32. The relationship between the action potential waveform amplitude and the twitch force amplitude is established through a quadratic function, and the relationship between the contraction time and the twitch force amplitude is defined based on an inverse power function: The maximum amplitude muMaxi of the action potential waveform of the i-th motor unit obtained by decomposition in all channels is compared with P i The quadratic function relationship is established as follows: P i =x(1)·muMax i 2 +x(2)·muuMax i +x(3) P i With T i There is an inverse power function relationship: Among them, T L is the longest contraction time, C is a constant; S33. Construct a gain variation model based on the time interval of the release sequence: The interval between the jth release of the i-th motor unit and the gain change are respectively recorded as ISI j and g i，j , then the two satisfy the relationship: ISI j =t i,j+1 -t i，j S34, superimpose the twitch forces of each motor unit to generate a total muscle force model: The resulting muscle contraction force is shown below: S35. Use nonlinear least squares method to optimize model parameters and complete the establishment of muscle strength estimation model: The unknown parameters x(1), x(2), and x(3) are obtained using the nonlinear least squares method, and the objective function is: The unknown parameters is the estimated force normalized by the maximum and minimum values, and F is the measured force normalized by the maximum and minimum values. After obtaining the unknown parameters, the target muscle force estimation model is obtained.

6. A muscle state analysis method based on electromyographic signal analysis of muscle strength according to claim 5, characterized in that: In S35, the nonlinear least squares method adjusts the model parameters through an iterative optimization algorithm to minimize the normalized error between the estimated force and the measured force; during the iteration process, the step size is dynamically adjusted based on parameter sensitivity to increase the convergence speed.

7. A muscle state analysis method based on electromyographic signal analysis of muscle strength according to claim 6, characterized in that: In S4, muscle activation is calculated by the following formula: The median frequency MF and the mean frequency MPF of the surface electromyographic signal are calculated as follows: Where D(f) is the power spectrum of the muscle surface electromyographic signal, and f represents the frequency of the surface electromyographic signal; Compare the frequency characteristics in the force state and the relaxation state to infer the degree of muscle fatigue. As muscle fatigue increases, the median frequency and average frequency will decrease, reflecting the decline in muscle performance after high-intensity activity. The degree of muscle fatigue is calculated by the following formula: Among them, MF max and MPF max Represents the median frequency and average frequency at maximum force, MF init and MPF init represent the median frequency and mean frequency during relaxation, respectively.

8. The muscle state analysis method based on electromyographic signal analysis of muscle strength according to claim 7, characterized in that: In S5, muscle activation and fatigue are combined to more comprehensively evaluate muscle status, and the combination of muscle activation and fatigue is integrated into a comprehensive evaluation index G: G=activation×(1-fatigue).

9. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

10. A terminal, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method according to any one of claims 1 to 8.

Images