Intelligent fatigue recognition method and device, electronic equipment and storage medium

Abstract

The invention provides an intelligent fatigue recognition method and device, electronic equipment and a storage medium, and the method comprises the steps: evaluating a fatigue index of a target user through two dimensions, i.e., a target conduction velocity and a target fractal dimension, the target conduction velocity being capable of reflecting a muscle fiber function, the target fractal dimension being capable of reflecting neural control complexity, and the target fractal dimension being capable of reflecting neural control complexity; one-sidedness of single index evaluation is avoided, and muscle fatigue and central fatigue can be distinguished more accurately; through multi-channel and multi-time-scale analysis, signal characteristics of different muscle areas and different contraction rhythms are covered, and errors of a single channel and a time scale are reduced. Furthermore, the peripheral fatigue index and the central fatigue index of the target user are calculated in combination with the baseline conduction velocity and the baseline fractal dimension obtained based on the steady state surface electromyogram signal of the target user, and the method can adapt to the muscle basic states of different users, so that the evaluation result is more personalized and higher in accuracy.

Description

Technical Field

[0001] This invention relates to the field of intelligent fatigue recognition technology, and in particular to an intelligent fatigue recognition method, device, electronic device, and storage medium. Background Technology

[0002] Muscle fatigue is a core pathophysiological mechanism underlying decreased athletic performance, increased risk of injury, and functional impairment. Its mechanisms can be broadly categorized into central fatigue and peripheral fatigue. Central fatigue is characterized by weakened neural drive, increased synchronization of motor units, and altered recruitment strategies, while peripheral fatigue is characterized by slowed muscle fiber conduction velocity, accumulation of metabolic waste, and decreased muscle membrane excitability. Accurately distinguishing between the two is of great value in assessing supercompensation in athletes, determining the physical characteristics of endurance and explosive power athletes, and diagnosing nerve damage. Therefore, accurately identifying central and peripheral fatigue has become a pressing problem in the field of fatigue identification technology. Summary of the Invention

[0003] In view of this, embodiments of the present invention provide an intelligent fatigue recognition method, device, electronic device, and storage medium to improve the accuracy of fatigue recognition.

[0004] According to one aspect of the present invention, an intelligent fatigue recognition method is provided, the method comprising: Multi-channel surface electromyography (EMG) signals of the target user are collected according to preset sampling time points to obtain the target EMG signal; the target EMG signal contains the sub-EMG signal sequence in each of the channels. Based on the target channel time delay between each channel in the target electromyographic signal, the target conduction velocity of the target electromyographic signal is calculated. For each channel in the target electromyography signal, the sub-electromyography signal sequence in the channel is resampled with different time steps and different starting points to obtain multiple sub-sequences corresponding to the channel; For each of the channels, the target curve length corresponding to each of the subsequences corresponding to each time step in the channel is calculated; Based on the target curve length corresponding to each different time step in each of the channels, the target fractal dimension of the target electromyographic signal is calculated. Based on the target conduction velocity, the target fractal dimension, and the target user's baseline data, the peripheral fatigue index and central fatigue index of the target user are calculated, wherein the baseline data includes the baseline conduction velocity and baseline fractal dimension calculated based on the surface electromyography signals of the target user in a steady state; Based on the comparison results between the peripheral fatigue index and the central fatigue index and the preset threshold range, the target fatigue type is determined.

[0005] In one possible embodiment, calculating the target conduction velocity of the target electromyographic signal based on the target channel delay between each of the channels in the target electromyographic signal includes: Based on the similarity of the sub-electromyography signals in each of the channels, the target channel delay between each adjacent channel is calculated; Based on the distance between each adjacent channel and the target channel delay between each adjacent channel, the local channel conduction velocity between adjacent channels is calculated. The target conduction velocity of the target electromyographic signal is obtained by weighted fusion of the conduction velocities of each local channel.

[0006] In one possible embodiment, calculating the target channel delay between adjacent channels based on the similarity of the sub-electromyography signals in each of the channels includes: The cross-correlation function of the adjacent channels is constructed based on the similarity of the sub-electromyography signals of adjacent channels and the channel delay between adjacent channels; The channel delay is set to different values ​​according to the time interval between the sampling time points, and the cross-correlation function value corresponding to each channel delay value is calculated; The maximum value of the cross-correlation function is determined based on the values ​​of each cross-correlation function; The channel delay corresponding to the maximum value of the cross-correlation function is used as the candidate channel delay between the adjacent channels; Based on the candidate channel delay, the cross-correlation function value corresponding to the candidate channel delay, and the cross-correlation function value corresponding to the adjacent sampling time points of the candidate channel delay, the candidate channel delay is corrected to obtain the target channel delay; The weighted fusion of the conduction velocities of each of the local channels to obtain the target conduction velocity of the target electromyographic signal includes: The local channel conduction velocities of each channel are weighted and fused according to the cross-correlation function values ​​between each channel and its adjacent channels to obtain the target conduction velocity of the target electromyographic signal.

[0007] In one possible embodiment, the step of calculating the target curve length corresponding to each time step in each channel, based on the curve length of each subsequence corresponding to each time step in the channel, includes: For each of the channels, the median of the curve lengths of each of the subsequences corresponding to each time step in the channel is taken as the target curve length corresponding to the time step in the channel.

[0008] In one possible embodiment, calculating the target fractal dimension of the target electromyographic signal based on the target curve length corresponding to each of the different time steps in each of the channels includes: For each of the channels, a weighted fitting is performed on each time step in the channel and the target curve length corresponding to each time step according to a preset scaling law relationship to obtain the subfractal dimension of the channel. The fitting weights used in the weighted fitting are the values ​​of the corresponding time steps. The target fractal dimension of the target electromyographic signal is obtained by weighted fusion of the sub-fractal dimensions in each channel.

[0009] In one possible embodiment, the method further includes: For each of the channels, the regression determination coefficients in the channel are calculated based on each time step in the channel, the target curve length corresponding to each time step, and the subfractal dimension. If the regression determination coefficient is higher than a preset coefficient threshold, the step of weighted fusion of the sub-fractal dimensions in each of the channels is performed to obtain the target fractal dimension of the target electromyographic signal. If the regression determination coefficient is lower than the coefficient threshold, the step of resampling the sub-electromyography signal sequence in each channel of the target electromyography signal with different time steps and different starting points is returned to obtain multiple sub-sequences corresponding to the channel. The weighted fusion of the sub-fractal dimensions in each of the aforementioned channels to obtain the target fractal dimension of the target electromyographic signal includes: The subfractal dimensions of each channel are weighted and fused according to the regression determination coefficients in each channel to obtain the target fractal dimension of the target electromyographic signal.

[0010] In one possible embodiment, calculating the regression determination coefficient for each of the channels, based on each time step in the channel, the target curve length corresponding to each time step, and the subfractal dimension, includes: Calculate the regression determination coefficient R in the channel according to the following formula. 2 :

[0011] Where k is the time step, w k Weights based on time step. Let k be the length of the target curve corresponding to time step k, and a and b be the fitting parameters, where -b = FD. i FD i Let i be the subfractal dimension in channel i. It is the weighted average of lnL(k). According to another aspect of the present invention, an intelligent fatigue recognition device is provided, the device comprising: The acquisition module is used to acquire multi-channel surface electromyography (EMG) signals of the target user according to preset sampling time points to obtain the target EMG signal; the target EMG signal includes sub-EMG signals in each of the channels; The first calculation module is used to calculate the target conduction velocity of the target electromyographic signal based on the target channel time delay between each channel in the target electromyographic signal; The resampling module is used to resample the sub-electromyography signal sequence in each channel of the target electromyography signal at different time steps and different starting points to obtain multiple sub-sequences corresponding to the channel. The second calculation module is used to calculate the target curve length corresponding to each time step in each channel based on the curve length of each subsequence corresponding to each time step in the channel. The third calculation module is used to calculate the target fractal dimension of the target electromyographic signal based on the target curve length corresponding to each different time step in each of the channels; The fourth calculation module is used to calculate the peripheral fatigue index and central fatigue index of the target user based on the target conduction velocity, the target fractal dimension and the baseline data of the target user, wherein the baseline data includes the baseline conduction velocity and the baseline fractal dimension calculated based on the surface electromyography signal of the target user in a steady state. The comparison module is used to determine the target fatigue type based on the comparison results between the peripheral fatigue index and the central fatigue index and the preset threshold range.

[0012] According to another aspect of the present invention, an electronic device is provided, comprising: Processor; and Stored program memory, The program includes instructions that, when executed by the processor, cause the processor to perform any of the intelligent fatigue recognition methods described above.

[0013] According to another aspect of the present invention, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause a computer to perform any of the intelligent fatigue recognition methods described above.

[0014] One or more technical solutions provided in the embodiments of the present invention collect target electromyography (EMG) signals of a target user, calculate target conduction velocity and target fractal dimension based on the target EMG signals, and calculate the peripheral fatigue index and central fatigue index of the target user based on the target conduction velocity, target fractal dimension and pre-collected baseline data of the target user. Then, based on the comparison results of the peripheral fatigue index and central fatigue index with preset threshold intervals, the target fatigue type of the target user is determined.

[0015] This invention evaluates a target user's fatigue index using two dimensions: target conduction velocity and target fractal dimension. Target conduction velocity reflects muscle fiber function, while target fractal dimension reflects the complexity of neural control, avoiding the limitations of single-indicator assessments and enabling a more accurate distinction between muscle fatigue and central fatigue. Multi-channel, multi-timescale analysis covers signal characteristics across different muscle regions and contraction rhythms, reducing errors associated with single channels and timescales. Furthermore, combining baseline data with the target user's peripheral and central fatigue indices allows for adaptation to different users' baseline muscle states, resulting in more personalized and accurate assessments. Attached Figure Description

[0016] Further details, features, and advantages of the invention are disclosed in the following description of exemplary embodiments in conjunction with the accompanying drawings, in which: Figure 1 A flowchart illustrating the intelligent fatigue recognition method provided by the present invention; Figure 2 This is a schematic diagram of a process for calculating the peripheral fatigue index and the central fatigue index in the intelligent fatigue recognition method provided by the present invention. Figure 3 A schematic diagram of a structure of the intelligent fatigue recognition device provided by the present invention; Figure 4 A structural block diagram of an exemplary electronic device that can be used to implement embodiments of the present invention is shown. Detailed Implementation

[0017] Embodiments of the present invention will now be described in more detail with reference to the accompanying drawings. While some embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the invention. It should be understood that the accompanying drawings and embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the invention.

[0018] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0019] The term "comprising" and its variations as used herein are open-ended, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Definitions of other terms will be given in the following description. It should be noted that the concepts of "first", "second", etc., mentioned in this invention are used only to distinguish different devices, modules, or units, and are not intended to limit the order of functions performed by these devices, modules, or units or their interdependencies.

[0020] It should be noted that the terms "a" and "a plurality of" used in this invention are illustrative rather than restrictive. Those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0021] The names of the messages or information exchanged between the multiple devices in the embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of these messages or information.

[0022] The following explanations of some terms used in this invention are provided to facilitate understanding by those skilled in the art: Surface electromyography (sEMG) refers to the comprehensive bioelectrical signals generated by the excitation of muscle fibers during the contraction or relaxation of skeletal muscle, which are collected by non-invasive electrodes attached to the skin surface.

[0023] Conductive velocity (CV) is a physical quantity that describes how fast bioelectrical signals propagate in tissues / fibers, and its unit is usually m / s (meters per second).

[0024] Fractal dimension (FD) is a quantitative indicator that describes the irregularity, complexity, and self-similarity of fractal structures. It is a core concept in fractal geometry, and its value directly reflects the complexity of the structure.

[0025] The Higuchi algorithm is used to calculate the fractal dimension. Its core idea is to resample the signal and calculate its curve length at different time scales (step size k). The length decreases linearly with the logarithm of k, and the negative value of the slope is the fractal dimension.

[0026] Scale law relationship: The scale law relationship in fractal theory refers to the fact that a certain characteristic quantity of a fractal structure or signal will exhibit a fixed power law relationship with the change of measurement scale. The core is that the characteristic quantity is proportional to the power function of the scale.

[0027] The channel activity index RMS, short for Root Mean Square, is the most commonly used time-domain characteristic index in electromyography (EMG) signal analysis. Its core purpose is to quantify the contraction intensity of muscles, i.e., channel activity.

[0028] Currently, the gold standard for distinguishing between central and peripheral fatigue still relies on invasive or external stimulation interventions, mainly including: 1. Twitch Interpolation Technique: Central fatigue requires external stimulation to induce. This method relies on invasive electrical stimulation, is complex to operate, and cannot be continuously monitored, making it unsuitable for routine training or clinical assessment.

[0029] 2. M-wave superimposition induced by electrical stimulation: This method distinguishes between central and peripheral contributions to endurance motor function. However, it requires external electrical stimulation, which can easily cause pain and muscle spasms, resulting in poor subject compliance. Therefore, it cannot be used for children, the elderly, or patients with nerve damage.

[0030] 3. Traditional electromyography (sEMG) time-frequency domain indices (such as root variance RMS and median frequency MPF) are combined with conduction velocity for fatigue identification. However, in this method, both RMS and MPF are affected by both peripheral metabolism and conduction velocity (CV), and cannot specifically separate the central mechanism. Furthermore, conduction velocity (CV) only reflects the peripheral and is not sensitive to central fatigue.

[0031] It is evident that none of the fatigue identification schemes in the relevant technologies can accurately identify central and peripheral fatigue in a non-invasive manner.

[0032] Based on this, the present invention provides an intelligent fatigue recognition method, device, electronic device, and storage medium. The intelligent fatigue recognition method provided by the present invention can be applied to any electronic device with intelligent fatigue recognition function, such as a server, computer, or mobile terminal, and the mobile terminal can be a smart wearable device. The present invention is described below with reference to the accompanying drawings: Figure 1 A flowchart illustrating the intelligent fatigue recognition method provided by the present invention may include the following steps: S101. Collect multi-channel surface electromyography (EMG) signals of the target user according to preset sampling time points to obtain target EMG signals; the target EMG signals contain sub-EMG signal sequences in each of the channels. S102. Calculate the target conduction velocity of the target electromyographic signal based on the target channel delay between each channel in the target electromyographic signal; S103. For each channel in the target electromyographic signal, the sub-electromyographic signal sequence in the channel is resampled with different time steps and different starting points to obtain multiple sub-sequences corresponding to the channel. S104. For each of the channels, calculate the target curve length corresponding to each of the subsequences corresponding to each time step in the channel; S105. Calculate the target fractal dimension of the target electromyographic signal based on the target curve length corresponding to each different time step in each channel. S106. Based on the target conduction velocity, the target fractal dimension, and the target user's baseline data, calculate the target user's peripheral fatigue index and central fatigue index, wherein the baseline data includes the baseline conduction velocity and baseline fractal dimension calculated based on the target user's surface electromyography signals under steady-state conditions. S107. Based on the comparison results between the peripheral fatigue index and the central fatigue index and the preset threshold range, determine the target fatigue type.

[0033] In this embodiment of the invention, target electromyography (EMG) signals of the target user are collected, target conduction velocity and target fractal dimension are calculated based on the target EMG signals, and peripheral fatigue index and central fatigue index of the target user are calculated based on the target conduction velocity, target fractal dimension and pre-collected baseline data of the target user. Then, based on the comparison results of peripheral fatigue index and central fatigue index with preset threshold intervals, the target fatigue type of the target user is determined.

[0034] This invention evaluates a target user's fatigue index using two dimensions: target conduction velocity and target fractal dimension. Target conduction velocity reflects muscle fiber function, while target fractal dimension reflects the complexity of neural control, avoiding the limitations of single-indicator assessments and enabling a more accurate distinction between muscle fatigue and central fatigue. Multi-channel, multi-timescale analysis covers signal characteristics across different muscle regions and contraction rhythms, reducing errors associated with single channels and timescales. Furthermore, combining baseline data with the target user's peripheral and central fatigue indices allows for adaptation to different users' baseline muscle states, resulting in more personalized and accurate assessments.

[0035] The following provides an exemplary description of S101-S106: The target user can be any user requiring fatigue recognition. In one possible embodiment, electrodes can be placed in a preset body region of the target user. This preset body region can be a region related to muscle contraction, such as upper limb muscles, lower limb muscles, trunk muscles, etc. As one possible implementation, a linear electrode array can be arranged according to the muscle direction. This linear electrode array can have 8 to 16 channels, that is, it can contain 8 to 16 electrodes, and multi-channel surface electromyography (EMG) signals of the target user can be collected at preset sampling time points. The spacing d between each electrode can be set according to the actual application scenario, such as 5 mm, 10 mm, etc. The number of sampling time points and the time interval between sampling time points can be set according to the actual application scenario. For example, the sampling rate can be set to ≥2 kHz, that is, at least 2000 signals are collected per second, and the interval between sampling time points is 500 μs. The collected raw EMG signals are multi-channel surface EMG signals. For ease of description, the EMG signals contained in a single channel are referred to as sub-EMG signal sequences in this invention.

[0036] The acquired raw electromyography (EMG) signals may contain noise and artifacts. Therefore, preprocessing of the raw EMG signals is necessary. This preprocessing process can include DC component removal, bandpass filtering, power frequency notch filtering, and artifact removal. DC component removal refers to removing stable DC offsets from the signal, which can be achieved by subtracting the average value of the raw EMG signal. Bandpass filtering retains only the effective frequency range of the raw EMG signal, filtering out irrelevant high and low frequency noise. The effective frequency of EMG signals is typically 10–500 Hz, so a bandpass filter can be used to filter signals with frequencies in this range. Power frequency notch filtering is used to filter AC interference from the power grid. The AC frequency in the power grid is typically 50 Hz, so a 50 Hz notch filter can be used to filter this interference. During EMG signal acquisition, sudden signal jumps may occur due to electrode displacement, large movements of the subject, or sudden signal interruptions due to electrode loosening, leading to a surge in noise. Artifact removal removes these obviously abnormal EMG signals, which can be achieved through methods such as threshold filtering and correlation analysis.

[0037] The raw electromyographic (EMG) signal is typically an EMG signal sequence over a relatively long period, such as a 30-minute EMG signal sequence for a target user. Since the target user's fatigue state changes over time, fatigue identification specifically refers to identifying the fatigue state of the target user at different times. Based on this, the raw EMG signal can be segmented according to a preset time window, resulting in multiple raw EMG signal windows. The length of this time window can be set according to the actual application scenario. Each raw EMG signal window contains raw EMG signals corresponding to multiple sampling time points. For example, with a sampling rate of 2kHz, the raw EMG signal sequence can be segmented into 1-second time windows, with each raw EMG signal window containing 2000 sampling time points. Furthermore, the target conduction velocity CV(T) and target fractal dimension FD(T) within each raw EMG signal window can be calculated. The specific value of the time T corresponding to the raw EMG signal window can be selected according to the actual application scenario, such as the start and end times of the time window. After preprocessing the original electromyographic signal window, the target electromyographic signal for subsequent analysis is obtained. The target conduction velocity and target fractal dimension of the target electromyographic signal can be calculated separately.

[0038] Target conduction velocity describes the speed at which an electrical signal propagates in the tissue of a target user and can be calculated using the distance between channels and the propagation time (i.e., delay) of the electrical signal between channels. In one possible embodiment, the target conduction velocity of the target electromyography (EMG) signal can be calculated based on the target channel delay between the channels in the target EMG signal through the following steps: S121. Based on the similarity of the sub-electromyography signals in each of the channels, calculate the target channel delay between each adjacent channel.

[0039] The time delay between adjacent channels can be calculated using the similarity between the electromyographic (EMG) signals of adjacent channels. Specifically, it can be calculated after how much time shift the similarity of the EMG signals of adjacent channels reaches its maximum value; the time shift corresponding to the maximum similarity is the target channel delay. For example, if the EMG signal at time t in channel A has the maximum similarity to the EMG signal at time t+τ in channel B, then τ can be determined as the signal transmission delay between channel A and channel B.

[0040] In one possible embodiment, the similarity of electromyography (EMG) signals between adjacent channels can be evaluated using a cross-correlation function, which describes the degree of similarity between two signals at different time delays. Since EMG signals are discrete signals acquired at discrete time points, the cross-correlation function between adjacent channels can be constructed by summing. For ease of description, the EMG signals in each channel are referred to as sub-EMG signals in this invention. The summation method described above refers to calculating the sum of the products of the sub-EMG signals at each time point of adjacent channels under a fixed time delay. As a possible implementation, to avoid the amplitude of the sub-EMG signals affecting the result of the cross-correlation function, the sum of the products can be normalized. Specifically, the sum of the products can be normalized by multiplying the sums of the sub-EMG signals at each sampling time point within the two channels. For example, the cross-correlation function can be expressed by the following formula: Formula (1) Where x is the electromyographic signal, i and i+1 are adjacent channels (i=1,2,…,N-1, N is the number of electrodes), and t is the current time point. Let τ be the sub-electromyography signal of channel i at time t, and τ be the time offset, i.e., the channel delay. Let be the sub-EMG signal of channel i+1 at time t+τ. This formula can be used to calculate the similarity between the two sub-EMG signals in channel i and channel i+1 after a shift in the sub-EMG signal on channel i over a given time period. Maximum. When When the maximum value is reached, it indicates that the two signals are most similar in shape and most aligned. The time offset τmax corresponding to the maximum value is the estimated candidate channel delay under the sampling accuracy.

[0041] As one possible implementation, τmax can be calculated by assigning values ​​to τ. As mentioned above, the sub-electromyography signal is a discrete signal acquired at different sampling time points. Therefore, the channel delay can be assigned values ​​based on the time interval between the sampling time points, and the maximum value of the cross-correlation function can be calculated. That is, the channel delay can be set to different values ​​according to the time interval between the sampling time points, and the cross-correlation function value corresponding to each channel delay value can be calculated; then, the maximum value of the cross-correlation function can be determined based on each cross-correlation function value. For example, τ can be assigned a multiple of the sampling time interval to obtain the candidate channel delay τ corresponding to the maximum value of the cross-correlation function. Max .

[0042] The candidate channel delay obtained through the above steps is a delay estimate under sampling precision. However, since the sampling time points are discrete, the actual peak value of the cross-correlation function may vary between sampling points. For example, the delay estimated through the above steps... That is, channel i+1 is delayed by three sampling time points compared to channel i, but the actual peak value of the cross-correlation function may occur at 2.4 or 3.6 sampling points. In one possible embodiment, in order to improve the accuracy of the target channel delay, the delay estimate under the sampling accuracy can be corrected to obtain a more accurate target channel delay.

[0043] As one possible implementation, the candidate channel delay can be corrected based on the candidate channel delay, the cross-correlation function value corresponding to the candidate channel delay, and the cross-correlation function value corresponding to the adjacent sampling time points of the candidate channel delay to obtain the target channel delay.

[0044] For example, the channel delay τmax estimated under the above sampling accuracy can be used as a candidate channel delay, and the target channel delay τ with higher accuracy can be obtained by using the three-point parabolic interpolation method on the candidate channel delay. peak The three-point parabolic interpolation method assumes that near the peak of the cross-correlation function curve, the shape of R(τ) approximates a parabola, and takes the peak point τ as the reference point. max And its two points τ to the left and right (max-1) and τ (max+1) To fit the vertex position of this parabola. Specifically, the target channel delay τ peak The following formula can be used for fitting and calculation: Formula (2) in, The cross-correlation function curve at the peak point τ max and its adjacent point τ (max-1) and τ (max+1) The numerator describes the degree of asymmetry of the peak, i.e., which side the peak tilts to; the denominator describes the curvature of the peak. The above formula (2) can be used to output the integer value. More precise peak position The unit is sampling points, which can be a decimal, for example, 2.37 sampling points.

[0045] S122. Calculate the local channel conduction velocity between adjacent channels based on the distance between each adjacent channel and the target channel delay between each adjacent channel. Local channel conduction velocity CV i Specifically, it can be calculated using the following formula: Formula (3) Where d is the electrode spacing. Let be the target channel delay in channel i.

[0046] S123. The conduction velocities of each local channel are weighted and fused to obtain the target conduction velocity of the target electromyographic signal.

[0047] Weight w of the conduction velocity of each local channel i The cross-correlation function of the corresponding channel can be used to determine the signal. As mentioned above, the cross-correlation function reflects the similarity of sub-EMG signals between adjacent channels. When the cross-correlation function value is high, the sub-EMG signal waveform is clear and reliable, and the corresponding weight w is high. i A higher value can be set; when the cross-correlation function value is low, the waveform clarity of the sub-EMG signal is poor, there may be noise, and the reliability is low. Therefore, the corresponding weight w should be adjusted accordingly. i It can be set to a smaller value.

[0048] For example, the weight w of the local channel conduction velocity of channel i can be calculated using the following formula. i : Formula (4) Accordingly, the target conduction velocity CV(T) can be obtained by weighted fusion of the conduction velocities of each local channel using the following formula: Formula (5) By employing the above technical solutions, and through multi-channel cross-correlation calculation and weighted fusion calculation of target transmission velocity, the accuracy of target transmission velocity estimation and noise resistance can be improved.

[0049] The target fractal dimension reflects the complexity of the structure. For electromyography (EMG) signals, a larger fractal dimension indicates a greater degree of muscle contraction, while the fractal dimension decreases during muscle fatigue. This invention utilizes an improved Higuchi algorithm to calculate the target fractal dimension. Specifically, for each channel of the target EMG signal, the sub-EMG signal sequences within that channel are resampled according to different time steps and starting points to obtain sub-sequences corresponding to each time step within each channel. The curve length of each sub-sequence is calculated, and based on the curve lengths of the sub-sequences corresponding to each time step within the channel, the target curve lengths corresponding to different time steps within that channel are determined. Then, according to the scaling law of fractal theory, different time steps within the channel and their corresponding target curve lengths are fitted to obtain the sub-fractal dimension of the channel. Finally, the sub-fractal dimensions of each channel are weighted and fused to obtain the target fractal dimension of the target EMG signal.

[0050] As one possible implementation, the target curve length corresponding to different time steps within each channel can be calculated using the following steps: S131. For each channel in the target electromyographic signal, the sub-electromyographic signal sequence in the channel is resampled with different time steps and different starting points to obtain multiple sub-sequences corresponding to the channel.

[0051] In this step, there can be multiple time steps and starting points. The time step can be set according to the sampling time points. For example, for a sub-EMG signal sequence with N sampling points, the time step can be set to k=1,2,…,k. max Typically, k is set. max = N / 10, where a larger k indicates a coarser time scale. Starting points m = 1.2…k are used to cover different starting points, completely splitting the sub-EMG signal sequence in the channel according to step size k, ensuring each starting position is covered, avoiding omissions, and guaranteeing the stationarity and shift invariance of scale statistics. For example, resampling a sub-EMG signal sequence containing N sampling time points yields the following sub-sequence: Formula (6) Where k is the time step, m is the starting point, and m = 1.2…k. The sub-electromyography sequence is obtained by resampling according to step size k and starting point m. This represents the electromyography (EMG) signal at the m-th sampling time point within the sub-EMG signal window. This represents the number of sampling time points contained in the subsequence obtained by the resampling. In each channel, multiple sub-electromyography sequences can be constructed using different scales k and different starting points m, which are used to estimate the curve length of the electromyography signal in the channel at different scales to achieve multi-scale self-similarity analysis.

[0052] S132. For each of the channels, calculate the target curve length corresponding to each of the subsequences corresponding to each time step in the channel.

[0053] The curve length of a subsequence can specifically be the normalized curve length of the subsequence. For example, for each subsequence obtained from resampling, the normalized curve length of the subsequence can be calculated using the following formula. : Formula (7) is the normalized curve length of the sub-electromyography sequence obtained by resampling at time step k and starting point m in the channel, and j is the time step factor. Let x(m+jk) and x(m+(j-1)k) be the (m+jk)th sub-EMG signal in the sub-EMG signal sequence of this channel. x(m+jk) and x(m+(j-1)k) are two adjacent sub-EMG signals in the sub-sequence obtained by resampling at time step k and starting point m. The length of this normalized curve reflects the rate of change or fluctuation complexity of the sub-sequence by representing the local change through the amplitude difference between adjacent points. (N-1) / N(m,k) is used for scale adjustment, making the lengths of different sub-EMG sequences comparable on the same N scale. 1 / k normalizes the length of scale k according to the index interval, thus obtaining the scaling relationship as k changes.

[0054] The traditional Higuchi algorithm uses the arithmetic mean of the subsequence curve lengths corresponding to each time step as the target subsequence curve length for that time step. However, since electromyographic signals are highly sensitive and easily affected by abnormal signals or noise, this embodiment of the invention uses a robust average instead of the arithmetic average to reduce the influence of abnormal signals or noise. Specifically, the normalized curve length L of each subsequence corresponding to time step k can be used... m (k) Take the median as the target curve length L(k) corresponding to the time step k to reduce the impact of individual outliers and thus improve the stability of the short window.

[0055] For each of the channels, a weighted fitting is performed on each time step in the channel and the target curve length corresponding to each time step according to a preset scaling law relationship to obtain the sub-fractal dimension of the channel. The fitting weight used in the weighted fitting is the value of the corresponding time step. The target fractal dimension of the target electromyographic signal can be obtained by weighted fusion of the sub-fractal dimensions in each channel.

[0056] The aforementioned pre-defined scaling law relationship can specifically be the scaling law relationship in fractal theory. The scaling law relationship in fractal theory refers to the fact that a certain characteristic quantity of a fractal structure or signal exhibits a fixed power-law relationship as the measurement scale changes. The core principle is that the characteristic quantity is directly proportional to a power function of the scale. Specifically, the scaling law relationship in fractal theory can be expressed by the following linear relationship: Formula (8) Where k is the time step, L(k) is the target curve length corresponding to time step k, and a and b are the fitting parameters to be determined, and the fractal dimension is... .

[0057] In this invention, the target curve length corresponding to different time steps in the channel can be fitted using least squares regression according to the aforementioned scaling law relationship, thus obtaining the subfractal dimension of the channel. Traditional Higuchi algorithms typically apply equal weights to all scales. As a possible implementation, to improve fitting stability and mitigate the impact of noise on regression results at small scales, weighted linear regression can be used for different scales k. Weighted linear regression is an extension of ordinary linear regression; its core is to assign differentiated weights to different sample points, allowing the model to prioritize fitting more reliable and important samples, thereby improving the accuracy and robustness of the regression results. In this invention, the fitting weights used in weighted linear regression can be set according to the actual application scenario. In one possible embodiment, the fitting weights can be set as the time step, i.e., the fitting weights. The slope b is estimated by weighted regression, and then the subfractal dimension FD in the channel is obtained. i .

[0058] A smaller k results in a larger sample size in each subsequence, but the interval between samples is extremely short, making L(k) more susceptible to amplification by random fluctuations from neighboring points and more affected by noise. A moderate k balances the sample size and sampling interval of the subsequence, allowing L(k) to better reflect the true structural changes and reducing its susceptibility to noise. Conversely, a larger k results in a smaller sample size in each subsequence, making L(k) more dependent on a few difference results and more susceptible to noise or random fluctuations. Therefore, by introducing a scale-dependent weight w'... k Small-scale L(k) noise is large but has small weights, which makes its residuals contribute little to the regression loss. When fitting, these noisy samples have low fitting priority, which can reduce the impact of noise on the regression results at small scales, thereby improving the stability of fractal dimension estimation under short time window conditions.

[0059] In practical applications, due to the presence of noise, the actual lnL(k) and k may not necessarily have a linear relationship. Therefore, the fitting difference can be calculated using the original lnL(k), the slope b (i.e. the negative of the subfractal dimension), and the fitting parameter a. If the difference is large, it means that there is a lot of noise, and the fitting result obtained cannot simulate the original signal well. Therefore, the subfractal dimension can be removed or its weight can be reduced when used later.

[0060] In one possible embodiment, to avoid outputting invalid fractal dimension values ​​when the scaling law does not hold or noise dominates, this embodiment can perform quality control on the fitting results. Specifically, for each channel, based on the length of the subsequence curve and the subfractal dimension corresponding to each time step in the channel, the regression determination coefficient in the channel can be calculated. The regression determination coefficient (R²) is a core indicator for measuring the fit of the regression model, representing the proportion of the variance in the dependent variable that the model can explain; the closer the value is to 1, the better the model fit. Specifically, based on the target curve length and fractal dimension corresponding to each time step in the channel, using each time step as a time step weight, the ratio between the weighted residual sum of squares and the weighted total deviation sum of squares corresponding to each time step in the channel can be calculated; the regression determination coefficient is calculated based on the ratio between the weighted residual sum of squares and the weighted total deviation sum of squares, wherein the regression determination coefficient is negatively correlated with the ratio between the weighted residual sum of squares and the weighted total deviation sum of squares.

[0061] For example, the coefficient of determination for regression can be calculated using the following formula, based on the weighted sum of squared residuals and the weighted sum of squared total deviations. : Formula (9) Where k is the time step, w k The time step weights can be set according to the actual application scenario; for example, they can be the same as the fitting weights mentioned above. This is the weighted average of lnL(k), where the weight w used in the weighted average calculation is the time step weight w. k .

[0062] If the regression determination coefficient is higher than a preset coefficient threshold, the step of weighted fusion of the sub-fractal dimensions in each of the channels is performed to obtain the target fractal dimension of the target electromyographic signal. If the regression determination coefficient is lower than the coefficient threshold, the step of resampling the sub-EMG signal sequences in each channel of the target EMG signal at different time steps and different starting points is returned to obtain the sub-sequences corresponding to the different time steps. For example, the coefficient threshold can be set to 0.90, then when... When, it represents a perfect fit. This indicates that the fit is meaningless. If If the current calculation result is marked as low quality, a suggestion will be made to resample or extend the window length; Then accept the result.

[0063] Then, subfractal dimensions with regression determination coefficients higher than a preset threshold can be weighted and fused. As one possible implementation, when weighting and fusing the subfractal dimensions of each channel, the channel activity index (RMS) can be used as the channel weight. In one possible embodiment, the regression determination coefficient R0 of the channel can also be used. 2 FD as the subfractal dimension in the channel i The weights are then used for weighted fusion. For example, the weighted fusion of the dimensions of each subfractal can be performed using the following formula: Formula (10) Among them, FD fusion R represents the target fractal dimension obtained after fusion. i 2 Let be the regression determination coefficient for channel i.

[0064] The traditional Higuchi algorithm is a single-channel algorithm and does not incorporate multi-channel computation. This invention addresses the issues of noise sensitivity, scale instability, and channel reliability differences in short-time-window, multi-channel electromyography signals by improving the traditional Higuchi algorithm from the perspectives of statistical estimation and multi-channel fusion. By introducing median robust statistics, scale-weighted regression, fitting instruction control, and multi-channel spatial weighted fusion mechanism, the stability and reliability of fractal dimension estimation in practical physiological signal applications are improved.

[0065] The decrease in the fractal dimension of electromyographic signals reflects central synchronization and is independent of changes in conduction velocity. When peripheral fatigue is dominant, conduction velocity decreases significantly and is the primary indicator, while the fractal dimension decreases slightly and is a secondary indicator. When central fatigue is dominant, the fractal dimension decreases significantly and is the primary indicator, while conduction velocity decreases slightly or remains stable and is a secondary indicator. Therefore, in this invention, the ratio of the decrease in conduction velocity to the decrease in fractal dimension can be used to distinguish fatigue types. As one possible implementation method, such as... Figure 2 As shown, the peripheral fatigue index and the central fatigue index can be calculated through the following steps: S151. Acquire the baseline electromyographic signal of the user in a steady state, and calculate the baseline conduction velocity and baseline fractal dimension of the baseline electromyographic signal.

[0066] The steady state refers to the target user at rest or during low-intensity muscle contraction. The acquisition method for baseline electromyography (EMG) signals, as well as the calculation process for baseline conduction velocity and baseline fractal dimension, can be found in the explanation above and will not be repeated here; only a brief explanation is provided. Based on the above calculation methods for target conduction velocity and target fractal dimension, baseline conduction velocity CV0 and baseline fractal dimension FD0 can be calculated from the baseline EMG signals.

[0067] S152. Calculate the rate of change of the target conduction velocity based on the target conduction velocity and the baseline conduction velocity; calculate the rate of change of the target fractal dimension based on the target fractal dimension and the baseline fractal dimension.

[0068] The rate of change of the target's conduction velocity and the rate of change of the target's fractal dimension can both be calculated in real time, such as by calculating the relative change rate of a real-time window. For example, the rate of change of the target's conduction velocity ΔCV(T) and the rate of change of the target's fractal dimension ΔFD(T) at time T can be calculated using the following formulas: Formula (11) Formula (12) in, Let T be the target propagation velocity. Let T be the target fractal dimension. Baseline conduction velocity, Given the baseline fractal dimension, at fatigue: .

[0069] S153. Calculate the peripheral fatigue index and the central fatigue index based on the target conduction velocity change rate and the target fractal dimension change rate according to the preset peripheral fatigue index formula and central fatigue index formula. In the peripheral fatigue index formula, the weight of the target conduction velocity change rate is higher than that of the target fractal dimension; in the central fatigue index formula, the weight of the target fractal dimension change rate is higher than that of the target conduction velocity change rate.

[0070] Peripheral fatigue primarily affects conduction velocity. Therefore, the peripheral fatigue index PFI(T) can be calculated by prioritizing the rate of change of the target conduction velocity and considering the rate of change of the target fractal dimension as a secondary factor. For example, the peripheral fatigue index can be calculated using the following formula: Formula (13) The decrease in conduction velocity (CV) is the primary factor in determining peripheral fatigue, with a weight of 1 (dominant). The decrease in the target fractal dimension (FD) serves as an auxiliary factor, and therefore its weight β can be set to a value within (0,1). For example, this weight can be set to 0.4. The negative sign makes the peripheral fatigue index (PFI) positive during fatigue, and thus, the larger the PFI value, the more severe the peripheral fatigue.

[0071] Central fatigue primarily affects fractal dimension. Therefore, the central fatigue index CFI(T) can be calculated by prioritizing the rate of change of the target fractal dimension and considering the rate of change of the target conduction velocity as a secondary factor. For example, the central fatigue index can be calculated using the following formula: Formula (13) A decrease in fractal dimension (FD) is a major indicator of central fatigue, with a weight of 1 (dominant). A decrease in conduction velocity (CV) serves as a cross-penalty, with the weight γ ranging from (0,1). For example, γ can be set to 0.2. A significant decrease in CV suggests peripheral dominance, thus reducing central confidence. The negative sign makes the fractal dimension index (CFI) positive during fatigue; a larger CFI value indicates more severe central fatigue.

[0072] In this invention, fatigue type can be determined by jointly using the peripheral fatigue index and the central fatigue index. In one possible embodiment, the target fatigue type can be determined by calculating the ratio between the central fatigue index and the peripheral fatigue index, and by comparing the central fatigue index, the peripheral fatigue index, and the ratio with a preset threshold range. For example, the ratio B(T) of central and peripheral fatigue can be calculated using the following formula: Formula (14) The threshold curves mentioned above can be set separately for the central fatigue index, the peripheral fatigue index, and the index ratio. For example, they can be set as follows: No obvious fatigue, The primary cause is peripheral fatigue; The main cause is central fatigue; It is a mixed fatigue.

[0073] While identifying the target fatigue type, threshold warnings can be issued to the target user, such as indicating the current fatigue type. In one possible embodiment, the fatigue index and fatigue identification results can be visualized, such as generating a real-time fatigue map, which can include the changes in peripheral fatigue index, central fatigue index, and fatigue identification calculation results over time.

[0074] This invention improves the accuracy and noise resistance of conduction velocity (CV) estimation through multi-channel cross-correlation calculation and weighted fusion; achieves short-window stable calculation of fractal dimension (FD) through multi-channel spatial fusion and robust averaging; and defines a two-dimensional fatigue index, fatigue type discrimination rules, and adaptive thresholds to achieve quantitative separation of fatigue types. It enables quantitative separation and dynamic tracking of central and peripheral fatigue, significantly improving accuracy through the complementary use of dual indicators, thus enhancing the scientific rigor and individualized precision of exercise training and rehabilitation assessment. Short-window stable calculation and multi-channel spatial fusion enhance real-time performance and robustness, allowing for scalability to wearable systems and overcoming the limitations of traditional single conduction velocity (CV) or fractal dimension (FD) indicators in comprehensively reflecting neuromuscular fatigue characteristics.

[0075] This invention can be widely applied in multiple fields such as sports science, rehabilitation medicine, and wearable health monitoring, providing a new two-dimensional signal feature model for the study of neuromuscular coupling mechanisms. It can also form a standardized central / peripheral fatigue index system as an evaluation standard for the future sports medicine and rehabilitation industry. For example, the intelligent recognition method provided by this invention can be applied to the following aspects: 1. Sports training and athlete monitoring: Combining wireless multi-conductive electromyography (EMG) patches or sportswear, it enables real-time separate monitoring of central and peripheral fatigue during training. It supports exercise load optimization, individualized training plan adjustments, and early warning of overtraining. When integrated with an AI-powered fitness management platform, it can generate individualized neuromuscular fatigue maps for competitive performance analysis.

[0076] 2. Neurological rehabilitation and disease diagnosis: Applicable to neuromuscular function assessment in patients with stroke, spinal cord injury, and Parkinson's disease. It differentiates between insufficient neural drive (central fatigue) and muscle metabolic disorders (peripheral fatigue) using CFI and PFI indices, guiding rehabilitation treatment plans. It can be extended to adaptive feedback parameters for myoelectric prostheses and brain-computer interface (BCI) systems.

[0077] 3. Wearable Health and Remote Monitoring System: Integrated into flexible wearable sEMG patches or rehabilitation monitoring garments, enabling continuous, non-invasive muscle fatigue detection. Suitable for fall risk assessment in the elderly, tracking of physical recovery, and monitoring of occupational fatigue (such as surgeons, drivers, assembly workers, etc.).

[0078] Based on the same inventive concept, according to another aspect of the present invention, an intelligent fatigue recognition device is provided, such as... Figure 3 As shown, the device 300 may include: Acquisition module 301 is used to acquire multi-channel surface electromyography (EMG) signals of a target user according to preset sampling time points to obtain target EMG signals; the target EMG signals include sub-EMG signals in each of the channels; The first calculation module 302 is used to calculate the target conduction velocity of the target electromyographic signal based on the target channel time delay between each channel in the target electromyographic signal; The resampling module 303 is used to resample the sub-electromyography signal sequence in each channel of the target electromyography signal with different time steps and different starting points to obtain multiple sub-sequences corresponding to the channel. The second calculation module 304 is used to calculate the target curve length corresponding to each time step in each channel based on the curve length of each subsequence corresponding to each time step in each channel. The third calculation module 305 is used to calculate the target fractal dimension of the target electromyographic signal based on the target curve length corresponding to each different time step in each of the channels. The fourth calculation module 306 is used to calculate the peripheral fatigue index and central fatigue index of the target user based on the target conduction velocity, the target fractal dimension and the baseline data of the target user, wherein the baseline data includes the baseline conduction velocity and the baseline fractal dimension calculated based on the surface electromyography signal of the target user in a steady state. The comparison module 307 is used to determine the target fatigue type based on the comparison results between the peripheral fatigue index and the central fatigue index and the preset threshold range.

[0079] In one possible embodiment, calculating the target conduction velocity of the target electromyographic signal based on the target channel delay between each of the channels in the target electromyographic signal includes: Based on the similarity of the sub-electromyography signals in each of the channels, the target channel delay between each adjacent channel is calculated; Based on the distance between each adjacent channel and the target channel delay between each adjacent channel, the local channel conduction velocity between adjacent channels is calculated. The target conduction velocity of the target electromyographic signal is obtained by weighted fusion of the conduction velocities of each local channel.

[0080] In one possible embodiment, calculating the target channel delay between adjacent channels based on the similarity of the sub-electromyography signals in each of the channels includes: The cross-correlation function of the adjacent channels is constructed based on the similarity of the sub-electromyography signals of adjacent channels and the channel delay between adjacent channels; The channel delay is set to different values ​​according to the time interval between the sampling time points, and the cross-correlation function value corresponding to each channel delay value is calculated; The maximum value of the cross-correlation function is determined based on the values ​​of each cross-correlation function; The channel delay corresponding to the maximum value of the cross-correlation function is used as the candidate channel delay between the adjacent channels; Based on the candidate channel delay, the cross-correlation function value corresponding to the candidate channel delay, and the cross-correlation function value corresponding to the adjacent sampling time points of the candidate channel delay, the candidate channel delay is corrected to obtain the target channel delay; The weighted fusion of the conduction velocities of each of the local channels to obtain the target conduction velocity of the target electromyographic signal includes: The local channel conduction velocities of each channel are weighted and fused according to the cross-correlation function values ​​between each channel and its adjacent channels to obtain the target conduction velocity of the target electromyographic signal.

[0081] In one possible embodiment, the step of calculating the target curve length corresponding to each time step in each channel, based on the curve length of each subsequence corresponding to each time step in the channel, includes: For each of the channels, the median of the curve lengths of each of the subsequences corresponding to each time step in the channel is taken as the target curve length corresponding to the time step in the channel.

[0082] In one possible embodiment, calculating the target fractal dimension of the target electromyographic signal based on the target curve length corresponding to each of the different time steps in each of the channels includes: For each of the channels, a weighted fitting is performed on each time step in the channel and the target curve length corresponding to each time step according to a preset scaling law relationship to obtain the subfractal dimension of the channel. The fitting weights used in the weighted fitting are the values ​​of the corresponding time steps. The target fractal dimension of the target electromyographic signal is obtained by weighted fusion of the sub-fractal dimensions in each channel.

[0083] In one possible embodiment, the method further includes: For each of the channels, the regression determination coefficients in the channel are calculated based on each time step in the channel, the target curve length corresponding to each time step, and the subfractal dimension. If the regression determination coefficient is higher than a preset coefficient threshold, the step of weighted fusion of the sub-fractal dimensions in each of the channels is performed to obtain the target fractal dimension of the target electromyographic signal. If the regression determination coefficient is lower than the coefficient threshold, the step of resampling the sub-electromyography signal sequence in each channel of the target electromyography signal with different time steps and different starting points is returned to obtain multiple sub-sequences corresponding to the channel. The weighted fusion of the sub-fractal dimensions in each of the aforementioned channels to obtain the target fractal dimension of the target electromyographic signal includes: The subfractal dimensions of each channel are weighted and fused according to the regression determination coefficients in each channel to obtain the target fractal dimension of the target electromyographic signal.

[0084] In one possible embodiment, calculating the regression determination coefficient for each of the channels, based on each time step in the channel, the target curve length corresponding to each time step, and the subfractal dimension, includes: Calculate the regression determination coefficient R in the channel according to the following formula. 2 :

[0085] Where k is the time step, w k Weights based on time step. Let k be the length of the target curve corresponding to time step k, and a and b be the fitting parameters, where -b = FD. i FD i Let i be the subfractal dimension in channel i. It is the weighted average of lnL(k).

[0086] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in this invention comply with relevant laws and regulations and do not violate public order and good morals.

[0087] An exemplary embodiment of the present invention also provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor. The memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to cause the electronic device to perform a method according to an embodiment of the present invention.

[0088] An exemplary embodiment of the present invention also provides a non-transitory computer-readable storage medium storing a computer program, wherein the computer program, when executed by a computer's processor, is used to cause the computer to perform a method according to an embodiment of the present invention.

[0089] An exemplary embodiment of the present invention also provides a computer program product, including a computer program, wherein, when executed by a computer's processor, the computer program is used to cause the computer to perform a method according to an embodiment of the present invention.

[0090] refer to Figure 4 The present invention will now be described in the form of a structural block diagram of an electronic device 400 that can serve as a server or client of the present invention, which is an example of a hardware device that can be applied to various aspects of the present invention. The electronic device is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0091] like Figure 4As shown, the electronic device 400 includes a computing unit 401, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 402 or a computer program loaded from a storage unit 408 into a random access memory (RAM) 403. The RAM 403 may also store various programs and data required for the operation of the electronic device 400. The computing unit 401, ROM 402, and RAM 403 are interconnected via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0092] Multiple components in electronic device 400 are connected to I / O interface 405, including: input unit 406, output unit 407, storage unit 408, and communication unit 409. Input unit 406 can be any type of device capable of inputting information to electronic device 400. Input unit 406 can receive input digital or character information and generate key signal inputs related to user settings and / or function control of electronic device. Output unit 407 can be any type of device capable of presenting information and may include, but is not limited to, a display, speaker, video / audio output terminal, vibrator, and / or printer. Storage unit 408 may include, but is not limited to, disks and optical discs. Communication unit 409 allows electronic device 400 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks, and may include, but is not limited to, modems, network cards, infrared communication devices, wireless communication transceivers, and / or chipsets, such as Bluetooth™ devices, WiFi devices, WiMax devices, cellular communication devices, and / or the like.

[0093] The computing unit 401 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 401 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 401 performs the various methods and processes described above. For example, in some embodiments, any of the intelligent fatigue recognition methods described above can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 408. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 400 via ROM 402 and / or communication unit 409. In some embodiments, the computing unit 401 can be configured to perform any of the intelligent fatigue recognition methods described above by any other suitable means (e.g., by means of firmware).

[0094] The program code used to implement the methods of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0095] In the context of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0096] As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, device, and / or apparatus (e.g., disk, optical disk, memory, programmable logic device (PLD)) for providing machine instructions and / or data to a programmable processor, including machine-readable media that receive machine instructions as machine-readable signals. The term "machine-readable signal" refers to any signal for providing machine instructions and / or data to a programmable processor.

[0097] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0098] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0099] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other.

Claims

1. A method for intelligent fatigue recognition, characterized in that, The method includes: Multi-channel surface electromyography (EMG) signals of the target user are collected according to preset sampling time points to obtain the target EMG signal; the target EMG signal contains the sub-EMG signal sequence in each of the channels. Based on the target channel time delay between each channel in the target electromyographic signal, the target conduction velocity of the target electromyographic signal is calculated. For each channel in the target electromyography signal, the sub-electromyography signal sequence in the channel is resampled with different time steps and different starting points to obtain multiple sub-sequences corresponding to the channel; For each of the channels, the target curve length corresponding to each of the subsequences corresponding to each time step in the channel is calculated; Based on the target curve length corresponding to each different time step in each of the channels, the target fractal dimension of the target electromyographic signal is calculated. Based on the target conduction velocity, the target fractal dimension, and the target user's baseline data, the peripheral fatigue index and central fatigue index of the target user are calculated, wherein the baseline data includes the baseline conduction velocity and baseline fractal dimension calculated based on the surface electromyography signals of the target user in a steady state; Based on the comparison results between the peripheral fatigue index and the central fatigue index and the preset threshold range, the target fatigue type is determined.

2. The method according to claim 1, characterized in that, The calculation of the target conduction velocity of the target electromyographic signal based on the target channel time delay between each channel in the target electromyographic signal includes: Based on the similarity of the sub-electromyography signals in each of the channels, the target channel delay between each adjacent channel is calculated; Based on the distance between each adjacent channel and the target channel delay between each adjacent channel, the local channel conduction velocity between adjacent channels is calculated. The target conduction velocity of the target electromyographic signal is obtained by weighted fusion of the conduction velocities of each local channel.

3. The method according to claim 2, characterized in that, The calculation of the target channel delay between adjacent channels based on the similarity of the sub-electromyography signals in each of the channels includes: The cross-correlation function of the adjacent channels is constructed based on the similarity of the sub-electromyography signals of adjacent channels and the channel delay between adjacent channels; The channel delay is set to different values ​​according to the time interval between the sampling time points, and the cross-correlation function value corresponding to each channel delay value is calculated; The maximum value of the cross-correlation function is determined based on the values ​​of each cross-correlation function; The channel delay corresponding to the maximum value of the cross-correlation function is used as the candidate channel delay between the adjacent channels; Based on the candidate channel delay, the cross-correlation function value corresponding to the candidate channel delay, and the cross-correlation function value corresponding to the adjacent sampling time points of the candidate channel delay, the candidate channel delay is corrected to obtain the target channel delay; The weighted fusion of the conduction velocities of each of the local channels to obtain the target conduction velocity of the target electromyographic signal includes: The local channel conduction velocities of each channel are weighted and fused according to the cross-correlation function values ​​between each channel and its adjacent channels to obtain the target conduction velocity of the target electromyographic signal.

4. The method according to claim 1, characterized in that, The step of calculating the target curve length corresponding to each time step in each channel, based on the curve length of each subsequence corresponding to each time step in the channel, includes: For each of the channels, the median of the curve lengths of each of the subsequences corresponding to each time step in the channel is taken as the target curve length corresponding to the time step in the channel.

5. The method according to claim 4, characterized in that, The calculation of the target fractal dimension of the target electromyographic signal based on the target curve length corresponding to each of the different time steps in each of the channels includes: For each of the channels, a weighted fitting is performed on each time step in the channel and the target curve length corresponding to each time step according to a preset scaling law relationship to obtain the subfractal dimension of the channel. The fitting weights used in the weighted fitting are the values ​​of the corresponding time steps. The target fractal dimension of the target electromyographic signal is obtained by weighted fusion of the sub-fractal dimensions in each channel.

6. The method according to claim 5, characterized in that, The method further includes: For each of the channels, the regression determination coefficients in the channel are calculated based on each time step in the channel, the target curve length corresponding to each time step, and the subfractal dimension. If the regression determination coefficient is higher than a preset coefficient threshold, the step of weighted fusion of the sub-fractal dimensions in each of the channels is performed to obtain the target fractal dimension of the target electromyographic signal. If the regression determination coefficient is lower than the coefficient threshold, the step of resampling the sub-electromyography signal sequence in each channel of the target electromyography signal with different time steps and different starting points is returned to obtain multiple sub-sequences corresponding to the channel. The weighted fusion of the sub-fractal dimensions in each of the aforementioned channels to obtain the target fractal dimension of the target electromyographic signal includes: The subfractal dimensions of each channel are weighted and fused according to the regression determination coefficients in each channel to obtain the target fractal dimension of the target electromyographic signal.

7. The method according to claim 6, characterized in that, For each of the aforementioned channels, based on each time step in the channel, the target curve length corresponding to each time step, and the subfractal dimension, the regression determination coefficient in the channel is calculated, including: Calculate the regression determination coefficient R in the channel according to the following formula. 2 : Where k is the time step, w k Weights based on time step. Let k be the length of the target curve corresponding to time step k, and a and b be the fitting parameters, where -b = FD. i FD i Let i be the subfractal dimension in channel i. It is the weighted average of lnL(k).

8. The method according to claim 1, characterized in that, The calculation of the peripheral fatigue index and central fatigue index of the target user based on the target conduction velocity, the target fractal dimension, and the baseline data of the target user includes: The baseline electromyography (EMG) signal of the user in a steady state was acquired, and the baseline conduction velocity and baseline fractal dimension of the baseline EMG signal were calculated. Calculate the rate of change of the target conduction velocity based on the target conduction velocity and the baseline conduction velocity; Based on the target fractal dimension and the baseline fractal dimension, calculate the rate of change of the target fractal dimension; According to the preset peripheral fatigue index formula and central fatigue index formula, the peripheral fatigue index and central fatigue index are calculated based on the target conduction velocity change rate and the target fractal dimension change rate. In the peripheral fatigue index formula, the weight of the target conduction velocity change rate is higher than that of the target fractal dimension; in the central fatigue index formula, the weight of the target fractal dimension change rate is higher than that of the target conduction velocity change rate.

9. An intelligent fatigue recognition device, characterized in that, The device includes: The acquisition module is used to acquire multi-channel surface electromyography (EMG) signals of the target user according to preset sampling time points to obtain the target EMG signal; the target EMG signal includes sub-EMG signals in each of the channels; The first calculation module is used to calculate the target conduction velocity of the target electromyographic signal based on the target channel time delay between each channel in the target electromyographic signal; The resampling module is used to resample the sub-electromyography signal sequence in each channel of the target electromyography signal at different time steps and different starting points to obtain the sub-sequences corresponding to the different time steps. The second calculation module is used to calculate the target curve length corresponding to each time step in each channel based on the curve length of each subsequence corresponding to each time step in each channel. The third calculation module is used to calculate the target fractal dimension of the target electromyographic signal based on the target curve length corresponding to each different time step in each of the channels; The fourth calculation module is used to calculate the peripheral fatigue index and central fatigue index of the target user based on the target conduction velocity, the target fractal dimension and the baseline data of the target user, wherein the baseline data includes the baseline conduction velocity and the baseline fractal dimension calculated based on the surface electromyography signal of the target user in a steady state. The comparison module is used to determine the target fatigue type based on the comparison results between the peripheral fatigue index and the central fatigue index and the preset threshold range.

10. An electronic device, comprising: processor; as well as Stored program memory, The program includes instructions that, when executed by the processor, cause the processor to perform the method according to any one of claims 1-8.

Images