Muscle relative load index generation method and multi-modal muscle state assessment method

Abstract

The present application relates to the technical field of motion monitoring, and particularly relates to a muscle relative load index generation method and a multi-modal muscle state evaluation method, comprising: synchronously collecting surface electromyography signals and inertial motion data; extracting muscle activation time series signals and motion intensity characteristic values; reading muscle activation baseline parameters and motion intensity reference values under a current motion scene; calculating relative activation through the muscle activation time series signals and the muscle activation baseline parameters; and calculating muscle relative load indexes through the motion intensity characteristic values, the motion intensity reference values, the relative activation, muscle group weight coefficients and motion intensity influence factors. The present application makes up for the lack of muscle physiological dimensions in single kinematic monitoring and the lack of motion intensity reference in single electromyography monitoring, can more comprehensively and accurately reflect the real muscle load state, and can continuously evaluate the muscle fatigue development process by comparing the change trend between the current electromyography spectrum characteristics and the individual reference state in real time.

Description

Technical Field

[0001] This invention relates to the field of motion monitoring technology, and in particular to a method for generating a muscle relative load index and a method for assessing multimodal muscle status. Background Technology

[0002] With the popularization of national fitness activities and the improvement of the level of precision in sports training, real-time and accurate assessment of muscle load during exercise is of great significance for improving training efficiency and reducing the risk of sports injuries. As a result, wearable muscle load monitoring technology has received widespread attention and application.

[0003] Existing muscle load monitoring technologies can be mainly divided into the following three categories: The first type is the kinematic monitoring scheme using pure inertial measurement units. This scheme collects motion time-series data by wearing accelerometers and gyroscopes on the limbs, and then calculates kinematic parameters such as gait, posture, and power to indirectly reflect the exercise load. However, this type of scheme can only obtain the mechanical performance of external limb movement and cannot collect electrophysiological signals at the muscle level, resulting in a fundamental deviation between the assessment results and the actual physiological load on the muscles.

[0004] The second category is physiological monitoring schemes based solely on surface electromyography (EMG). These schemes collect bioelectrical signals generated by muscle contraction using surface EMG electrodes attached to the skin, assessing muscle activation and fatigue based on signal amplitude and spectral characteristics. However, this type of scheme relies on the absolute amplitude of the EMG signal for assessment. Initial use requires users to complete a time-consuming maximum voluntary contraction calibration process, resulting in a high barrier to entry and a poor user experience. Furthermore, the EMG amplitude is affected by multiple interference factors such as skin impedance, electrode offset, subcutaneous fat thickness, and ambient temperature, making it difficult to guarantee the stability of absolute measurement accuracy. Additionally, the commonly used fixed thresholds for judging muscle state cannot adapt to the physiological differences of different users, nor can they be dynamically adjusted as the user's training level improves or their physiological state changes, leading to a continuous decline in assessment accuracy over long-term use.

[0005] The third category is a simple combination of multiple sensors, with some devices attempting to integrate both inertial measurement units and surface electromyography (SEMG) electrodes. However, this type of solution only performs simple temporal alignment of the two types of data, without constructing a unified quantitative evaluation index for deep fusion; the output content remains independent electromyographic parameters and kinematic parameters, failing to form a comprehensive load index that can simultaneously reflect the degree of muscle physiological activation and limb movement intensity. Summary of the Invention

[0006] This invention provides a method for generating a muscle relative load index and a multimodal muscle state assessment method, which can solve the problems pointed out in the background art.

[0007] To solve the above technical problems, the present invention provides the following technical solution: A method for generating a muscle relative load index, comprising: Simultaneously acquire surface electromyographic signals of the target muscle group, as well as inertial motion data of the target area; Based on the surface electromyography signal, the muscle activation time-series signal of the target muscle group is extracted, and based on the inertial motion data, the motion intensity feature value of the target part is extracted; Read the baseline parameters of muscle activation and the reference value of exercise intensity under the current exercise scenario and intensity range; The real-time relative activation level is calculated using the muscle activation time-series signal and the muscle activation baseline parameters. The muscle relative load index is calculated using the exercise intensity characteristic value, exercise intensity reference value, relative activation degree, as well as the preset muscle group weight coefficient and exercise intensity influencing factor.

[0008] Furthermore, the formula for calculating the muscle relative load index is as follows: ; In the formula, The relative load index of the muscle; The characteristic value of the motion intensity; This refers to the reference value for exercise intensity; The relative activation level; The weighting coefficient for the muscle group; This refers to the factor influencing the intensity of the exercise.

[0009] Furthermore, the surface electromyography signal is sequentially subjected to bandpass filtering, full-wave rectification, envelope extraction, and normalization to obtain the muscle activation time-series signal.

[0010] Furthermore, attitude calculation is performed on the inertial motion data to extract the motion intensity feature values.

[0011] Multimodal muscle status assessment methods include: Collect users' physiological attribute information, and based on the physiological attribute information, retrieve a matching set of group baseline parameters from a preset group baseline database as the current effective set of baseline parameters. The effective set of baseline parameters includes the muscle activation baseline parameters and exercise intensity reference values. The muscle relative load index is calculated by performing the muscle relative load index generation method described above; Valid training data is selected, and a personal baseline parameter set corresponding to the user, exercise scenario, intensity range, target muscle group, and target body part is iteratively constructed. When the personal baseline meets the preset convergence condition, the personal baseline parameter set replaces the group baseline parameter set as the current new effective baseline parameter set. When the personal baseline is determined to be invalid, pending update, or insufficient data, it reverts to the group baseline, or a weighted combination of the group baseline and the historical personal baseline is adopted. The relative muscle load index is compared with the deviation judgment threshold corresponding to the effective baseline parameter set; when the relative muscle load index exceeds the deviation judgment threshold and continues for a preset duration, it is determined that the muscle load amplitude has deviated.

[0012] Furthermore, the determination of muscle load amplitude deviation is performed only when the signal quality of the surface electromyography signal is higher than the set standard.

[0013] Furthermore, at least two acquisition channels are set up, and each acquisition channel independently acquires surface electromyography signals corresponding to different target muscle groups; For any of the acquisition channels, if the signal quality of the surface electromyography signal is lower than the set standard, the signal amplification gain is increased. If, after increasing the gain, the signal quality of the surface electromyography signal is still lower than the set standard, then the acquired data from the acquisition channel is discarded.

[0014] Furthermore, the validity conditions of the training data include: The signal quality of the surface electromyography signal is higher than the set standard; the training duration reaches the minimum duration of the exercise scenario, the exercise intensity is within the intensity range, the data loss duration is less than the preset duration, and the data value is within the normal range of historical distribution.

[0015] Furthermore, the preset convergence condition is: Under the same motion scenario and intensity range, more than a preset number of effective training data are accumulated, and the coefficient of variation of the baseline parameters in the most recent consecutive training sessions is lower than a preset threshold.

[0016] Furthermore, the personal baseline is determined to be in need of updating when at least one of the following conditions is detected: The user has not performed the exercise scenario training for more than a preset number of days, the target body part has changed, the surface electromyography signal acquisition device has changed, the parameters of the personal baseline have fluctuated beyond a preset threshold, or the user has actively chosen to re-establish the personal baseline.

[0017] The beneficial effects of this invention are as follows: The muscle relative load index generation method in this invention compensates for the lack of muscle physiological dimensions in single kinematic monitoring and the lack of exercise intensity reference in single electromyography monitoring by simultaneously collecting surface electromyography signals and inertial motion data. It calculates relative activation based on muscle activation baseline parameters, eliminating the dependence on absolute electromyography amplitude and maximum voluntary contraction calibration, and can adapt to the physiological differences of different users. At the same time, it deeply integrates relative activation and exercise intensity characteristics into a unified load quantification index, which can more comprehensively and accurately reflect the true muscle load state compared to the simple superposition of multi-sensor data.

[0018] This invention enables the continuous establishment of an individual user reference state during exercise. By comparing the changing trends between the current electromyographic spectrum characteristics and the individual reference state in real time, the muscle fatigue development process can be continuously assessed. Based on the degree of fatigue change, different levels of early warning information can be dynamically output, thereby realizing real-time fatigue monitoring and training risk alerts applicable to wearable sports scenarios. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A flowchart of the method for generating the muscle relative load index; Figure 2 This is a flowchart of a multimodal muscle state assessment method. Detailed Implementation

[0021] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0022] Example 1 like Figure 1 As shown, the method for generating the muscle relative load index includes: A1: Simultaneously acquire surface electromyographic signals of the target muscle group, as well as inertial motion data of the target area; In this step, the acquisition of surface electromyography signals is achieved by attaching electromyography electrodes to the skin surface of the target muscle group and synchronously acquiring the bioelectrical signals generated during muscle contraction at a preset sampling rate; the acquisition of inertial motion data is achieved by fixing an inertial measurement unit to the target part of the limb and synchronously acquiring the acceleration and angular velocity time-series data of that part during the movement. A2: Extracting the temporal signal of muscle activation of the target muscle group based on surface electromyography signals, and extracting the motion intensity feature value of the target part based on inertial motion data; A3: Read the muscle activation baseline parameters and exercise intensity reference values ​​under the current exercise scenario and intensity range; In this step, the muscle activation baseline parameter is a reference benchmark used by the system to compare and determine whether the muscle activation level deviates from the user's normal state in real time; while the exercise intensity reference value is a benchmark value used to perform dimensionless normalization on the real-time exercise intensity characteristics. In some embodiments of the present invention, during the initial use phase, the system can specifically load muscle activation baseline parameters of the corresponding group baseline from the group baseline library according to the user's gender, age, and training experience; as effective training data accumulates, the system can iteratively generate personal baselines maintained in layers according to exercise scenario, intensity range, target muscle group, and target body part through a sliding window statistical algorithm; after the personal baseline reaches the convergence condition, it can replace the group baseline as the basis for real-time comparison; when the personal baseline fails, it can revert to the group baseline, or adopt a weighted combination of the two to ensure the continuity of the evaluation process; In other embodiments of the present invention, the system can preset corresponding group reference values ​​as exercise intensity reference values ​​for different sports scenarios such as running, cycling, and strength training; or, it can generate individual reference values ​​as exercise intensity reference values ​​based on the user's personal historical data; the above two options depend on the selection of the above group baseline or individual baseline. By calculating the ratio of the real-time exercise intensity feature value to the reference value, the same units are eliminated, thus eliminating the dimensional differences between different scenarios and different feature forms.

[0023] A4: Real-time relative activation is calculated using the muscle activation time-series signal and muscle activation baseline parameters; A5: The muscle relative load index is calculated using exercise intensity characteristic values, exercise intensity reference values, relative activation, and preset muscle group weight coefficients and exercise intensity influencing factors.

[0024] The muscle relative load index generation method in this embodiment compensates for the lack of muscle physiological dimensions in single kinematic monitoring and the lack of exercise intensity reference in single electromyography monitoring by simultaneously collecting surface electromyography signals and inertial motion data. It calculates relative activation based on muscle activation baseline parameters, eliminating the dependence on absolute electromyography amplitude and maximum voluntary contraction calibration, and can adapt to the physiological differences of different users. At the same time, it deeply integrates relative activation and exercise intensity characteristics into a unified load quantification index, which can more comprehensively and accurately reflect the true muscle load state compared to the simple superposition of multi-sensor data.

[0025] This invention enables the continuous establishment of an individual user reference state during exercise. By comparing the changing trends between the current electromyographic spectrum characteristics and the individual reference state in real time, the muscle fatigue development process can be continuously assessed. Different levels of warning information can be dynamically output based on the degree of fatigue change, thereby achieving real-time fatigue monitoring and training risk alerts suitable for wearable sports scenarios. As a preferred embodiment, the formula for calculating the muscle relative load index is: ; In the formula, The muscle relative load index; The characteristic value of exercise intensity; This is a reference value for exercise intensity; Relative activation level; This refers to the muscle group weighting coefficient; specifically, it applies to the gastrocnemius muscle. A value of 1.0 is used for the quadriceps. A value of 1.2 is used for the gluteus maximus. The value is 1.1; the values ​​for other muscle groups are not listed one by one and can all be set by the system; the above values ​​are normalized reference values ​​and are only examples of embodiments, and are not intended to limit the scope of protection of this invention.

[0026] γ is an exercise intensity influencing factor used to quantify the nonlinear amplification effect of exercise intensity on muscle load. When exercise intensity exceeds the reference value, γ causes a superlinear increase in the muscle relative load index, reflecting the additional metabolic stress and mechanical injury risk experienced by muscles during high-intensity exercise. The value of γ ranges from 0.3 to 0.5, obtained based on fitting of a large amount of exercise experimental data, representing that for every 100% increase in exercise intensity beyond the reference value, the muscle relative load index increases by an additional 30%-50%. As a preferred embodiment of the above, the surface electromyography signal is sequentially subjected to bandpass filtering, full-wave rectification, envelope extraction and normalization to obtain the muscle activation time-series signal.

[0027] As an example, the sampling frequency of the surface electromyography signal can be 1000Hz.

[0028] In some embodiments of the present invention, the surface electromyography (EMG) signal is first subjected to a 20Hz-450Hz bandpass filter using a 4th-order Butterworth filter to remove power frequency interference, baseline drift, and motion artifacts. The filtered signal is then subjected to full-wave rectification, followed by low-pass filtering with a cutoff frequency of 3Hz-5Hz to extract the envelope signal. Finally, the envelope signal is normalized to obtain a time-series signal reflecting the real-time activation level of the muscle. Through these processing methods, the temporal variation characteristics of muscle activation are fully preserved, while eliminating individual differences in the absolute amplitude of the original EMG signal.

[0029] During implementation, the surface electromyography (EMG) signals of a single target muscle group are filtered, rectified, extracted, and normalized, resulting in a continuously changing one-dimensional numerical sequence over time. Each moment corresponds to the strength of muscle contraction activation at that time. This allows us to obtain the time-series signal of muscle activation, which is a dynamic real-time data stream.

[0030] For example, the gastrocnemius muscle channel is sampled at 1000Hz, and an activation value is output every 10ms. A time sequence is: [0.21, 0.35, 0.68, 0.42, 0.25...]. This entire continuous sequence is the muscle activation time sequence signal of the gastrocnemius muscle; correspondingly, the amplitude of the real-time muscle activation time sequence signal is 0.68.

[0031] In some embodiments of the present invention, the muscle activation baseline parameters are a set of static benchmark parameters stored hierarchically according to user, exercise scenario, intensity range and target muscle group, used to measure the user's normal exertion level, including sub-parameters such as benchmark amplitude and normal fluctuation range; divided into group baseline and converged individual baseline, with the individual baseline being the user's own normal standard.

[0032] For example, for a 32-year-old woman, the set of personal baseline parameters corresponding to the intensity range of running and the gastrocnemius muscle is {baseline amplitude: 0.45, lower limit of normal fluctuation: 0.30, upper limit of normal fluctuation: 0.60}. The entire set is collectively referred to as the muscle activation baseline parameters under this category.

[0033] Based on the above example, in step A4, the amplitude of the real-time muscle activation time-series signal of 0.68 is specifically compared with the baseline amplitude of 0.45 in the currently effective muscle activation baseline parameter to obtain the real-time relative activation of 1.51 with the user's own normal state as a reference. This is used to quantify the degree of deviation of the current muscle activation level from the individual baseline, representing that the current muscle activation intensity is 51% higher than the user's normal average level.

[0034] As a preferred embodiment of the above, attitude calculation is performed on the inertial motion data to extract the motion intensity feature value of the target part.

[0035] Specifically, inertial motion data is acquired using a three-axis accelerometer and a three-axis gyroscope; for example, the sampling frequency of the inertial motion data can be 200Hz.

[0036] After the data acquisition is completed, in some embodiments of the present invention, the Mahony complementary filtering algorithm is used to perform attitude calculation by fusing the three-axis accelerometer and the three-axis gyroscope. This can suppress the cumulative drift error caused by pure integral calculation, ensure the accuracy of the motion intensity characteristics of the installation part, and then extract specific feature values ​​of the installation part as motion intensity feature values.

[0037] In some embodiments of the present invention, the motion intensity characteristic value is at least one of the following characteristic values: angular velocity amplitude, acceleration vector and amplitude, cadence, and peak acceleration during the gait cycle obtained based on inertial motion data; the corresponding motion intensity characteristic value can be selected according to different sports scenarios. For example, cadence or angular velocity amplitude is preferred for running scenarios, angular velocity amplitude is preferred for cycling scenarios, and peak acceleration is preferred for strength training scenarios.

[0038] In the implementation process, if a non-angular velocity form of evaluation index is used, it is preferable to further map and convert the angular velocity amplitude into a dimensionless and equivalent motion intensity characteristic value to eliminate the dimensional differences between different motion scenarios and to quantify the intensity of limb movements. Example 2 This embodiment, based on Embodiment 1, provides a specific implementation scenario for generating the muscle relative load index. All parameter values ​​are example values ​​used in this embodiment and can be adjusted according to different sports, target muscle groups, training populations, and experimental calibration results. This does not constitute a limitation on the scope of protection of this invention. Taking the running scenario as an example, the example user is a 32-year-old woman with 1 year of running training experience, the target muscle group is the gastrocnemius muscle, and the target area is the outer side of the calf. B1: The bioelectric signals generated by muscle contraction are collected by electromyography electrodes attached to the skin surface of the gastrocnemius muscle at a sampling rate of 1000 Hz; and the acceleration and angular velocity time-series data of the target part of the lower leg during movement are collected synchronously by an integrated triaxial accelerometer and triaxial gyroscope at a sampling rate of 200 Hz. The two types of signals are synchronized based on a unified timestamp. B2: The surface electromyography (EMG) signals were sequentially subjected to a 4th-order Butterworth bandpass filter (20Hz-450Hz), full-wave rectification, and a low-pass filter with a cutoff frequency of 4Hz to extract the envelope signal. The envelope signal was then normalized to obtain a time-series signal reflecting the real-time muscle activation level. For inertial motion data, the Mahony complementary filtering algorithm was used to perform attitude calculations based on the fusion of triaxial accelerometer and triaxial gyroscope data. After suppressing integral drift, the real-time angular velocity amplitude of the lower leg was extracted as a motion intensity feature value. In this embodiment, taking jogging as an example, the characteristic value of exercise intensity is taken. =5.0 rad / s; In this embodiment, after the user has completed 5 effective running training sessions, the individual baseline for the corresponding dimension of the gastrocnemius muscle reaches the convergence condition, and the following steps are performed: B3: The system reads the current running intensity range, the personal baseline for the gastrocnemius muscle, and the corresponding personal muscle activation baseline parameters, while also reading the exercise intensity reference value corresponding to the running scenario. =5.0 rad / s, and the preset weighting coefficient of the gastrocnemius muscle group. =1.0, Influence factor of exercise intensity For example, =0.4; B4: The relative activation level is obtained by comparing the amplitude of the real-time muscle activation time-series signal with the baseline amplitude in the currently effective personal muscle activation baseline parameters. In this embodiment, relative activation is used. For example, a value of 1.26 corresponds to an approximately 26% increase in muscle activation compared to the individual's baseline. B5: Incorporate the characteristic value of exercise intensity Exercise intensity reference values and relative activation and the preset weight coefficient of the gastrocnemius muscle group Factors affecting exercise intensity Substitute the values ​​into the following formula to calculate the muscle relative load index. : =1.764.

[0039] Example 3 Multimodal muscle condition assessment methods, such as Figure 2 As shown, it includes: S1: Collect the user's physiological attribute information, and retrieve the matching set of group baseline parameters from the preset group baseline database based on the physiological attribute information as the current effective set of baseline parameters. The effective set of baseline parameters includes muscle activation baseline parameters and exercise intensity reference values. The physiological attribute information in this step, as described in Example 2 above, includes at least gender, age, and years of training experience. This step does not require a complex calibration process. The retrieved effective baseline parameter set serves as the initial parameter set, providing a benchmark for the subsequent calculation of the muscle relative load index. S2: Perform the muscle relative load index generation method as described in Example 1 to calculate the muscle relative load index; The implementation process of this step is as described in Example 1, and will not be repeated here; in step A3 of Example 1, the muscle activation baseline parameters are loaded from the group baseline or individual baseline. Based on this, this example also includes the following steps: S3: Filter valid training data and iteratively construct a personal baseline parameter set corresponding to the user, exercise scenario, intensity range, target muscle group, and target body part; when the personal baseline meets the preset convergence condition, replace the group baseline parameter set with the personal baseline parameter set as the current effective baseline parameter set; when the personal baseline is determined to be invalid, pending update, or insufficient data, revert to the group baseline, or use a weighted combination of the group baseline and historical personal baselines; This step involves the iterative construction and switching of individualized baselines, enabling long-term system adaptation. In some embodiments of this invention, the system first selects data that meets the validity criteria from the training data and includes it in baseline statistics. Specifically, it can iteratively construct a unique set of personal baseline parameters in layers according to the dimensions of user, exercise scenario, intensity range, target muscle group, and target body part. When the data accumulation and parameter stability of the personal baseline reach the preset convergence condition, the system switches the effective baseline from the general group baseline to the personalized personal baseline, realizing the upgrade of the evaluation benchmark from general to individual, and continuously improving the adaptability and accuracy of load assessment. In some embodiments of the present invention, the weighted combination of the group baseline and the historical individual baseline is achieved by the following formula: ; ; In the formula, Let be the effective baseline at time t. As a historical individual baseline, Let w be the population baseline and w be the adaptive weight. The initial value of the corresponding adaptive weight is λ > 0, and Δt is the number of days the baseline failed; as an example, The value can be 0.8, and λ can be 0.1. Of course, the above values ​​are not intended to limit the scope of protection of this invention. In this embodiment It can serve as a hybrid transitional baseline between individual and group baselines. As the individual baseline gradually becomes ineffective, the system gradually fades out of the individual baseline and fades into the group baseline. S4: Compare the muscle relative load index with the deviation judgment threshold corresponding to the effective baseline parameter set; when the muscle relative load index exceeds the deviation judgment threshold and continues for a preset duration, it is judged as a deviation of muscle load amplitude.

[0040] For example, if a user is using the system for the first time and there is no effective training data or personal baseline generated, the system will match the group baseline library based on gender, age, and training experience, and load the group baseline parameter set for that group as the current effective baseline: Effective baseline parameter set {muscle activation baseline parameter: {baseline amplitude 0.40}, exercise intensity reference value: 5.0 rad / s}; After accumulating 5 effective training sessions, the personal baseline converges, and the effective baseline is switched to the personal baseline parameter set. Under this group, the effective training sample size and the coefficient of variation of the baseline parameters both meet the convergence condition. The system replaces the effective baseline and switches to the exclusive personal baseline parameter set: Effective baseline parameter set {muscle activation baseline parameters: {baseline amplitude 0.45}, exercise intensity reference value: 5.0 rad / s}; In the example above, the baseline amplitude was updated from the general group value of 0.40 to the user's own normal value of 0.45, making the evaluation benchmark more in line with the individual's exertion characteristics.

[0041] The multimodal muscle state assessment method in this embodiment is initialized based on matching user physiological attributes with a group baseline, allowing for rapid activation without complex calibration. The muscle relative load index generated by multimodal fusion ensures the comprehensiveness of the load assessment. During implementation, a hierarchical iterative construction of individual baselines and adaptive switching of effective benchmarks continuously improves the personalized adaptability and accuracy of the assessment. A dual-judgment rule identifies deviations in muscle load amplitude, filtering out instantaneous normal fluctuations during exercise, reducing false alarms, and ensuring the reliability of deviation judgment.

[0042] As a preferred embodiment of the above, the determination of muscle load amplitude deviation is performed only when the signal quality of the surface electromyography signal is higher than the set standard.

[0043] In some embodiments of the present invention, the signal quality of surface electromyography signals is evaluated by at least one of the following indicators: signal-to-noise ratio, signal amplitude stability, waveform morphology consistency, and baseline drift. The set standard is the preset threshold corresponding to each of the above indicators. When two or more indicators are set, and all indicators meet the corresponding preset threshold, the signal quality is determined to be higher than the set standard.

[0044] In this preferred embodiment, by adding a precondition for surface electromyography signal quality verification before performing muscle load amplitude deviation judgment, invalid electromyography data with severe noise interference and acquisition failure can be filtered out, avoiding incorrect load deviation judgment based on distorted signals, and effectively reducing the probability of misjudgment of load abnormality.

[0045] As a preferred embodiment of the above, at least two acquisition channels are provided, and each acquisition channel independently acquires surface electromyography signals corresponding to different target muscle groups. For any acquisition channel, if the signal quality of the surface electromyography signal is lower than the set standard, the signal amplification gain will be increased. If the signal quality of the surface electromyography signal is still lower than the set standard after increasing the gain, the acquired data from the acquisition channel will be discarded.

[0046] This preferred solution is specifically designed for underwater environments. First, it compensates for weak electromyographic signals by increasing the amplification gain of a single poor-quality channel. If the compensation is ineffective, the data of the failed channel is directly discarded. This avoids the interruption of the overall assessment due to the failure of a single channel in the underwater environment, and can improve the stability and data integrity of muscle state assessment in underwater motion scenarios.

[0047] In the above embodiments, the setting standard can be a fixed threshold set by the system factory, or it can be adaptively configured according to the motion scenario and the environment in which the acquisition channel is used; and the gain amplification factor can specifically include a basic level, such as 200 times, and an underwater compensation high gain level, such as 400 times; when the channel signal quality is not up to standard, it switches from the basic level to the high gain level.

[0048] As a preferred embodiment of the above, the validity conditions for the training data include: The signal quality of the surface electromyography signal is higher than the set standard; the training duration reaches the minimum duration of the exercise scenario, the exercise intensity is within the intensity range, the data loss duration is less than the preset duration, and the data value is within the normal range of historical distribution.

[0049] Under the above conditions, if the signal quality of the surface electromyography (EMG) signal is higher than the set standard, it can filter out distorted EMG data caused by excessive noise and poor electrode contact, and only effective signals with a qualified signal-to-noise ratio are used for baseline iteration; by limiting the training time, it is ensured that there is a sufficient amount of data in a single training session, avoiding insufficient sampling information due to too short a duration; regarding the requirements for exercise intensity, extreme samples with excessively low or high intensity can be removed to prevent extreme exercise states from interfering with the accuracy of the normal baseline; if the data loss time is less than the preset time, the total duration of signal disconnection and transmission loss is limited to avoid distortion of sample features caused by a large amount of missing data; and when the requirement that the data value is within the normal range of the historical distribution is met, extreme abnormal feature values ​​caused by equipment failure and movement errors can be filtered out to ensure that the data used for baseline iteration closely matches the user's normal exercise performance.

[0050] The settings for standard, minimum duration, allowable range, preset duration, and normal range of historical distribution can be fixed values ​​preset by the system at the factory, or they can be adaptively adjusted according to different exercise scenarios and target muscle groups.

[0051] As a preferred embodiment of the above, the preset convergence condition is: Under the same exercise scenario and intensity range, accumulate more than a preset number of effective training data, and the coefficient of variation of the baseline parameters in the most recent consecutive training sessions is lower than the preset threshold.

[0052] For example, the preset number of times is 5, and the preset threshold is 10%. During convergence determination, the coefficient of variation is calculated for each baseline parameter. Only when the coefficients of variation of all core parameters are lower than the preset threshold is the individual baseline under that category considered to have reached convergence. The formula for calculating the coefficient of variation is parameter standard deviation / parameter mean. In this preferred embodiment, the number of the most recent consecutive times can be customized according to the baseline convergence stability requirements. Typically, this consecutive number is set to be equal to the preset number of data accumulations, but it can also be set separately to a fixed value less than the preset number.

[0053] The above classification refers to a uniquely matched combination unit of the same user, exercise scenario, intensity range, target muscle group and target position; the coefficient of variation of each baseline parameter is calculated separately for all valid training data accumulated within the combination unit.

[0054] During implementation, an individual's baseline will be deemed to require updating if at least one of the following conditions is detected: The following conditions are considered: the user has not performed exercise scenario training for more than a preset number of days; the target body part has changed; the surface electromyography signal acquisition device has changed; the fluctuation of personal baseline parameters exceeds a preset threshold; or the user actively chooses to re-establish their personal baseline. In some embodiments of the present invention, the determination of whether the fluctuation of personal baseline parameters exceeds the preset threshold can refer to the determination logic of calculating the coefficient of variation of baseline parameters based on the most recent consecutive training in the aforementioned convergence conditions. If the statistically obtained coefficient of variation exceeds the corresponding preset threshold, the baseline parameter fluctuation is determined to be excessive.

[0055] In this preferred solution, the baseline re-accumulation process is triggered in a timely manner for each of the above situations to avoid deviations in the baseline due to changes in usage conditions or physiological states, and to ensure that the baseline is adapted to the user's actual movement characteristics in the long term.

[0056] As a preferred embodiment of the above, the comparison between the muscle relative load index and the deviation judgment threshold adopts the following grading rule: When the muscle relative load index L(t) is less than 1.2, it is determined to be in the normal load range; When the muscle relative load index L(t) is greater than or equal to 1.2 and less than 1.5, it is judged to be slightly high; When the muscle relative load index L(t) is greater than or equal to 1.5 and less than 2.0, it is judged as moderately high and a warning message is output. When the muscle relative load index L(t) is greater than or equal to 2.0, it is judged as severely high and an alarm is output. The deviation judgment threshold is the above-mentioned graded boundary value, which can be adaptively adjusted according to the statistical distribution of the user's personal historical data.

[0057] Those skilled in the art should understand that this invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to this invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A multimodal muscle state assessment method, characterized in that, include: Collect users' physiological attribute information, and based on the physiological attribute information, retrieve a matching set of group baseline parameters from a preset group baseline database as the current effective set of baseline parameters. The effective set of baseline parameters includes the muscle activation baseline parameters and exercise intensity reference values. Simultaneously acquire surface electromyographic signals of the target muscle group, as well as inertial motion data of the target area; Based on the surface electromyography signal, the muscle activation time-series signal of the target muscle group is extracted, and based on the inertial motion data, the motion intensity feature value of the target part is extracted; Read the baseline parameters of muscle activation and the reference value of exercise intensity under the current exercise scenario and intensity range; The real-time relative activation level is calculated using the muscle activation time-series signal and the muscle activation baseline parameters. The muscle relative load index is calculated using the exercise intensity characteristic value, exercise intensity reference value, relative activation degree, and preset muscle group weight coefficient and exercise intensity influencing factor. The formula for calculating the muscle relative load index is as follows: ； In the formula, The relative load index of the muscle; The characteristic value of the motion intensity; This refers to the reference value for exercise intensity; The relative activation level; The weighting coefficient for the muscle group; The aforementioned factors affecting exercise intensity; Filter effective training data and iteratively construct a set of personal baseline parameters corresponding to the user, exercise scenario, intensity range, target muscle group and target body part; When the individual baseline meets the preset convergence condition, the individual baseline parameter set replaces the group baseline parameter set and becomes the new effective baseline parameter set. When the individual baseline is determined to be invalid, pending update, or insufficient data, it falls back to the group baseline, or a weighted combination of the group baseline and the historical individual baseline is used; The relative muscle load index is compared with the deviation judgment threshold corresponding to the effective baseline parameter set; when the relative muscle load index exceeds the deviation judgment threshold and continues for a preset duration, it is determined that the muscle load amplitude has deviated.

2. The multimodal muscle state assessment method according to claim 1, characterized in that, The surface electromyography signal is sequentially subjected to bandpass filtering, full-wave rectification, envelope extraction and normalization to obtain the muscle activation time-series signal.

3. The multimodal muscle state assessment method according to claim 1, characterized in that, The attitude calculation is performed on the inertial motion data, and the motion intensity feature value is extracted.

4. The multimodal muscle state assessment method according to claim 1, characterized in that, The determination of muscle load amplitude deviation is performed only when the signal quality of the surface electromyography signal is higher than the set standard.

5. The multimodal muscle state assessment method according to claim 1, characterized in that, At least two acquisition channels are set up, and each acquisition channel independently acquires surface electromyography signals corresponding to different target muscle groups; For any of the acquisition channels, if the signal quality of the surface electromyography signal is lower than the set standard, the signal amplification gain is increased. If, after increasing the gain, the signal quality of the surface electromyography signal is still lower than the set standard, then the acquired data from the acquisition channel is discarded.

6. The multimodal muscle state assessment method according to claim 1, characterized in that, The validity conditions of the training data include: The signal quality of the surface electromyography signal is higher than the set standard; the training duration reaches the minimum duration of the exercise scenario, the exercise intensity is within the intensity range, the data loss duration is less than the preset duration, and the data value is within the normal range of historical distribution.

7. The multimodal muscle state assessment method according to claim 1, characterized in that, The preset convergence condition is: Under the same motion scenario and intensity range, more than a preset number of effective training data are accumulated, and the coefficient of variation of the baseline parameters in the most recent consecutive training sessions is lower than a preset threshold.

8. The multimodal muscle state assessment method according to claim 1, characterized in that, The personal baseline is determined to be in need of updating when at least one of the following conditions is detected: The user has not performed the exercise scenario training for more than a preset number of days, the target body part has changed, the surface electromyography signal acquisition device has changed, the parameters of the personal baseline have fluctuated beyond a preset threshold, or the user has actively chosen to re-establish the personal baseline.

Images