Intelligent motion evaluation and early warning system based on multi-source biosignal fusion

The intelligent exercise assessment and early warning system, which integrates multi-source biosignals, addresses the shortcomings of single-signal assessment in existing systems. It enables precise assessment of muscle fatigue and cardiovascular load, improves the accuracy and environmental adaptability of exercise risk, and provides personalized exercise risk warnings.

CN122320525APending Publication Date: 2026-07-03深圳蜜拓蜜健康管理有限公司
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202610288200.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-03-10
Publication Date
2026-07-03

AI Technical Summary

Technical Problem

Existing exercise assessment and early warning systems rely on a single biological signal and do not fully consider the synergistic effect of multiple physiological indicators such as muscle fatigue and sympathetic nerve activity. They are difficult to achieve accurate exercise risk warnings in complex environments and are easily affected by sweat secretion factors, leading to misjudgment of assessment results.

Method used

An intelligent exercise assessment and early warning system employing multi-source biosignal fusion collects exercise data through devices such as surface electromyography (EMG) sensors, heart rate sensors, blood oxygen sensors, skin conductance sensors, and acceleration sensors. It combines data related to muscle fatigue, cardiovascular load, and risk warning, and uses muscle fatigue correction index, cardiovascular load dynamic coefficient, and exercise risk comprehensive early warning index for accurate assessment, taking into account the influence of environmental factors.

Benefits of technology

It enables precise quantification of muscle fatigue, dynamic assessment of cardiovascular load, and multi-dimensional scenario-based early warning of exercise risks, improving the accuracy and environmental adaptability of assessments, reducing the misjudgment rate, and providing personalized exercise risk warnings.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122320525A_ABST
    Figure CN122320525A_ABST
Patent Text Reader

Abstract

This invention discloses an intelligent exercise assessment and early warning system based on multi-source biosignal fusion, belonging to the field of exercise risk early warning technology. It collects multi-source data, including electromyography (EMG), heart rate, blood oxygen, skin conductance, and acceleration, from various wearable sensors to assess muscle fatigue, cardiovascular load, exercise status, and environmental factors. After preprocessing, the system uses quantitative models such as the muscle fatigue correction index and the cardiovascular load dynamic coefficient to calculate a comprehensive exercise risk early warning index by integrating physiological state and environmental factors. This results in a normalized risk classification, corresponding to low, medium, and high risk levels and marked with different color icons. This system overcomes the limitations of single signals, considering the synergistic effects of multi-dimensional physiological indicators and the influence of environmental factors, improving the accuracy of assessment in complex scenarios. It can effectively intervene before risks develop and is suitable for scenarios such as mass fitness and professional training. It is integrated into lightweight wearable devices without interfering with the exercise process.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of sports risk early warning technology, specifically to an intelligent sports assessment and early warning system based on multi-source biological signal fusion. Background Technology

[0002] With the popularization of the national fitness concept and the upgrading of sports and health needs, intelligent sports assessment and early warning systems have become a core technology direction in the field of sports and health. Existing systems are usually based on single or a few biological signals, such as heart rate and acceleration, to build assessment models. By collecting physiological data during exercise and combining it with preset thresholds, they can realize sports risk warnings. Such systems have been widely used in scenarios such as mass fitness and professional sports training, providing users with basic sports status monitoring and risk warnings, and reducing the probability of sports injuries to a certain extent.

[0003] Although existing systems possess basic exercise assessment capabilities, they still face numerous technical limitations in practical applications. These systems often rely on single biosignals for exercise load assessment, failing to adequately consider the synergistic effects of multiple physiological indicators such as muscle fatigue and sympathetic nerve activity. Furthermore, current health management programs struggle to incorporate external environmental factors like exercise scenarios into their assessment systems, leading to decreased accuracy in complex environments. Additionally, most devices employ limited biosignal acquisition methods, and some data acquisition is susceptible to interference from sweat secretion, resulting in data bias and misjudgments. Therefore, existing exercise management systems lack sufficient early risk prediction capabilities, hindering intervention before risks materialize. Their status assessment and early warning functions still require improvement. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent motion assessment and early warning system based on multi-source biosignal fusion, which solves the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an intelligent exercise assessment and early warning system based on multi-source biosignal fusion, comprising an exercise information acquisition component and an information preprocessing component. The exercise information acquisition component acquires exercise data during the exercise process, including muscle fatigue-related data, cardiovascular load-related data, and risk warning-related data. The acquired exercise data is input into the information preprocessing component for cleaning. The early warning system also includes a state recognition component, where the cleaned exercise data is input. The state recognition component:

[0006] Based on the root mean square value of electromyography (EMG) signal at time t, the resting root mean square baseline value, the average frequency of EMG signal at time t, the resting average frequency baseline value, the cumulative change in body surface temperature, and the temperature influence coefficient in muscle fatigue-related data, a muscle fatigue correction index is output. The muscle fatigue correction index enables more accurate quantification of fatigue state.

[0007] Based on the real-time heart rate at time t, the average resting heart rate, the maximum heart rate, the blood oxygen saturation at time t, the average resting blood oxygen saturation, the skin conductance tetany level at time t, the baseline value of the resting skin conductance tetany level, and the weighting coefficient in the cardiovascular load-related data, a dynamic coefficient of cardiovascular load is output, which enables a more comprehensive assessment of exercise load.

[0008] Based on the normalized motion acceleration value at time t, maximum motion acceleration at time t, real-time wind speed, respiratory rate at time t, resting average respiratory rate at time t, respiratory amplitude index at time t, resting average respiratory amplitude index at time t, altitude at time t, initial altitude, environmental comprehensive correction coefficient, and weighting coefficient in the risk warning related data, a comprehensive motion risk warning index is output. More accurate motion risk warnings are achieved through the comprehensive motion risk warning index.

[0009] Optionally, the state recognition component includes a muscle state calculation unit, a cardiovascular state calculation unit, and a state comprehensive calculation unit, wherein the cardiovascular state calculation unit further includes a signal processing module.

[0010] Optionally, the processing logic of the muscle state calculation unit is as follows: based on the resting state, quantify the energy change of electromyographic signals during exercise to reflect the relative change of muscle activation intensity; based on the resting state, quantify the relative change of the average frequency of electromyographic signals to reflect the change of muscle fiber recruitment type; couple the influence of temperature change on electromyographic signals to the spectral change to correct the spectral feature distortion caused by temperature drift; and convert the relative change into a normalized index between 0 and 200 for easy and intuitive assessment of fatigue level.

[0011] When skin temperature rises, the weight of the rate of spectral change is reduced by the reciprocal function to offset the temperature-induced spectral drift of the electromyographic signal. When skin temperature falls, the electromyographic signal is less affected by temperature and does not require correction, thus maintaining the original weight of the rate of spectral change.

[0012] Optionally, the processing logic of the cardiovascular status calculation unit is as follows: First, using resting heart rate as the benchmark and maximum heart rate as the upper limit, the relative load intensity of the current heart rate is quantified to reflect the basic load of the cardiovascular system. Then, using resting blood oxygen as the benchmark, the relative decrease in blood oxygen saturation is quantified to reflect the reduction in oxygen transport capacity. Next, using resting skin conductance as the benchmark, the relative increase in skin conductance tonicity level is quantified to reflect the degree of sympathetic nerve stress. The effects of oxygen transport capacity and sympathetic nerve stress are coupled to the heart rate reserve to correct the limitations of simple heart rate assessment. The muscle fatigue index is coupled to the cardiovascular load to reflect the indirect effect of muscle fatigue on the cardiovascular system. Finally, a normalized cardiovascular load index between 0 and 200 is output.

[0013] Optionally, the processing logic of the signal processing module is as follows: subtract the trend line fitted by the LOESS local weighted regression from the original electrodermal signal, filter the detrended signal with a 1Hz low-pass Butterworth filter, and divide the difference between the filtered signal and the real-time baseline by the real-time baseline to obtain the relative rate of change.

[0014] After converting the quantified value of the sweat effect into a compensation coefficient using a reciprocal function, the filtered signal is multiplied by the compensation coefficient to obtain the compensated EDA signal. Finally, the skin electrical tetany level at time t can be obtained by calculating the 60-second moving average.

[0015] Optionally, the processing logic of the state comprehensive calculation unit is as follows: it couples cardiovascular load and muscle fatigue to reflect the comprehensive load of the physiological system; it couples physiological load with normalized exercise intensity to reflect the amplification effect of exercise mechanical load on physiological load; it reduces the risk assessment value under high wind speed through a wind speed correction function to offset the false increase in physiological load caused by wind resistance; it quantifies the relative changes in respiratory rate and respiratory amplitude based on resting breathing; it amplifies the impact of abnormal respiratory characteristics on risk through an exponential function; it quantifies the relative amplitude of altitude change based on initial altitude; it amplifies the impact of altitude increase on risk through an exponential function; it integrates the influence of environmental factors such as wind speed, respiratory characteristics, altitude, seasonal climate, diurnal rhythm, and noise into the risk assessment; and finally outputs a normalized risk index between 0 and 500.

[0016] Optionally, the early warning system also includes an early warning component, which calculates the comprehensive early warning index of motion risk and inputs it into the early warning component.

[0017] Optionally, the processing logic of the warning component is as follows: the current movement status is divided into low risk, medium risk and high risk. When the comprehensive warning index of movement risk is less than 100, it is in low risk and the warning icon turns yellow. When the comprehensive warning index of movement risk is in the range of 100 to 250, it is in medium risk and the warning icon turns orange. When the comprehensive warning index of movement risk is higher than 250, it is in high risk and the warning icon turns red.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0019] I. This invention achieves precise quantification of muscle fatigue state by fusing the root mean square and average frequency features of electromyography (EMG) signals and combining them with a skin temperature correction function. The final output is a normalized fatigue index, which optimizes the distortion problem of single EMG signal evaluation in existing systems and eliminates the influence of individual muscle physiological differences and environmental temperature on the evaluation results, providing a more reliable physiological basis for exercise load assessment.

[0020] II. This invention achieves dynamic and comprehensive assessment of cardiovascular load through multi-source coupling of heart rate reserve, changes in blood oxygen saturation, skin conductance tetany level, and muscle fatigue index. Among them, heart rate reserve quantifies the basic load of the cardiovascular system, changes in blood oxygen saturation reflect the reduction in oxygen transport capacity, skin conductance tetany level reflects the stress state of the sympathetic nervous system, and muscle fatigue index couples the indirect effects of muscle fatigue on the cardiovascular system. Each parameter achieves differentiated contribution through weighting coefficients, optimizing the limitations of existing systems that rely solely on heart rate assessment, realizing the synergistic assessment of cardiovascular load and multi-dimensional physiological indicators, and improving the accuracy of exercise risk warning.

[0021] Furthermore, in acquiring data on skin conductance intensity levels, the influence of sweating on the data acquisition of athletes was eliminated, making the data more consistent with the actual situation and the assessment results more accurate.

[0022] Third, this invention achieves multi-dimensional, scenario-based assessment of exercise risk by coupling the dynamic coefficient of cardiovascular load, the muscle fatigue correction index, normalized acceleration, and multi-dimensional environmental factors. Furthermore, the environmental comprehensive correction coefficient incorporates the influence of factors such as seasonal climate, circadian rhythm, and noise, optimizing the existing system's insufficient adaptability to environmental factors and the lag in static assessment. This enables early prediction and real-time warning of exercise risk, providing a personalized and precise solution for risk assessment in different exercise scenarios. Attached Figure Description

[0023] Figure 1 This is a flowchart of the early warning process of the present invention.

[0024] Figure 2 This is a schematic diagram of the state recognition component of the present invention. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Example: Please refer to Figure 1 and Figure 2 This invention provides an intelligent exercise assessment and early warning system based on multi-source biological signal fusion, including an exercise information acquisition component and an information preprocessing component. The exercise information acquisition component acquires exercise data during the exercise process, including muscle fatigue-related data, cardiovascular load-related data, and risk warning-related data.

[0027] The muscle fatigue-related data includes the root mean square (RMS) value of the electromyography (EMG) signal at time t, calculated using a surface electromyography sensor attached to the abdominal region of the target muscle group and a 500ms sliding window with a 50% overlap ratio. emg (t), that is Where N is the number of sampling points within the window, X i The root mean square (RMS) baseline value of the electromyography (EMG) at the i-th sampling point within the window is obtained by acquiring signals from the target muscle group using a surface electromyography (SEM) sensor after the user maintains a resting posture for 1 minute. emg (0) The raw signal is acquired by the surface electromyography sensor. The power spectral density is obtained by performing a fast Fourier transform on the signal within the sliding window. The average frequency ME of the electromyography signal at time t is obtained by calculating the frequency-weighted average value of the 0-500Hz frequency band. emg (t), that is Where f is the frequency value, and P(f) is the power spectral density corresponding to frequency f, the user maintains a resting posture for 1 minute, and the target muscle group signals are collected through a surface electromyography sensor. The resting average frequency reference value ME is obtained by calculating the average frequency of the signal within 1 minute. emg (0), The cumulative change in body surface temperature ΔT is collected in real time by attaching a flexible thin-film wearable temperature sensor to the surface of the target muscle group. skin ΔT skin =T skin (t)-T skin (0), where T skin (t) represents the skin temperature at time t, T skin (0) represents the initial skin temperature at the start of exercise, with the temperature influence coefficient divided into upper limb muscles k = 0.1℃. -1 Lower limb muscle k = 0.2℃ -1Core muscle k = 0.3℃ -1 .

[0028] Furthermore, cardiovascular load-related data include real-time heart rate HR(t) at time t obtained through a wristband heart rate sensor, and the average resting heart rate HR obtained by calculating the average heart rate over 5 minutes of continuous heart rate data collected by the heart rate sensor after maintaining a resting state for 5 minutes. rest The maximum heart rate (HR) was calculated using the formula. max The calculation method is: HR max =208-0.7×age, the blood oxygen saturation SpO2(t) at time t is obtained by a blood oxygen sensor. After maintaining a resting state for 5 minutes, continuous blood oxygen data is collected by a blood oxygen sensor, and the average blood oxygen saturation within 5 minutes is calculated to obtain the resting average blood oxygen saturation SpO2(0). The raw skin electrical signal is collected by a flexible electrode wearable skin electrical sensor attached to the skin on the inside of the wrist. After sweat compensation, EDA is obtained. compensated (t), and then the EDA level of the skin conductance at time t is calculated using the 60-second moving average method. tonic (t);

[0029] The specific calculation method is as follows: EDA tonic (t)=1 / n∑ N i=1 EDA compensated (t-i), where N is the number of sampling points within 60 seconds. Continuous skin conductance signals are collected using a wearable skin conductance sensor after maintaining a resting state for 5 minutes. The resting skin conductance tetany level baseline value (EDA) is obtained by calculating the average tetany level over 5 minutes. tonic (0), and dynamic weighting coefficients α, β and γ, for endurance exercise α = 0.7, β = 0.2, γ = 0.1; for strength training α = 0.5, β = 0.3, γ = 0.2; for mixed exercise α = 0.6, β = 0.25, γ = 0.15.

[0030] Furthermore, the risk warning-related data includes real-time acquisition of X, Y, and Z axis acceleration signals using a lightweight triaxial accelerometer worn on the waist, and calculation of the triaxial acceleration magnitude. Then divide by the maximum dynamic acceleration ACC. max The normalized acceleration value ACC at time t obtained after (t) norm (t), the maximum acceleration ACC at time t is obtained by dynamically updating using the exponential smoothing method. max (t), the specific update method is as follows:

[0031] ACC max (t)=0.5×ACC max(t-1)+0.5×max(ACC(t-60s),t), where ACC max (t-1) represents the maximum acceleration at the previous moment, and max(ACC(t-60s),t) represents the maximum acceleration value within the past 60 seconds. The real-time wind speed ENV is directly acquired through a wearable miniature wind speed sensor. wind The resting average respiratory rate (EDR) was obtained by directly acquiring the respiratory rate EDR(t) at time t through a chest breathing belt sensor, maintaining a resting state for 5 minutes, and collecting continuous respiratory data through a breathing sensor. rest The average respiratory amplitude index (RAD) at time t was obtained by directly collecting data from the chest breathing belt sensor. After maintaining a resting state for 5 minutes, continuous chest expansion data was collected by the breathing sensor, and the average respiratory amplitude index over the 5 minutes was calculated to obtain the resting average respiratory amplitude index (RAD). rest The system measures real-time atmospheric pressure P using a built-in barometric pressure sensor, calculates the altitude ALT(t) at time t using the international standard atmospheric formula, obtains the initial altitude ALT0 automatically measured by the built-in barometric pressure sensor, and calculates the user's optimal exercise temperature T based on the user's exercise history data and physiological characteristics. opt Combined with the current seasonal average temperature T current The seasonal climate adaptation coefficient S is calculated as follows:

[0032] S=tan[(min(max(T current ,T opt -10),T opt +10)-T opt ) / 5], where the temperature is limited to T opt Within a range of ±10℃, to avoid excessive influence from extreme temperatures, the user's peak exercise time t is obtained based on analysis of the user's exercise history data. peak The biological clock coefficient C, calculated using the formula based on the current time t, is as follows:

[0033] C = -sin[(π×(t-t)] peak [)) / 12], measuring the equivalent continuous A-weighted sound level L in the environment using a wearable miniature noise sensor. eq Through the formula N = (L eq The acoustic environment quality coefficient N obtained by (-60) / 40, and the weighting coefficient ω 1-9 This allows for the quantification of the impact weights of cardiovascular load, muscle fatigue, wind speed, altitude changes, seasonal climate, diurnal rhythm, noise level, respiratory rate changes, and respiratory amplitude changes on comprehensive exercise risk assessment.

[0034] In this embodiment, the early warning system also includes a status recognition component. The collected motion data is input into the information preprocessing component, which cleans the motion data. The cleaned motion data is then input into the status recognition component.

[0035] Specifically, the state recognition component includes a muscle state calculation unit, which is based on the root mean square (RMS) value of the electromyography signal at time t in muscle fatigue-related data. emg (t), Root Mean Square Reference Value at Rest (RMS) emg (0) The average frequency ME of the electromyographic signal at time t emg (t), Resting average frequency reference value ME emg (0) Cumulative change in body surface temperature ΔT skin The temperature influence coefficient k is used to output the muscle fatigue correction index (FCI), which enables more accurate quantification of fatigue state.

[0036] Specifically, using the resting state as a baseline, the energy changes of electromyographic (EMG) signals during exercise are quantified to reflect the relative changes in muscle activation intensity. Using the resting state as a baseline, the relative changes in the average frequency of EMG signals are quantified to reflect the shift in muscle fiber recruitment types. The influence of temperature changes on EMG signals is coupled to the spectral changes, correcting spectral distortion caused by temperature drift. The relative changes are converted into a normalized exponent between 0 and 200 for easy and intuitive assessment of fatigue levels. The specific calculation method is as follows:

[0037]

[0038] When muscles are fatigued, the synchronicity of muscle fiber contraction decreases, and the energy intensity of electromyography (EMG) signals will show a trend of first increasing and then decreasing. Using the resting RMS as a benchmark, the ratio can eliminate individual muscle physiological differences and focus on the relative changes during exercise. Similarly, when muscles are fatigued, fast-twitch muscle fibers gradually withdraw from work due to their easy fatigue, and the proportion of slow-twitch muscle fibers increases, resulting in a decrease in the average frequency of EMG signals. Calculating the relative rate of change can quantify the degree of this spectral drift and reflect the fatigue process. Increased skin temperature will cause baseline drift and spectral distortion of EMG signals. This environmental interference can be offset by a temperature correction function. When the temperature drops, the EMG signal is less affected. Therefore, the correction coefficient is set to 1 to avoid over-correction.

[0039] Electromyographic energy ratio reflects the "quantity" of muscle activation, while the spectral change rate reflects the "quality" of fatigue. Combining the two can provide a more comprehensive assessment of fatigue status. Temperature correction enhances the environmental adaptability of the assessment, and the final results can be converted into percentages for easier understanding.

[0040] When skin temperature rises, the weight of the rate of spectral change is reduced using a reciprocal function to offset the temperature-induced spectral drift in the electromyographic (EMG) signal. When skin temperature decreases, the EMG signal is less affected by temperature and requires no correction, maintaining the original weight of the rate of spectral change. The specific calculation method is as follows:

[0041]

[0042] Elevated skin temperature can cause a pseudo-decrease in the average frequency of electromyographic signals (a spectral shift not caused by fatigue). This pseudo-fatigue can be corrected by reducing the weight of the rate of spectral change using a reciprocal function, k×ΔT. skin Since it is a dimensionless quantity, the function output is ensured to be between 0 and 1 to avoid over-correction. When the skin temperature decreases, the spectral characteristics of the electromyographic signal are less affected, and the pseudo-drift can be ignored. Therefore, the correction coefficient is set to 1 to maintain the original evaluation value of the spectral change rate.

[0043] For k, considering the different fat thickness and blood circulation capacity of different muscle groups, the core muscles (such as the abdomen and back) have thicker fat, and temperature changes have a greater impact on electromyographic signals, so a higher correction coefficient is set; the upper limb muscles have thinner fat, and temperature has a smaller impact, so a lower correction coefficient is set. Therefore, the above reference value is set to achieve differentiated temperature correction.

[0044] Furthermore, the status recognition component also includes a cardiovascular status calculation unit, which calculates the real-time heart rate HR(t) at time t and the resting average heart rate HR based on cardiovascular load-related data. rest Maximum heart rate (HR) max Oxygen saturation SpO2(t) at time t, mean resting oxygen saturation SpO2(0), and EDA level of skin conductance at time t. tonic (t), Resting skin electrical conductivity level baseline value (EDA) tonic (0) and weighting coefficients α, β and γ, output the dynamic coefficient of cardiovascular load DCL, and achieve a more comprehensive exercise load assessment through the dynamic coefficient of cardiovascular load DCL;

[0045] Specifically, firstly, using resting heart rate as the baseline and maximum heart rate as the upper limit, the relative load intensity of the current heart rate is quantified to reflect the basal load of the cardiovascular system. Then, using resting blood oxygen saturation as the baseline, the relative decrease in blood oxygen saturation is quantified to reflect the reduction in oxygen transport capacity. Next, using resting skin conductance as the baseline, the relative increase in skin conductance tonicity is quantified to reflect the degree of sympathetic nerve stress. The effects of oxygen transport capacity and sympathetic nerve stress are coupled to heart rate reserve, correcting the limitations of simple heart rate assessment. Finally, the muscle fatigue index is coupled to cardiovascular load to reflect the indirect impact of muscle fatigue on the cardiovascular system. The final output is a normalized cardiovascular load index between 0 and 200. The specific calculation method is as follows:

[0046]

[0047]

[0048]

[0049] More specifically, traditional heart rate assessment only focuses on absolute heart rate, ignoring the differences between individual baseline heart rate and maximum heart rate. In this embodiment, heart rate reserve quantifies cardiovascular load through relative values, eliminating individual physiological differences and more accurately reflecting the degree of stimulation of the cardiovascular system by exercise. When the exercise load is too high, oxygen consumption increases and blood oxygen saturation decreases, which will cause the cardiovascular system to work harder to maintain oxygen supply. Therefore, heart rate reserve can be corrected by blood oxygen saturation rate, reflecting the impact of oxygen transport capacity on cardiovascular load.

[0050] Similarly, during exercise, the sympathetic nervous system is excited, and the level of skin conductance increases, reflecting the stress state of the cardiovascular system. The rate of change in skin conductance can be used to correct the heart rate reserve, reflecting the impact of autonomic nervous regulation on cardiovascular load. When muscles are fatigued, the body needs more energy to maintain exercise, which indirectly increases the load on the cardiovascular system. Therefore, coupling the FCI index to DCL can reflect the interaction between muscle fatigue and cardiovascular load, improving the comprehensiveness of the assessment.

[0051] For the weighting coefficient α, oxygen transport capacity is the limiting factor in endurance exercise, so it is given a higher weight. In strength training, muscle strength is the main limiting factor, and the impact of oxygen transport is relatively small, so it is given a lower weight. For the weighting coefficient β, the sympathetic nerve excitation level is higher in strength training, which has a greater impact on cardiovascular load, so it is given a higher weight. In endurance exercise, the sympathetic nerve excitation level is relatively stable, so it is given a lower weight. Similarly, for the weighting coefficient γ, muscle fatigue has a greater indirect impact on cardiovascular load in strength training, so it is given a higher weight. In endurance exercise, muscle fatigue develops more slowly and has a smaller impact on cardiovascular load, so it is given a lower weight.

[0052] Furthermore, the cardiovascular status calculation unit also includes a signal processing module. The processing logic of the signal processing module is as follows: first, subtract the trend line fitted by the LOESS local weighted regression from the original electrodermal signal.

[0053]

[0054] EDA detrended (t) represents the detrended EDA signal, eliminating baseline drift caused by factors such as prolonged wear and slow changes in skin humidity, highlighting rapidly changing components related to sympathetic nerve activity. raw(t) represents the raw skin electrodermal signal collected at time t during exercise, which includes various components such as sympathetic nerve activity, sweat secretion, and environmental interference. It is directly collected by a wearable skin electrodermal sensor with flexible electrodes.

[0055] The raw electrodermal signal contains slow baseline drift (such as signal drift caused by changes in skin humidity or prolonged sensor wear). LOESS regression can effectively remove this slow trend and highlight the fast-changing components related to sympathetic nerve activity. a = 0.1 is the smoothing parameter. A smaller value indicates that the fitted trend line is closer to the original signal and can remove baseline drift more accurately. LOESS local weighted regression is a non-linear data smoothing algorithm, which belongs to the existing algorithm processing method. By weighting local data points and fitting a regression curve, it can extract the slow changing trend in the original signal while retaining the fast fluctuation components of the signal.

[0056] The detrended signal is filtered using a 1Hz low-pass Butterworth filter:

[0057]

[0058] EDA filtered (t) represents the filtered electrodermal signal, which removes high-frequency noise and motion artifacts, retains effective signal components related to sympathetic nerve activity, and improves the accuracy of electrodermal characteristics. c =1Hz is the cutoff frequency, S(t) represents the quantified value of the influence of sweat at time t during exercise. Quantifying the degree of interference of sweat secretion on the skin electrical signal provides a basis for subsequent sweat compensation. EDA baseline (t) represents the real-time baseline of the skin electrical signal at time t during the exercise, reflecting the basic level change of the skin electrical signal;

[0059] The Butterworth filter is a linear phase filter designed to achieve the flattest frequency response within the passband. It can effectively remove noise while preserving the original phase information of the signal to the greatest extent. It is an existing technology, and the specific filtering principle will not be elaborated here.

[0060] The relative rate of change is obtained by dividing the difference between the filtered signal and the real-time baseline by the real-time baseline.

[0061]

[0062] EDA baseline (t) represents the real-time baseline of the EDA signal at time t during the exercise, updated in real time using exponential smoothing: EDA baseline (t)=EDA baseline (t-1)×(1-γ)+EDA filtered(t)×γ, where γ=0.01 is the smoothing coefficient. When the baseline drift detection value D(t)>0.2, γ is adjusted to 0.1 to trigger a fast update. The baseline drift detection value D(t) is shown below.

[0063] S(t) represents the quantified value of the sweat effect at time t during exercise, reflecting the degree of influence of current sweat secretion on the skin electrical signal. After converting the quantified sweat effect value into a compensation coefficient using a reciprocal function, the filtered signal is multiplied by the compensation coefficient to obtain the compensated EDA signal, i.e.:

[0064]

[0065]

[0066] Where k(t) represents the sweat compensation coefficient at time t during exercise. The greater the influence of sweat, the smaller the compensation coefficient, which can effectively counteract the false increase in skin electrical signal caused by sweat. 0.5 can balance the compensation intensity and signal fidelity. EDA compensated (t) represents the skin electrical signal after sweat compensation, which eliminates the interference of sweat secretion on the signal. The distortion of the skin electrical signal caused by sweat secretion is canceled out by multiplication operation, so as to obtain the signal reflecting the true sympathetic nerve activity. Finally, the skin electrical tetany level EDA at time t can be obtained by calculating a 60-second moving average. tonic (t):

[0067]

[0068] Furthermore, during exercise, the real-time baseline EDA of the skin's electrical activity at time t is... baseline After (t) calculation, baseline drift detection will be performed, specifically as follows:

[0069]

[0070] D(t) represents the baseline drift detection value at time t during the movement, reflecting the degree of deviation between the current skin electrodermal signal and the real-time baseline. If D(t) > 0.2, the update coefficient will be changed to 0.1, as described above.

[0071] Furthermore, the state recognition component also includes a state integration calculation unit, which is based on the normalized motion acceleration value (ACC) at time t in the risk warning related data. norm (t), maximum acceleration ACC at time t max (t), Real-time wind speed ENV wind respiratory rate at time t (EDR(t)) and resting mean respiratory rate (EDR) rest , respiratory amplitude index at time t (RAD(t)) and resting mean respiratory amplitude index (RAD)rest The comprehensive motion risk warning index (CRI) is output by taking the altitude ALT(t) at time t, the initial altitude ALT0, the environmental comprehensive correction coefficient, and the weighting coefficient. More accurate motion risk warnings are achieved through the CRI.

[0072] Specifically, the processing logic of the state comprehensive calculation unit is as follows: It couples cardiovascular load and muscle fatigue to reflect the comprehensive load of the physiological system; it couples physiological load with normalized exercise intensity to reflect the amplification effect of exercise mechanical load on physiological load; it reduces the risk assessment value under high wind speeds through a wind speed correction function to offset the false increase in physiological load caused by wind resistance; it quantifies the relative changes in respiratory rate and amplitude based on resting breathing, and amplifies the impact of abnormal respiratory characteristics on risk through an exponential function; it quantifies the relative amplitude of altitude changes based on initial altitude, and amplifies the impact of altitude increase on risk through an exponential function; it integrates the influence of environmental factors such as wind speed, respiratory characteristics, altitude, seasonal climate, diurnal rhythm, and noise into the risk assessment; and finally, it outputs a normalized risk index between 0 and 500. The specific calculation method is as follows:

[0073]

[0074]

[0075]

[0076]

[0077] More specifically, cardiovascular load and muscle fatigue are the core physiological basis of exercise risk, and their relative contributions to risk can be reflected by weighted coupling; cardiovascular load is the main source of acute risk, so it is given a higher weight, ω1 is 0.7, while muscle fatigue is the main source of chronic risk, so it is given a lower weight, ω2 is 0.2.

[0078] The same physiological load carries different risks at different exercise intensities; the higher the exercise intensity, the higher the risk. By coupling with normalized acceleration, physiological load can be correlated with actual exercise intensity, improving the fit of risk assessment. When exercising against the wind, increased wind resistance can lead to a false increase in physiological load, but the actual exercise risk does not increase proportionally. Therefore, by using a correction function that adds 1 to the denominator, the risk assessment value under high wind speeds can be reduced, avoiding misjudgment. The impact of wind speed on risk is "corrective" rather than "contributive," hence its weight ω3 = 0.1 s / m.

[0079] Similarly, abnormal changes in respiratory rate and amplitude are early warning signals of exercise risks. An exponential function can amplify the impact of these abnormalities on risk, enabling early warning. The characteristic of an exponential function is that when the input is positive, the output increases rapidly, effectively highlighting abnormal changes in respiratory characteristics. Increased altitude leads to a decrease in oxygen partial pressure, increasing the burden on the cardiovascular and respiratory systems, resulting in an exponential increase in risk. An exponential function can amplify the impact of altitude changes on risk, reflecting the risk characteristics of high-altitude exercise.

[0080] Since changes in respiratory rate are an early warning signal of exercise risk, their impact needs to be amplified when abnormalities occur. Therefore, ω4 is set to 0.6. Changes in respiratory amplitude have a relatively small impact on risk, but they need to be assessed in conjunction with respiratory rate. Therefore, ω5 is set to 0.4. Finally, changes in altitude have a significant impact on risk, but they need to be balanced with respiratory characteristics and physiological load. Therefore, a moderate initial weight is set for ω6, which is set to 0.15. Increased altitude leads to a decrease in oxygen partial pressure, which increases the load on the cardiovascular and respiratory systems, and the risk increases exponentially. An exponential function can amplify the impact of altitude changes on risk and reflect the risk characteristics of high-altitude exercise.

[0081]

[0082] D represents the comprehensive environmental correction factor. Environmental factors such as seasonal climate, circadian rhythms, and noise can affect physical function and athletic performance. The comprehensive environmental correction factor D can be used to incorporate these factors into the risk assessment system, improving the comprehensiveness of the assessment. Specifically:

[0083] Both excessively high and low temperatures increase the risk of exercise. Using the structure 1 + ω7 × S, when the seasonal climate adaptation coefficient S is positive (temperature is higher than the optimum temperature), the correction coefficient is greater than 1, and the risk assessment value increases; when the seasonal climate adaptation coefficient S is negative (temperature is lower than the optimum temperature), the correction coefficient is less than 1, and the risk assessment value decreases. This reflects the two-way impact of climate on risk. The impact of seasonal climate on exercise risk is moderate, so a moderate initial weight ω7 is set to 0.15.

[0084] Exercising during off-peak exercise times results in lower physical function and increased risk. Using a structure of 1 + ω8 × C, when the biological clock coefficient C is positive (current time is close to peak exercise time), the correction coefficient is greater than 1, and the risk assessment value decreases. When the biological clock coefficient C is negative (current time is far from peak exercise time), the correction coefficient is less than 1, and the risk assessment value increases. This reflects the impact of the biological clock on risk. Circadian rhythms have a significant impact on exercise risk; therefore, a relatively high initial weight ω8 of 0.2 is set.

[0085] High noise levels can be distracting and increase the risk of motion. Using the structure 1 + ω9 × N, when the ambient noise quality coefficient N is positive (noise is above the comfort threshold), the correction coefficient is greater than 1, and the risk assessment value increases. When the ambient noise quality coefficient N is negative (noise is below the comfort threshold), the correction coefficient is less than 1, and the risk assessment value decreases. This reflects the impact of noise on risk. The impact of noise level on motion risk is relatively small, so a low initial weight is set, with ω9 being 0.1.

[0086] More specifically, the seasonal climate adaptation coefficient S is calculated as follows: first, the current temperature is limited to T. opt Within a range of ±10℃, the difference from the optimum temperature is calculated and divided by 5. Finally, it is converted to a coefficient between -1 and 1 using the hyperbolic tangent function, i.e.:

[0087]

[0088] The hyperbolic tangent function has the characteristic that the output is 0 when the input is 0, and the output approaches ±1 as the absolute value of the input increases. It can convert the difference between the air temperature and the optimum temperature into a smooth adaptation coefficient. Limiting the air temperature range can avoid over-correction caused by extreme temperatures, ensuring that the coefficient is within a reasonable range. opt The optimal exercise temperature for users is calculated based on their exercise history data. 5 is a scaling factor for the temperature difference. A value of 5 makes the coefficient approach ±1 when the temperature changes by 10℃, reflecting the significant impact of temperature on exercise risk.

[0089] The biological clock coefficient C is calculated as follows: calculate the difference between the current time and the peak exercise time, convert it to radians, take the sine value, and then take the negative value. That is:

[0090]

[0091] The period of the sine function is 24 hours, which reflects the circadian rhythm characteristics of the biological clock; taking a negative value makes the coefficient 1 during peak exercise time and negative during off-peak time, which conforms to the influence of the biological clock on exercise risk. peak The peak exercise time for users is calculated based on their exercise history data; 12 is the scaling factor for the time difference. A value of 12 makes the time difference 12 hours with a coefficient of 1, reflecting the maximum impact of the biological clock on exercise risk.

[0092] The environmental sound quality factor N is calculated as follows: subtract the comfort threshold of 60 dB from the equivalent continuous A-weighted sound level, and then divide by 40, that is:

[0093]

[0094] 60 dB is the noise threshold for a comfortable exercise environment. When the noise exceeds 60 dB, the coefficient is positive, and the risk assessment value increases; when the noise is below 60 dB, the coefficient is negative, and the risk assessment value decreases; dividing by 40 makes the coefficient 1 when the noise reaches 100 dB (high noise environment), reflecting the significant impact of high noise on exercise risk, where L eq The environmental equivalent continuous A-weighted sound level was obtained through a wearable noise sensor.

[0095] Furthermore, the early warning system also includes an early warning component, which is equipped with an early warning icon. After calculating the Comprehensive Risk Early Warning Index (CRI) for motion, the data is input into the early warning component. The early warning component categorizes the current motion status into low risk, medium risk, and high risk. When the CRI is less than 100, the system is considered low risk, and the early warning icon turns yellow. When the CRI is between 100 and 250, the system is considered medium risk, and the early warning icon turns orange. When the CRI is greater than 250, the system is considered high risk, and the early warning icon turns red.

[0096] It should be emphasized that all data that needs to be acquired through the collection device in this embodiment can be integrated into lightweight wearable devices, such as fitness trackers, fitness belts, and fitness vests, including but not limited to the above-mentioned wearable devices, to ensure that the data is acquired smoothly without interfering with the person's exercise process.

[0097] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An intelligent exercise evaluation and early warning system based on multi-source biosignal fusion, characterized in that: It includes a motion information acquisition component and an information preprocessing component. The motion information acquisition component acquires motion data during the exercise process, including data related to muscle fatigue, cardiovascular load, and risk warning. The collected motion data is input into the information preprocessing component, which cleans the motion data. The early warning system also includes a status recognition component. The cleaned motion data is input into the status recognition component, which is: Based on the root mean square value of electromyography (EMG) signal at time t, the resting root mean square baseline value, the average frequency of EMG signal at time t, the resting average frequency baseline value, the cumulative change in body surface temperature, and the temperature influence coefficient in muscle fatigue-related data, a muscle fatigue correction index is output. The muscle fatigue correction index enables more accurate quantification of fatigue state. Based on the real-time heart rate at time t, the average resting heart rate, the maximum heart rate, the blood oxygen saturation at time t, the average resting blood oxygen saturation, the skin conductance tetany level at time t, the baseline value of the resting skin conductance tetany level, and the weighting coefficient in the cardiovascular load-related data, a dynamic coefficient of cardiovascular load is output, which enables a more comprehensive assessment of exercise load. Based on the normalized motion acceleration value at time t, maximum motion acceleration at time t, real-time wind speed, respiratory rate at time t, resting average respiratory rate at time t, respiratory amplitude index at time t, resting average respiratory amplitude index at time t, altitude at time t, initial altitude, environmental comprehensive correction coefficient, and weighting coefficient in the risk warning related data, a comprehensive motion risk warning index is output. More accurate motion risk warnings are achieved through the comprehensive motion risk warning index.

2. The intelligent exercise evaluation and early warning system based on multi-source biosignal fusion according to claim 1, characterized in that: The state recognition component includes a muscle state calculation unit, a cardiovascular state calculation unit, and a state comprehensive calculation unit. The cardiovascular state calculation unit also includes a signal processing module.

3. The intelligent motion assessment and early warning system based on multi-source biosignal fusion according to claim 2, characterized in that: The processing logic of the muscle state calculation unit is as follows: based on the resting state, the energy change of electromyographic signals during exercise is quantified to reflect the relative change of muscle activation intensity; based on the resting state, the relative change of the average frequency of electromyographic signals is quantified to reflect the change of muscle fiber recruitment type; the influence of temperature change on electromyographic signals is coupled to the spectral change to correct the spectral feature distortion caused by temperature drift; and the relative change is converted into a normalized index between 0 and 200 for easy and intuitive assessment of fatigue level. When skin temperature rises, the weight of the rate of spectral change is reduced by the reciprocal function to offset the temperature-induced spectral drift of the electromyographic signal. When skin temperature falls, the electromyographic signal is less affected by temperature and does not require correction, thus maintaining the original weight of the rate of spectral change.

4. The intelligent motion assessment and early warning system based on multi-source biosignal fusion according to claim 3, characterized in that: The processing logic of the cardiovascular status calculation unit is as follows: First, using resting heart rate as the benchmark and maximum heart rate as the upper limit, the relative load intensity of the current heart rate is quantified to reflect the basic load of the cardiovascular system. Then, using resting blood oxygen as the benchmark, the relative decrease in blood oxygen saturation is quantified to reflect the reduction in oxygen transport capacity. Next, using resting skin conductance as the benchmark, the relative increase in skin conductance tonicity level is quantified to reflect the degree of sympathetic nerve stress. The effects of oxygen transport capacity and sympathetic nerve stress are coupled to the heart rate reserve to correct the limitations of simple heart rate assessment. The muscle fatigue index is coupled to the cardiovascular load to reflect the indirect effect of muscle fatigue on the cardiovascular system. Finally, a normalized cardiovascular load index between 0 and 200 is output.

5. The intelligent motion assessment and early warning system based on multi-source biosignal fusion according to claim 4, characterized in that: The processing logic of the signal processing module is as follows: subtract the trend line fitted by the LOESS local weighted regression from the original electrodermal signal, filter the detrended signal with a 1Hz low-pass Butterworth filter, and divide the difference between the filtered signal and the real-time baseline by the real-time baseline to obtain the relative rate of change. After converting the quantified value of the sweat effect into a compensation coefficient using a reciprocal function, the filtered signal is multiplied by the compensation coefficient to obtain the compensated EDA signal. Finally, the skin electrical tetany level at time t can be obtained by calculating the 60-second moving average.

6. The intelligent motion assessment and early warning system based on multi-source biosignal fusion according to claim 5, characterized in that: The processing logic of the state comprehensive calculation unit is as follows: it couples cardiovascular load and muscle fatigue to reflect the comprehensive load of the physiological system; it couples physiological load with normalized exercise intensity to reflect the amplification effect of exercise mechanical load on physiological load; it reduces the risk assessment value under high wind speed through a wind speed correction function to offset the false increase in physiological load caused by wind resistance; it quantifies the relative changes in respiratory rate and respiratory amplitude based on resting breathing; it amplifies the impact of abnormal respiratory characteristics on risk through an exponential function; it quantifies the relative amplitude of altitude change based on initial altitude; it amplifies the impact of altitude increase on risk through an exponential function; it integrates the influence of environmental factors such as wind speed, respiratory characteristics, altitude, seasonal climate, diurnal rhythm, and noise into the risk assessment; and finally outputs a normalized risk index between 0 and 500.

7. The intelligent motion assessment and early warning system based on multi-source biosignal fusion according to any one of claims 2-6, characterized in that: The early warning system also includes an early warning component, which calculates the comprehensive early warning index of motion risk and inputs it into the early warning component.

8. The intelligent motion assessment and early warning system based on multi-source biosignal fusion according to claim 7, characterized in that: The processing logic of the warning component is as follows: the current movement status is divided into low risk, medium risk and high risk. When the comprehensive movement risk warning index is less than 100, it is in low risk and the warning icon turns yellow. When the comprehensive movement risk warning index is between 100 and 250, it is in medium risk and the warning icon turns orange. When the comprehensive movement risk warning index is higher than 250, it is in high risk and the warning icon turns red.