Wristband multi-mode perception-based health assessment and intelligent management system

Abstract

The invention discloses a health assessment and intelligent management system based on wrist strap multi-mode perception, and relates to the technical field of intelligent health assessment, and the system comprises a data collection and preprocessing module which is used for periodically collecting multi-mode wrist strap perception data, and carrying out the preprocessing of the collected data; the humidity influence correction module is used for evaluating the influence degree of humidity on the multi-mode wrist strap sensing data and correcting the multi-mode wrist strap sensing data with humidity interference by adopting a regression model; the health assessment analysis module is used for assessing the health state of the user and generating a corresponding health level label; and the health intervention feedback module is used for formulating a graded health strategy report based on the health grade label and dynamically adjusting a health intervention plan in combination with an intervention effect. The problem that in the prior art, interference of humidity changes on skin resistance and other physiological signal data cannot be effectively removed, and consequently the health assessment result is inaccurate is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent health assessment, in particular to a health assessment and intelligent management system based on wristband multi-modal sensing. BACKGROUND

[0002] With the rapid development of modern technology, the field of health management and health assessment has gradually introduced intelligent technology, especially in multi-modal data acquisition and analysis. Many health monitoring systems have been able to use intelligent sensing devices to collect physiological signals of users in real time, and convert these data into valuable information. These systems not only can monitor personal health in real time, but also can predict potential health risks to a certain extent, and provide personalized health management solutions for users. In recent years, with the popularity of wearable devices and the progress of data processing technology, health assessment and intelligent intervention systems based on physiological signals have gradually been applied in the fields of medicine, sports health, and elderly care.

[0003] For example, the invention with publication number CN115497627A discloses a health assessment intelligent management system, relating to the technical field of health assessment management. The data acquisition unit of the present application is used to record the information provided by the old, weak, sick and disabled for subsequent input; the data input unit is used to input the physical indicators of the old, weak, sick and disabled into the health assessment management system; the data processing unit is used to process the indicator data; the data output unit is used to output and display the graphics generated by the data processing unit.

[0004] For example, the invention with publication number CN120977582A discloses an intelligent health assessment analysis method and system based on health management, including the following steps: obtaining the direction conflict path in the monitoring period, tracking the offset change and positioning the synchronous divergence position, extracting the trend switching section and rhythm repeating path, extending the fluctuation range to divide the bifurcation structure, identifying the trend start and end and tracking the direction trend, and obtaining the health status classification judgment label group. In the present application, by identifying the direction conflict and rhythm fluctuation change in the index path, the continuous trajectory structure is constructed and the trend differentiation section is extracted, the path continuation sequence and fluctuation rhythm distribution are combined, the rhythm tracking of state changes and trend turning points is completed, the label sequence with direction perception and rhythm recognition characteristics is generated under multi-time sequence change, the response ability of health status division to trend evolution is enhanced, and the capture ability of evaluation to potential state change and the adaptability of classification and layering are improved.

[0005] However, despite the significant progress made by current health management systems in many areas, there are still some problems in the existing technology that limit its widespread application and further improvement of effectiveness. First, most systems still rely on single physiological data for evaluation, lacking comprehensive analysis of multi-modal data, resulting in insufficient accuracy and comprehensiveness of evaluation results. Second, existing health assessment and intervention systems often lack sufficient real-time and adaptability, often unable to adjust intervention strategies in real time according to changes in health status, resulting in insufficient timeliness and individualization of intervention measures.

[0006] Therefore, in view of the above problems, there is an urgent need for a health assessment and intelligent management system based on wristband multi-modal perception. SUMMARY

[0007] Technical problems solved

[0008] In view of the deficiencies of the prior art, the present application provides a health assessment and intelligent management system based on wristband multi-modal perception, which solves the problem that the prior art cannot effectively remove the influence of humidity changes on skin resistance and other physiological signal data, resulting in inaccurate health assessment results.

[0009] Technical solutions

[0010] To achieve the above purpose, the present application realizes the following technical solutions: a health assessment and intelligent management system based on wristband multi-modal perception, comprising: a data acquisition and preprocessing module for periodically acquiring multi-modal wristband perception data, and performing time alignment, noise suppression, outlier rejection, missing value completion and scale normalization processing on the acquired data to generate preprocessed multi-modal wristband perception data; a humidity influence correction module for constructing a reference value set based on the preprocessed multi-modal wristband perception data, evaluating the influence of humidity on the multi-modal wristband perception data, determining whether the current sampling exists humidity interference, and correcting the multi-modal wristband perception data with humidity interference using a regression model; a health assessment and analysis module for constructing a perception feature data set based on the humidity corrected multi-modal wristband perception data, evaluating the health status of the user, and determining the health level of the user according to the evaluation result, and generating the corresponding health level label; a health intervention feedback module for formulating a graded health strategy report based on the health level label, and continuously monitoring the change of the user's health status, dynamically adjusting the health intervention plan combined with the intervention effect, and realizing the health management closed loop.

[0011] Further, the specific steps of periodically collecting multi-modal wristband sensing data and performing time alignment, noise suppression, abnormality rejection, missing value completion, and scale normalization on the collected data to generate pre-processed multi-modal wristband sensing data are as follows: set a fixed-width time window as a sampling period, periodically collect multi-modal wristband sensing data, and the multi-modal wristband sensing data includes electromyography signals, electrodermal activity signals, pulse wave signals, wrist skin temperature, heart rate, respiratory rate, wristband humidity, and skin resistance; based on the least squares interpolation algorithm, synchronize and correct the multi-source time series of the collected multi-modal wristband sensing data, and use the double exponential smoothing filter algorithm to smooth and suppress the sudden spikes and high-frequency disturbances in the multi-modal wristband sensing data; automatically identify and reject abnormal segments in the multi-modal wristband sensing data through the local outlier factor algorithm, and use the Lagrange interpolation method to reconstruct the continuity of short missing intervals; and scale the multi-modal wristband sensing data to a standard scale and unify the numerical values by using the Z-Score standardization algorithm.

[0012] Further, the specific steps of constructing a reference value set based on the pre-processed multi-modal wristband sensing data are as follows: based on the pre-processed multi-modal wristband sensing data, arrange them in the order of sampling time, calculate the skin resistance difference value of adjacent sampling periods, when the continuous N skin resistance difference values are lower than the difference threshold value, calculate the mean value of each multi-modal wristband sensing data in the corresponding sampling period, and record it as the reference value of the corresponding data, and collectively construct a reference value set.

[0013] Further, the specific steps of evaluating the influence of humidity on multi-modal wristband sensing data are as follows: extract real-time multi-modal wristband sensing data, calculate the absolute difference between the electrodermal activity signal, electromyography signal, and pulse wave signal and the corresponding reference value, and sum the three absolute difference values to obtain the signal difference total term; calculate the square of the difference between the wrist skin temperature and the skin resistance and the corresponding reference value, add the two square differences, add one, and then take the reciprocal to obtain the temperature resistance correction term; take the natural logarithm of the absolute difference between the wristband humidity and the corresponding reference value after adding one to obtain the humidity change logarithmic term; multiply the signal difference total term and the temperature resistance correction term, and then multiply the humidity logarithmic term to obtain the humidity influence evaluation value.

[0014] Further, the specific steps for determining whether humidity interference exists in the current sampling and using a regression model to correct the multimodal wristband sensing data with humidity interference are as follows: Real-time comparison of the humidity impact assessment value and the humidity impact threshold. When the humidity impact assessment value is less than or equal to the humidity impact threshold, the current sampling is considered normal and no processing is performed. When the humidity impact assessment value is greater than the humidity impact threshold, the current sampling is considered to have humidity interference, and the humidity correction process begins: Extract multimodal wristband sensing data and humidity impact assessment values ​​from historical sampling periods where the skin resistance difference is greater than the difference threshold. Construct a regression model based on the extreme gradient boosting regression algorithm, and input the multimodal wristband sensing data from historical periods and the corresponding benchmark value set into the regression model for training and optimization. Input the humidity impact assessment value and multimodal wristband sensing data from the current sampling period into the trained regression model, and output the humidity-corrected multimodal wristband sensing data.

[0015] Furthermore, the specific steps for constructing a sensing feature dataset based on humidity-corrected multimodal wristband sensing data are as follows: Based on humidity-corrected multimodal wristband sensing data, calculate the mean and variance of each multimodal wristband sensing data in the most recent M sampling periods, and extract the maximum skin temperature and minimum skin resistance to jointly construct the sensing feature dataset.

[0016] Further, the specific steps for assessing a user's health status are as follows: Extract the sensory feature dataset and the baseline value set; sum the mean values ​​of skin electrokinetic activity signal, electromyographic signal, and pulse wave signal sequentially, and then multiply by the signal fusion coefficient to obtain the signal fusion term; divide the absolute value of the difference between the mean heart rate and the baseline heart rate value by the heart rate sensitivity coefficient, and take the exponential function value of the natural constant e with the negative of the obtained ratio as the exponent to obtain the heart rate deviation index term; calculate the square of the difference between the maximum skin temperature and the minimum skin resistance value to obtain the temperature and humidity difference correction term; add one to the absolute value of the difference between the mean respiratory rate and the baseline respiratory rate value, and then take the natural logarithm to obtain the respiratory rate adjustment term; multiply the signal fusion term and the heart rate deviation index term, and divide by the sum of the constant one, the temperature and humidity difference correction term, and the respiratory rate adjustment term to obtain the health assessment value.

[0017] Furthermore, the specific steps for determining the user's health level based on the assessment results and generating the corresponding health level label are as follows: The health assessment value... Respectively compared with the first-level health threshold and secondary health threshold Real-time comparison is performed to determine the user's health level and generate a corresponding health level label: when S≤ When the current state is determined to be healthy, it is marked as Level 1 healthy; when < < When the current state is determined to be sub-healthy, it is marked as Level 2 health; when ≥ When the health level label is the third level health, it is determined that the current health state is unhealthy, and the third level health label is marked; the health level label, the health assessment value, the perception feature data set and the corresponding multi-modal wristband perception data are stored in structure, and the health assessment data set is constructed.

[0018] Further, the specific steps of formulating a hierarchical health strategy report based on the health level label are as follows: extracting the health assessment data set, and generating a hierarchical health strategy report based on the health level label: when the label is the first level health, the health assessment data set is extracted to generate a health maintenance report; when the label is the second level health, the health assessment data set is extracted and a sub-health adjustment report is generated, and a health intervention plan is pushed through the mobile terminal; when the label is the third level health, the health assessment data set is extracted and a health intervention report is generated, and a physical examination warning is issued.

[0019] Further, the specific steps of continuously monitoring the change of the user's health state and dynamically adjusting the health intervention plan combined with the intervention effect to realize the health management closed loop are as follows: taking one week as an intervention period, extracting the health assessment values before and after each intervention period, dividing the difference between the health assessment value after intervention and the health assessment value before intervention by the health assessment value before intervention to obtain the health assessment change ratio; multiplying the difference between the health assessment value after intervention and the first level health threshold by the health adjustment coefficient, taking the opposite number of the product as the index, calculating the exponential function value of the natural constant e, then taking the inverse of the obtained exponential function value plus one, and subtracting the obtained inverse value from the constant one to obtain the health improvement factor; multiplying the health assessment change ratio by the health improvement factor to obtain the intervention effect evaluation value; dynamically adjusting the health intervention plan combined with the intervention effect evaluation value and the health assessment data set in the current intervention period, and generating a period health progress report to realize the health management closed loop.

[0020] Beneficial effects

[0021] The present application has the following beneficial effects:

[0022] (1) The health assessment and intelligent management system based on wristband multi-modal perception can effectively eliminate the interference of external humidity on skin resistance and other physiological signals by introducing a humidity influence correction module. The data after humidity interference is corrected using a regression model, which improves the data accuracy in the health assessment process and avoids misleading evaluation due to humidity changes.

[0023] (2) The health assessment and intelligent management system based on wristband multi-modal perception can comprehensively reflect the health state of the user by comprehensively evaluating the health based on various multi-modal wristband perception data, avoid the deviation caused by a single data source, and improve the reliability and accuracy of the evaluation result.

[0024] (3) The health assessment and intelligent management system based on wristband multi-modal sensing can make a hierarchical intervention strategy by combining the health level label of the user, and continuously track the change of the health status. By combining the intervention effect evaluation, the health intervention plan is dynamically optimized, and the adaptability and continuity of the health management are effectively improved.

[0025] (4) The health assessment and intelligent management system based on wristband multi-modal sensing can capture the change of the health trend in real time, identify the potential health risks in time, and prevent the occurrence of diseases by periodically collecting and analyzing the physiological data of the user. The system can provide customized health management suggestions according to the early warning of the health trend data, and enhance the forward-looking and preventive nature of the health monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 FIG. 1 is a structural diagram of the health assessment and intelligent management system based on wristband multi-modal sensing;

[0027] Figure 2 FIG. 2 is a flowchart of the health assessment and intelligent management system based on wristband multi-modal sensing;

[0028] Figure 3 FIG. 3 is a health level determination diagram based on the health assessment value. DETAILED DESCRIPTION

[0029] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0030] Please refer to Figures 1-3The embodiment of the present application provides a technical scheme: a health assessment and intelligent management system based on wristband multi-modal perception, comprising: a data acquisition and preprocessing module, configured to periodically acquire multi-modal wristband perception data, and perform time alignment, noise suppression, abnormality rejection, missing value completion and scale normalization processing on the acquired data to generate preprocessed multi-modal wristband perception data; a humidity influence correction module, configured to construct a reference value set based on the preprocessed multi-modal wristband perception data, evaluate the influence degree of humidity on the multi-modal wristband perception data, determine whether there is humidity interference in the current sampling, and correct the multi-modal wristband perception data with humidity interference by using a regression model; a health assessment and analysis module, configured to construct a perception feature data set based on the humidity corrected multi-modal wristband perception data, evaluate the health status of the user, determine the health level of the user according to the evaluation result, and generate a corresponding health level label; and a health intervention feedback module, configured to formulate a hierarchical health strategy report based on the health level label, continuously monitor the change of the health status of the user, dynamically adjust the health intervention plan combined with the intervention effect, and realize a health management closed loop.

[0031] As shown in Figure 2 The whole process is carried out around the continuous acquisition of multi-modal wristband perception data. First, multi-dimensional physiological signals are acquired under a fixed sampling period, and preprocessing operations such as time alignment, noise suppression, abnormality rejection and scale unification are sequentially performed to make the original signals have consistent features. Then, the humidity influence evaluation value is calculated based on the reference value set, and it is judged whether the signal of this period is disturbed by humidity according to the humidity influence evaluation value. If the humidity influence is within the threshold, the current period data is directly used; if the humidity deviation is significant, the humidity correction result is generated by calling the regression model based on the historical interference samples to ensure the data quality of the subsequent feature extraction link. Then, statistical features are extracted from the multi-modal data of continuous periods to form a feature data set for health analysis. By calculating the health-related quantitative value and comparing it with the multi-level health threshold, the health level of this period is obtained, and the hierarchical strategy content is generated accordingly. The system records the health changes in the set intervention period, calculates the intervention effect evaluation value, and updates the intervention plan combined with the feature changes in the period; at the same time, the corresponding health progress report is output to support the continuous health management process.

[0032] Specifically, the multi-modal wristband sensing data is periodically collected, and the collected data is time-aligned, noise-suppressed, abnormality-removed, missing value-completed and scale-normalized. The specific steps for generating the pre-processed multi-modal wristband sensing data are as follows: First, set a fixed-width time window as a sampling period, and periodically collect multi-modal wristband sensing data. The multi-modal wristband sensing data collected in each sampling period includes electromyography signals collected by a wristband surface electromyography sensor, electrodermal activity signals collected by a wristband skin conductance sensor, pulse wave signals collected by a wristband PPG sensor, wrist skin temperature recorded in real time by a wristband high-precision NTC thermistor, heart rate collected by a wristband heart rate sensor, respiration rate collected by a wristband respiration sensing unit, wristband humidity collected by a wristband humidity sensor, and skin resistance value collected by a wristband skin resistance sensor. The collection process uses an accurate clock synchronization device embedded in the wristband to ensure that the collection time points of all sensors in the wristband are consistent in each period. The collected multi-modal wristband sensing data is corrected synchronously. The least squares interpolation algorithm is used to calibrate the multi-source time series, so that the data of different sensors can be accurately corresponded at the same time point. This process calculates the deviation of each sensor data and uses interpolation method for smooth correction to ensure that the data of multiple signal sources have consistent time steps. Especially in the case of time offset of different sensors, the interpolation algorithm effectively aligns the data, avoiding the influence of time alignment error on the analysis result. A double exponential smoothing filter algorithm is used to smooth and suppress the sudden spikes and high-frequency disturbances in the multi-modal wristband sensing data. Double exponential smoothing filtering is a commonly used time series smoothing technique, especially suitable for signal data with large noise. In this step, the double exponential smoothing algorithm eliminates the noise signals with large fluctuations by weighted average of data, reducing the influence of instantaneous disturbance and high-frequency noise on data analysis. This makes the final data more stable and easy for subsequent processing. A local outlier factor algorithm is used to automatically identify and remove abnormal segments in the multi-modal wristband sensing data. The local outlier factor algorithm identifies outliers by evaluating the density difference of data points in the local area. This algorithm can automatically determine which data points deviate from the normal mode according to the local structure of the data, so it can effectively detect abnormal signals such as extreme data points caused by sensor failure, external interference or other irrelevant factors. After removing the abnormal data by the local outlier factor algorithm, the accuracy of the remaining data is greatly improved, which helps to improve the accuracy of subsequent analysis. For missing signal data, Lagrange interpolation method is used to reconstruct the missing values. Lagrange interpolation is a commonly used polynomial interpolation method that can estimate missing data values based on the surrounding known data points. Through this method, the continuity of the data can be ensured, and the influence of missing data on the analysis result can be avoided. The interpolation process uses the weighted average of adjacent data points to fill in the missing signals, thereby ensuring the integrity of the entire data set.All collected multi-modal wristband sensing data are scale-standardized and numerically unified using the Z-Score standardization algorithm. Z-Score standardization subtracts each data point from the mean of the dataset it belongs to and divides by the standard deviation, eliminating the influence of dimension differences between different sensors, allowing each signal measure to be compared under the same standard, and making subsequent calculation formulas expressed in dimensionless form.

[0033] In this embodiment, by performing precise time alignment, noise suppression, outlier rejection, and missing value completion on the multi-modal wristband sensing data, the accuracy and consistency of the data can be significantly improved. After synchronization correction and multiple algorithm optimization of each sensor data, the high quality and comparability of the signal are ensured, thereby providing a reliable data foundation for subsequent health assessment analysis. This data processing flow reduces the influence of external interference and data errors, making the final output of the health assessment result more accurate and reliable.

[0034] Specifically, the specific steps of constructing the reference value set based on the pre-processed multi-modal wristband sensing data are as follows: first, based on the pre-processed multi-modal wristband sensing data, arrange in the order of sampling time to ensure the time sequence of the data. Then, calculate the skin resistance difference value of adjacent sampling periods to analyze the change of skin resistance on the time axis. When the continuous N skin resistance difference values are lower than the preset difference threshold value, it indicates that the skin resistance change is small and the data tends to be stable; wherein the value range of N is 3 to 5; the difference threshold value is obtained based on the steady-state distribution characteristics of the skin resistance time series in the historical sampling period. Specifically: extract the skin resistance sequence collected continuously for multiple days, calculate the skin resistance difference value of adjacent sampling periods, and form a skin resistance difference value set; perform steady-state statistics based on the median absolute deviation on the set, take the median of the difference value set plus the amplification amount of the median absolute deviation as the stable boundary, and define the stable boundary as the difference threshold value. This threshold value can reflect the natural fluctuation range of the user's skin resistance under no interference, and is used to identify the stable physiological interval. On this basis, the mean values of each multi-modal wristband sensing data in the corresponding sampling period are calculated, including electromyographic signals, electrodermal activity signals, pulse wave signals, wrist skin temperature, heart rate, respiratory rate, wristband humidity, and skin resistance. The mean value can effectively smooth the instantaneous fluctuations in the data and avoid interference from abnormal values or noise in a short period of time; these mean values are used as the reference values of the corresponding data, and all reference values are collected together to form a complete reference value set, providing a stable and reliable reference for subsequent data analysis.

[0035] In this embodiment, through the process of constructing a reference value set based on the pre-processed multi-modal wristband sensing data, stable and representative reference data can be effectively provided for subsequent analysis. By calculating the difference of skin resistance and selecting the mean value of the last N sampling periods as the reference value when the change tends to be stable, the interference of short-term fluctuations and noise can be eliminated, ensuring the reliability and consistency of the data. Taking the mean steady state as the full-modal reference period further enhances the stability of the data, so that each multi-modal data can better reflect the user's long-term health status when analyzed, rather than being affected by instantaneous abnormalities, thereby providing a solid data foundation for subsequent health assessment and intervention strategy formulation.

[0036] Specifically, the specific steps of evaluating the degree of influence of humidity on the multi-modal wristband sensing data are as follows: first, the real-time multi-modal wristband sensing data is extracted, the absolute value of the difference between the skin electrical activity signal, the electromyographic signal and the pulse wave signal and the corresponding reference value is calculated respectively, and the absolute values of the three differences are summed to obtain a signal difference total term. This term can comprehensively reflect the difference between each physiological signal and its reference value, and capture the fluctuation of each signal caused by humidity change. In this way, the difference of each signal is fully evaluated. Next, the square of the difference between the wrist skin temperature and the skin resistance and the corresponding reference value is calculated respectively, and the two difference squares are added to one, and then the reciprocal is taken to obtain a temperature resistance correction term. This term can enhance the adaptability of temperature and resistance changes to humidity influence through nonlinear correction, and eliminate the possible mutual interference of the two. Especially when the environmental humidity increases, the skin surface is more likely to form a continuous water film, and the water penetrates into the stratum corneum to change the hydration degree of the stratum corneum, which increases the number of conductive paths on the skin surface and significantly improves the sensitivity of skin resistance to humidity change. The temperature resistance correction term accordingly compensates for the resistance fluctuation driven by humidity. By introducing the nonlinear characteristics of the reciprocal function, the response of temperature and resistance to humidity change is further adjusted to more accurately reflect its role in the overall data. Then, the absolute value of the difference between the wristband humidity and the corresponding reference value is added to one and then the natural logarithm is taken, and the obtained natural logarithm value is added to one to obtain a humidity change logarithmic term. Since the subjective perception of human body to external stimulus intensity usually conforms to the logarithmic relationship described by Weber-Fechner law in most scenarios, the use of logarithmic transformation can make the mapping of humidity change in the numerical space more close to the real change rate of physiological perception, and enhance the correspondence between the humidity change logarithmic term and the physiological state perception. Through the introduction of the natural logarithm function, this term can effectively compress the extreme value of humidity change, so that the influence of humidity on the signal presents a logarithmic decreasing effect, thereby making the influence of humidity fluctuation on the overall evaluation more smooth and conforming to the actual physiological change. Finally, the product of the signal difference total term and the temperature resistance correction term and the humidity logarithmic term are multiplied in turn to obtain the humidity influence evaluation value. In this way, the influence factors of each item of data are combined, and the use of product ensures that the multiple influences of humidity on different physiological signals are fully quantified, and unnecessary deviations are effectively reduced. In addition, the product method enhances the interaction between each factor in the mathematical model, ensuring that the evaluation value of humidity interference has higher accuracy and reliability.

[0037] wherein the specific calculation formula of the humidity influence evaluation value is:

[0038] ;

[0039] in the formula, represents the humidity influence evaluation value, represents the skin electrical activity signal, represents the myoelectric signal, represents the pulse wave signal, represents the wrist skin temperature, represents the skin resistance, represents the wristband humidity, represents the skin electric activity signal reference value, represents the myoelectric signal reference value, represents the pulse wave signal reference value, represents the skin temperature reference value, represents the skin resistance reference value, represents the wristband humidity reference value.

[0040] In this embodiment, by introducing a plurality of nonlinear functions and comprehensive calculation methods, the method effectively quantifies the influence of humidity on multi-modal wristband sensing data. The combination of signal difference sum term, temperature resistance correction term and humidity change logarithm term makes the influence of humidity on each physiological signal be evaluated in multiple dimensions and multiple levels, further enhancing the accuracy and robustness of the evaluation results. Using nonlinear adjustment functions such as inverse and logarithmic functions, the influence of extreme values is reasonably suppressed, ensuring data smoothness and stability during the evaluation process, thereby providing an accurate and adaptive evaluation method. This method not only effectively reduces the error caused by humidity fluctuations, but also improves the reliability of overall data analysis, providing a more stable and scientific basis for subsequent health evaluation.

[0041] Specifically, the specific steps of determining whether the current sampling exists humidity interference and correcting the multi-modal wristband sensing data with humidity interference by using the regression model are as follows: first, the humidity influence evaluation value and the humidity influence threshold value are compared in real time, when the humidity influence evaluation value is less than or equal to the humidity influence threshold value, it is determined that the current sampling is normal, and the data does not need to be further processed; when the humidity influence evaluation value is greater than the humidity influence threshold value, it is determined that the current sampling exists humidity interference, and the humidity correction process is entered. The humidity influence threshold value is automatically determined based on the historical humidity influence evaluation value sequence: first, the humidity influence evaluation values of the historical sampling period for consecutive days are extracted, and the humidity influence evaluation value set is constructed in chronological order; then, the steady-state segmentation scanning is performed on the set, the segments with stable skin resistance changes are identified, and the humidity influence evaluation values in these segments are taken as reference samples without obvious humidity interference; then, the median of the reference samples is calculated, and the median absolute deviation of the median is calculated; on this basis, the humidity influence threshold value is obtained by adding the amount of the median to the median absolute deviation. The humidity influence threshold value can represent the normal fluctuation range of the humidity influence evaluation value without humidity interference, and is used to determine whether the current sampling period exists humidity interference. By comparing the data in the historical sampling period, the interference mode of the humidity change on the skin resistance and other signals can be effectively identified. Based on the extreme gradient boosting regression algorithm, the regression model is constructed by learning these historical data, and the model will learn the relationship between the humidity change and the multi-modal wristband sensing data, so as to capture the regularity of the humidity influence. The reason for using the extreme gradient boosting regression algorithm is that the influence of humidity interference on the skin activity signal, the electromyography signal and the pulse wave signal presents obvious nonlinear characteristics, and is also accompanied by coupling relationship among multi-dimensional features. Traditional linear filtering methods relying on fixed filtering kernel cannot effectively eliminate this kind of structural interference, while the extreme gradient boosting regression algorithm can mine complex nonlinear mapping relationship through multiple rounds of residual fitting, and correct the humidity residual interference that cannot be removed by traditional filtering. The regression model can accurately simulate the influence function between humidity and physiological signals under complex humidity change conditions. After the training and optimization of the regression model are completed, the humidity influence evaluation value and the multi-modal wristband sensing data of the current sampling period are input into the regression model, the current sampling data are corrected by the model, and the multi-modal wristband sensing data corrected by the humidity are output; the corrected data can more truly reflect the physiological signals affected by the environmental humidity, thereby ensuring the accuracy and reliability of the data in the health assessment process and improving the accuracy of the final health state determination.

[0042] In this embodiment, the humidity interference is corrected by introducing the extreme gradient boosting regression algorithm, which can effectively eliminate the influence of humidity change on the multi-modal wristband sensing data, thereby significantly improving the accuracy and reliability of the data. This method establishes the mapping relationship between humidity and physiological signals through learning historical data, which can accurately simulate the interference of humidity on the signal, ensuring that the corrected data is more consistent with the actual physiological state. This correction mechanism can ensure the effectiveness of multi-modal wristband sensing data in complex humidity environments, thereby improving the accuracy and consistency of health assessment results.

[0043] Specifically, the specific steps of constructing the sensing feature dataset based on the humidity-corrected multi-modal wristband sensing data are as follows: based on the humidity-corrected multi-modal wristband sensing data, the mean and variance of each multi-modal wristband sensing data in the last M sampling periods are calculated, and the value of M is in the range of 3 to 10 sampling periods to ensure the timeliness and stability of the data. For each signal data, the mean and variance in the M periods are calculated to extract the central tendency and fluctuation amplitude of the data, providing stable feature information for subsequent health assessment. In addition, the maximum value of skin temperature and the minimum value of skin resistance are extracted from the humidity-corrected data. These two characteristic values are selected as key health assessment indicators, and their causal relationship is based on the following principles: the maximum value of skin temperature usually reflects the adaptability of the human body to external environmental changes, especially when the temperature fluctuates sharply, the extreme value of skin temperature can reflect the body's self-regulation ability; the minimum value of skin resistance is often related to the body's water balance and electrolyte balance, and a lower skin resistance value may indicate the body's sensitivity to environmental changes or potential physiological abnormalities. Based on the above multi-dimensional statistical indicators and extreme value characteristics, the mean, variance, maximum skin temperature, and minimum skin resistance of each multi-modal wristband sensing data are combined to construct the sensing feature dataset, so that the feature structure covers the central tendency, fluctuation characteristics, and extreme response characteristics, thereby more comprehensively representing the user's real-time physiological state and improving the accuracy and sensitivity of subsequent health assessment.

[0044] In this embodiment, by constructing the sensing feature dataset based on the humidity-corrected multi-modal wristband sensing data, this process can effectively capture the stability and fluctuation characteristics of each physiological signal, reducing the influence of external environmental interference factors on health assessment. By reasonably selecting the number of sampling periods, the timeliness and stability of the data are ensured, and by combining key physiological characteristics, the comprehensiveness and reliability of data analysis are enhanced, providing more accurate and comprehensive basic data support for subsequent health assessment.

[0045] Specifically, the specific steps of evaluating the health status of the user are as follows: extracting the perception feature dataset and the reference value set, sequentially adding the mean values of the electrodermal activity signal, the electromyogram signal and the pulse wave signal, and then multiplying the signal fusion coefficient to obtain a signal fusion term. The signal fusion coefficient is obtained by comprehensively analyzing the contribution of different signals to health evaluation, combining the fluctuation amplitudes of the electrodermal activity signal, the electromyogram signal and the pulse wave signal, and using a particle swarm optimization algorithm, and the value range is usually between 1.0 and 5.0. The signal fusion term reflects the comprehensive effect of the three physiological signals of electrodermal activity, electromyogram activity and pulse wave signal, and by introducing the signal fusion coefficient, the contributions of different signals in health evaluation are appropriately weighted, thereby improving the accuracy of comprehensive health evaluation. Next, the absolute value of the difference between the mean heart rate and the heart rate reference value is divided by the heart rate sensitivity coefficient, and the inverse of the obtained ratio is taken as the exponent, and the exponential function value of the natural constant e is obtained to obtain the heart rate deviation index term. The heart rate sensitivity coefficient is obtained by analyzing the mean and variance of the user's historical heart rate data, using a regression analysis algorithm, and the value range is generally between to The introduction of the nonlinear characteristics of the exponential function makes the deviation of the heart rate more significant in health evaluation, especially in the case of a large difference between the heart rate and the reference value, which can highlight its impact on health. Then, the square of the difference between the maximum skin temperature and the minimum skin resistance is calculated to obtain the temperature and humidity difference correction term. This term reflects the extreme changes between skin temperature and skin resistance, and when the difference between the two indicators is large, it may indicate that the user's body temperature regulation ability or skin condition is abnormal, so this correction term helps to identify physiological stress reactions caused by environmental changes. Next, the absolute value of the difference between the mean respiratory frequency and the respiratory frequency reference value is added to a natural logarithm, to obtain the respiratory frequency adjustment term. The introduction of the natural logarithm function smooths the difference between the respiratory frequency and the reference value, avoiding the excessive influence of extreme differences on the evaluation results, and also enhances the performance of the respiratory frequency deviation in the evaluation. Finally, the product of the signal fusion term and the heart rate deviation index term is divided by the sum of the constant one, the temperature and humidity difference correction term and the respiratory frequency adjustment term to obtain the health evaluation value. The comprehensive formula organically combines various feature data, improves the comprehensiveness and adaptability of health evaluation through weighting and nonlinear adjustment, so that the final evaluation value is more consistent with the user's real health status, and can more accurately reflect the user's physiological condition.

[0046] wherein the specific calculation formula of the health evaluation value is:

[0047] ;

[0048] In the formula, represents the health evaluation value, represents the mean value of the electrodermal activity signal, represents a mean value of the electromyography signal, represents a mean value of the pulse wave signal, represents a mean value of the heart rate, represents a reference value of the heart rate, represents a maximum value of the skin temperature, represents a minimum value of the skin resistance, represents a mean value of the respiration rate, represents a reference value of the respiration rate, represents a sensitivity coefficient of the heart rate, represents a fusion coefficient of the signals.

[0049] In the embodiment, Table 1 is a health assessment value data table listing health assessment values and related calculation data of five users. Specifically, for User 1: the average skin electrical activity signal is 7.53, the average electromyography signal is 15.65, the average pulse wave signal is 49.30, the average heart rate is 75.88, the heart rate reference value is 72.61, the maximum skin temperature is 36.98, the minimum skin resistance is 120.73, the average respiration rate is 14.12, the respiration rate reference value is 12.54, the heart rate sensitivity coefficient is 1.5, and the corresponding health assessment value is 11.68. For User 2: the average skin electrical activity signal is 8.10, the average electromyography signal is 13.78, the average pulse wave signal is 52.45, the average heart rate is 80.24, the heart rate reference value is 75.34, the maximum skin temperature is 37.25, the minimum skin resistance is 130.65, the average respiration rate is 15.32, the respiration rate reference value is 13.04, the heart rate sensitivity coefficient is 1.5, and the corresponding health assessment value is 3.25. For User 3: the average skin electrical activity signal is 6.85, the average electromyography signal is 14.32, the average pulse wave signal is 47.89, the average heart rate is 78.92, the heart rate reference value is 71.92, the maximum skin temperature is 36.77, the minimum skin resistance is 125.94, the average respiration rate is 14.55, the respiration rate reference value is 12.93, the heart rate sensitivity coefficient is 1.5, and the corresponding health assessment value is 1.01. For User 4: the average skin electrical activity signal is 7.92, the average electromyography signal is 16.44, the average pulse wave signal is 50.67, the average heart rate is 82.15, the heart rate reference value is 73.43, the maximum skin temperature is 37.19, the minimum skin resistance is 128.47, the average respiration rate is 15.10, the respiration rate reference value is 13.17, the heart rate sensitivity coefficient is 1.5, and the corresponding health assessment value is 1.01. For User 5: the average skin electrical activity signal is 6.45, the average electromyography signal is 17.11, the average pulse wave signal is 50.32, the average heart rate is 77.41, the heart rate reference value is 72.79, the maximum skin temperature is 36.86, the minimum skin resistance is 120.29, the average respiration rate is 14.90, the respiration rate reference value is 12.78, the heart rate sensitivity coefficient is 1.5, and the corresponding health assessment value is 4.88.

[0050] Table 1 Health assessment value data table

[0051]

[0052] As Figure 3The figure shows the health assessment values of five users and the determination results of their health levels. The column chart uses different colors to distinguish the health status: green column represents first-level health; yellow column represents second-level health; red column represents third-level health. The blue dotted line in the figure represents the first-level health threshold , , the orange dotted line represents the second-level health threshold , , and the red dotted line represents the third-level health threshold , which is valued at 5.0. As can be seen from the figure, the health assessment value of user 1 is greater than the second-level health threshold , marked as third-level health; the health assessment values of user 2 and user 5 are between the first-level health threshold and the second-level health threshold , marked as second-level health; the health assessment values of user 3 and user 4 are less than the first-level health threshold Figure 3 , marked as first-level health.

[0053] In this embodiment, by combining the weighted and nonlinear adjustment of multiple physiological signals, the accuracy and adaptability of health assessment are effectively improved. The introduction of signal fusion coefficient, heart rate sensitivity coefficient and temperature and humidity difference correction term makes the assessment more comprehensive, and can fully consider the effect of different physiological signals and reasonably adjust the influence of each data in health assessment. Especially in dealing with physiological abnormalities such as heart rate fluctuation, skin temperature and skin resistance change, the assessment process can more sensitively reflect the influence of these changes on the user's health status. In addition, by reasonably weighting the data features and introducing exponential, logarithmic and other nonlinear functions, the ability to cope with extreme data is enhanced, and the final assessment result is more accurate, which helps to fully reflect the user's health status.

[0054] Specifically, according to the assessment result, the health level of the user is determined, and the specific steps of generating the corresponding health level label are as follows: first, the health assessment value is compared with the first-level health threshold and the second-level health threshold respectively, the health level of the user is determined, and the corresponding health level label is generated: when ≤ , it is determined that the current health status is healthy, marked as first-level health; when < < , it is determined that the current health status is sub-health, marked as second-level health; when ≥ At this time, it is determined that the current state is unhealthy, and a health level of three is marked. This determination process accurately distinguishes different health levels according to multi-level health threshold criteria, reflecting the user's current overall health level, avoiding ambiguous health criteria. On this basis, the health level label not only provides intuitive feedback on the user's health status, but also helps health management personnel identify the user's health risks, and then tailor health recommendations for the user. By structurally storing the health level label, health assessment value, perception feature data set, and corresponding multi-modal wristband perception data, the user's health information can be effectively integrated to form a health assessment data set. In this process, the data is stored in a standardized manner to ensure the uniformity and standardization of the data, which provides a reliable foundation for subsequent health analysis, trend prediction, and decision support. This data storage method allows real-time updates of health assessments, and also facilitates long-term health monitoring, ensuring data traceability and effectiveness.

[0055] In this embodiment, by comparing the health assessment value with the health threshold, the health level of the user can be accurately determined, and the corresponding health label can be generated. This process ensures the high accuracy and operability of health assessment, and provides comprehensive support for subsequent health management by updating the health status in real time and storing related data. The classification of health levels not only helps users understand their own health status, but also lays the foundation for personalized health recommendations. In addition, the structured storage of health assessment data improves the traceability and effectiveness of the data, providing a reliable basis for long-term health monitoring and health management decisions.

[0056] Specifically, the health assessment data set is extracted, and a graded health strategy report is generated based on the health level label: when the health level is one, the health assessment data set is extracted to generate a health maintenance report, which includes suggestions for maintaining the current health status, focusing on maintaining healthy eating and exercise habits, ensuring adequate rest time, and regularly checking health to ensure the continuous stability of the health status. When the health level is two, the health assessment data set is extracted and a sub-health adjustment report is generated, which includes suggestions to increase exercise, set reasonable work and rest time, manage stress and emotions, improve diet structure, and avoid excessive fatigue. In addition, combined with the mobile terminal push function, a health intervention plan is provided to ensure that the user effectively intervenes according to the suggestions. When the health level is three, the health assessment data set is extracted and a health intervention report is generated, and a warning is issued to the user to conduct a comprehensive physical examination, and personalized diet and exercise programs are provided for the user based on the physical examination results. At the same time, the report will prompt the user to pay attention to potential health problems and suggest timely intervention measures.

[0057] In this embodiment, by generating a hierarchical health strategy report based on the health level label, personalized health recommendations and interventions can be provided according to different health states. The report content provides practical adjustment schemes for each health level, ensuring that users receive professional guidance in health maintenance, sub-health adjustment, and health intervention. Through intelligent health data analysis and report generation, the accuracy and relevance of health management are improved, helping users make scientific health decisions in different health states and ensuring the effectiveness and sustainability of the health intervention process.

[0058] Specifically, the specific steps of the health management closed loop are as follows: a week is set as an intervention period, the health assessment values before and after each intervention period are extracted, the difference between the health assessment value after intervention and the health assessment value before intervention is divided by the health assessment value before intervention to obtain the health assessment change ratio, which represents the relative change amplitude of the health assessment value in a single intervention period; the difference between the health assessment value after intervention and the first health threshold is multiplied by the health adjustment coefficient, the inverse of the product is taken as the exponent, the exponential function value of the natural constant e is calculated, and then the inverse of the obtained exponential function value is added to one, and the obtained inverse value is subtracted by the constant one to obtain the health improvement factor. The combination of the exponential function and the inverse function constitutes a monotonic change mapping, and the increase of the health assessment value after intervention relative to the first health threshold makes the exponential term decrease and the health improvement factor increase accordingly. The non-linear mapping is used to suppress small fluctuations and highlight the sensitivity of changes across threshold intervals; the health adjustment coefficient is obtained by least squares regression algorithm fitting based on the health assessment values before and after intervention in the historical intervention period and the corresponding health assessment change ratio, and the value range is 0.5 to 3.0; the health assessment change ratio is multiplied by the health improvement factor to obtain the intervention effect evaluation value, which is used to represent the comprehensive quantitative result of the relative improvement amplitude and the threshold neighborhood improvement degree; the health intervention plan is dynamically adjusted in combination with the intervention effect evaluation value and the health assessment data set in the current intervention period, and the period health progress report is generated to realize the health management closed loop; wherein the strategy updating rule is to increase the intervention intensity by one level when the intervention effect evaluation value increases continuously for K intervention periods and the current health level remains unchanged, to decrease the intervention intensity by one level when the intervention effect evaluation value decreases continuously for K intervention periods, and to decrease the intervention intensity by one level when the change amplitude of the intervention effect evaluation value in the continuous K intervention periods is lower than the stability threshold and the health level is changed from the second health to the first health; wherein K is 2 to 4, and the intervention intensity level refers to the intervention intensity corresponding to different health levels when the health intervention plan is developed, including multiple intervention schemes from light to heavy. If the highest intervention intensity in the current level group is reached during the adjustment process, the intervention intensity will not be further increased, and the current intervention intensity will be maintained. If the lowest intervention intensity is reached, the intervention intensity will not be further decreased, and the original intervention intensity will be maintained.

[0059] The specific calculation formula of the intervention effect evaluation value is as follows: ; In the formula, represents the intervention effect evaluation value, represents the health evaluation value before intervention, represents the health evaluation value after intervention, represents the first health threshold value, represents the health adjustment coefficient.

[0060] In the embodiment, by introducing the periodic change of the health evaluation value into the unified quantification link, and combining the health evaluation change ratio, the health improvement factor, the intervention effect evaluation value and the linkage criterion of the health evaluation data set, a continuous intervention regulation mechanism based on data driving can be formed, so that the adjustment of the health intervention plan has a calculable basis and a traceable logic, the matching degree of the intervention rhythm and the real state change of the user is improved, the risk of frequent switching of strategies caused by short-term fluctuations is reduced, the stability and consistency of the intervention intensity adjustment are enhanced, and therefore more continuous quantification support is provided for the periodic health progress report.

[0061] It should be noted that, in this document, the terms such as first and second are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment.

[0062] The preferred embodiments of the application disclosed above are only used to help explain the application. The preferred embodiments do not describe all the details, nor limit the application to the specific embodiments described. Obviously, many modifications and variations can be made according to the content of the present specification. The present specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the application, so that those skilled in the art can well understand and utilize the application. The application is limited only by the claims and their full scope and equivalents.

Claims

1. A health assessment and intelligent management system based on wristband multimodal sensing, characterized in that, Includes the following modules: The data acquisition and preprocessing module is used to periodically acquire multimodal wristband sensing data and perform time alignment, noise suppression, anomaly removal, missing value completion and scale normalization on the acquired data to generate preprocessed multimodal wristband sensing data. The humidity impact correction module is used to construct a set of benchmark values ​​based on the preprocessed multimodal wristband sensing data, assess the degree of influence of humidity on the multimodal wristband sensing data, determine whether there is humidity interference in the current sampling, and use a regression model to correct the multimodal wristband sensing data with humidity interference. The health assessment and analysis module is used to construct a sensing feature dataset based on humidity-corrected multimodal wristband sensing data, assess the user's health status, determine the user's health level based on the assessment results, and generate corresponding health level labels. The health intervention feedback module is used to develop tiered health strategy reports based on health level labels, continuously monitor changes in users' health status, and dynamically adjust health intervention plans based on intervention effects to achieve a closed loop in health management.

2. The health assessment and intelligent management system based on wristband multimodal perception according to claim 1, characterized in that: The specific steps for periodically collecting multimodal wristband sensing data and performing time alignment, noise suppression, anomaly removal, missing value completion, and scale normalization on the collected data to generate preprocessed multimodal wristband sensing data are as follows: A fixed-width time window is set as one sampling period to periodically collect multimodal wristband sensing data, which includes electromyography signals, skin electrical activity signals, pulse wave signals, wrist skin temperature, heart rate, respiratory rate, wristband humidity, and skin resistance. For the acquired multimodal wristband sensing data, the multi-source time series are synchronously corrected based on the least squares interpolation algorithm, and the sudden spikes and high-frequency disturbances in the multimodal wristband sensing data are smoothed and suppressed using the double exponential smoothing filter algorithm. Abnormal segments in the multimodal wristband sensing data are automatically identified and removed using the local outlier algorithm, and the continuity of short-term missing intervals is reconstructed using the Lagrange interpolation method. Finally, the multimodal wristband sensing data is scaled and numerically unified using the Z-Score normalization algorithm.

3. The health assessment and intelligent management system based on wristband multimodal perception according to claim 1, characterized in that: The specific steps for constructing the baseline value set based on the preprocessed multimodal wristband sensing data are as follows: Based on the preprocessed multimodal wristband sensing data, the data are arranged in the order of sampling time. The skin resistance difference between adjacent sampling periods is calculated. When N consecutive skin resistance differences are lower than the difference threshold, the mean of each multimodal wristband sensing data in the corresponding sampling period is calculated and recorded as the benchmark value of the corresponding data. Together, they form a benchmark value set.

4. The health assessment and intelligent management system based on wristband multimodal perception according to claim 1, characterized in that: The specific steps for assessing the impact of humidity on multimodal wristband sensing data are as follows: Real-time multimodal wristband sensing data is extracted, and the absolute values ​​of the differences between the skin conductance signal, electromyography signal, and pulse wave signal and the corresponding baseline values ​​are calculated respectively. The three absolute values ​​of the differences are summed to obtain the signal difference sum term. Calculate the squared differences between the wrist skin temperature and skin resistance and the corresponding baseline values, add the squared differences together, add one, and take the reciprocal to obtain the temperature resistance correction term; add one to the absolute value of the difference between the wristband humidity and the corresponding baseline value, take the natural logarithm, and add one to the natural logarithm to obtain the humidity change logarithm term; multiply the product of the signal difference sum term and the temperature resistance correction term and the humidity logarithm term in sequence to obtain the humidity impact assessment value.

5. The health assessment and intelligent management system based on wristband multimodal perception according to claim 4, characterized in that: The specific steps for determining whether humidity interference exists in the current sample and using a regression model to correct the multimodal wristband sensing data with humidity interference are as follows: The system compares the humidity impact assessment value and the humidity impact threshold in real time. When the humidity impact assessment value is less than or equal to the humidity impact threshold, the current sampling is considered normal and no action is taken. When the humidity impact assessment value is greater than the humidity impact threshold, the current sampling is considered to have humidity interference, and the humidity correction process begins. This involves extracting multimodal wristband sensing data and humidity impact assessment values ​​from historical sampling periods where the skin resistance difference is greater than the difference threshold. A regression model is constructed based on the extreme gradient boosting regression algorithm, and the multimodal wristband sensing data from historical periods and the corresponding benchmark value set are input into the regression model for training and optimization. Finally, the humidity impact assessment value and multimodal wristband sensing data from the current sampling period are input into the trained regression model, and the humidity-corrected multimodal wristband sensing data is output.

6. The health assessment and intelligent management system based on wristband multimodal perception according to claim 1, characterized in that: The specific steps for constructing the sensing feature dataset based on the humidity-corrected multimodal wristband sensing data are as follows: Based on the humidity-corrected multimodal wristband sensing data, the mean and variance of each multimodal wristband sensing data in the most recent M sampling periods are calculated, and the maximum skin temperature and minimum skin resistance are extracted to construct a sensing feature dataset.

7. The health assessment and intelligent management system based on wristband multimodal perception according to claim 1, characterized in that: The specific steps for assessing a user's health status are as follows: Extract the sensory feature dataset and the baseline value set. Sum the mean values ​​of skin electrokinetic activity, electromyography, and pulse wave signals sequentially, then multiply by the signal fusion coefficient to obtain the signal fusion term. Divide the absolute value of the difference between the mean heart rate and the baseline heart rate value by the heart rate sensitivity coefficient, and use the negative of the resulting ratio as the exponent to take the exponential function value of the natural constant e to obtain the heart rate deviation exponent term. Calculate the square of the difference between the maximum skin temperature and the minimum skin resistance value to obtain the temperature and humidity difference correction term. Add one to the absolute value of the difference between the mean respiratory rate and the baseline respiratory rate value, and take the natural logarithm to obtain the respiratory rate adjustment term. Multiply the signal fusion term and the heart rate deviation exponent term, and divide by the sum of the constant one, the temperature and humidity difference correction term, and the respiratory rate adjustment term to obtain the health assessment value.

8. The health assessment and intelligent management system based on wristband multimodal perception according to claim 7, characterized in that: The specific steps for determining a user's health level based on the assessment results and generating a corresponding health level label are as follows: Health assessment values Respectively compared with the first-level health threshold and secondary health threshold Real-time comparison is performed to determine the user's health level and generate a corresponding health level label: when ≤ When the current state is determined to be healthy, it is marked as Level 1 healthy; when < < At that time, the person is determined to be in a sub-healthy state and marked as Level 2 healthy; when ≥ When the current state is determined to be unhealthy, it is marked as Level 3 healthy; The health level labels, health assessment values, sensory feature datasets, and corresponding multimodal wristband sensory data are structured and stored to construct a health assessment dataset.

9. The health assessment and intelligent management system based on wristband multimodal perception according to claim 8, characterized in that: The specific steps for developing a tiered health strategy report based on health level labels are as follows: Extract the health assessment dataset and generate a tiered health strategy report based on health level labels: When the label is Level 1 Health, extract the health assessment dataset to generate a health maintenance report; When the label is Level 2 Health, extract the health assessment dataset and generate a sub-health adjustment report, and push the health intervention plan through the mobile terminal; When the label is Level 3 Health, extract the health assessment dataset, generate a health intervention report, and issue a physical examination warning.

10. The health assessment and intelligent management system based on wristband multimodal perception according to claim 1, characterized in that: The specific steps for continuously monitoring changes in the user's health status and dynamically adjusting the health intervention plan based on the intervention effect to achieve a closed loop in health management are as follows: One week is defined as an intervention cycle. Health assessment values ​​before and after each intervention cycle are extracted. The difference between the post-intervention health assessment value and the pre-intervention health assessment value is divided by the pre-intervention health assessment value to obtain the health assessment change ratio. The difference between the post-intervention health assessment value and the first-level health threshold is multiplied by the health adjustment coefficient. The exponential function value of the natural constant e is calculated using the negative of the product as the exponent. One is added to the exponential function value, and the reciprocal is taken. The reciprocal value is then subtracted from the constant to obtain the health improvement factor. The health assessment change ratio is multiplied by the health improvement factor to obtain the intervention effect assessment value. By combining the intervention effect assessment value with the health assessment dataset within the current intervention cycle, the health intervention plan is dynamically adjusted, and a periodic health progress report is generated, thus achieving a closed loop in health management.

Images