Physical health personalized exercise prescription generation method and device based on biofeedback

Abstract

The invention relates to a physical health personalized exercise prescription generation method and device based on biofeedback. The method comprises the steps that biological signals such as the heart rate, the blood pressure and the breathing frequency of a user during exercise are obtained, and a pre-trained signal denoising model is adopted for filtering processing to form a biological signal set; performing multi-dimensional recognition and pattern classification on the biological signal set by using a support vector machine algorithm, and extracting a multi-dimensional feature data set containing the cardiopulmonary function intensity, the metabolic efficiency level and the motion stress response; and inputting the data set into a pre-trained physique evaluation classifier to output a quantitative index set, matching the basic exercise parameters based on a mapping relationship between the quantitative index set and a preset physique health standard library, and generating an initial exercise prescription containing the exercise type, the intensity threshold, the duration and the intermittent period parameter. According to the method, objective quantitative generation of the exercise prescription is realized, and the personalized health intervention precision is improved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent health management technology, and in particular relates to a method and device for generating personalized exercise prescriptions for physical health based on biofeedback. Background Technology

[0002] With the rapid development of intelligent health management technology, biofeedback monitoring technology based on wearable devices has emerged. This technology can dynamically track a user's exercise status by collecting physiological signals such as heart rate and blood pressure in real time. This leads to the current mainstream methods of exercise prescription formulation: In traditional methods, exercise prescriptions usually rely on physician experience or standardized physical fitness test results (such as VO2max test), combined with static health questionnaire data to generate a fixed-intensity exercise plan; some intelligent systems recommend training plans with preset intensity ranges through statistical analysis of historical exercise data. However, the current methods of generating exercise prescriptions have significant drawbacks: strong reliance on subjective experience, manually formulated prescriptions are easily affected by physician cognitive biases and cannot quantitatively match real-time changes in the user's physical condition (such as fluctuations in cardiopulmonary function); insufficient utilization of biosignals, existing systems mostly use a single heart rate threshold to control intensity, ignoring the correlation of multi-dimensional signals such as blood pressure and respiratory rate, leading to inaccurate assessment of metabolic efficiency and exercise stress response; lack of dynamic adaptability, traditional methods cannot optimize prescription parameters in real time based on the user's historical exercise characteristics (such as duration stability and recovery efficiency) and changes in health needs, resulting in a lag in health risk warnings. Summary of the Invention

[0003] Therefore, it is necessary to provide a method and device for generating personalized exercise prescriptions for physical health based on biofeedback that can solve the above problems.

[0004] Firstly, this application provides a method for generating personalized exercise prescriptions for physical health based on biofeedback, including:

[0005] The system acquires the user's biological signals during exercise and filters them using a pre-established signal denoising model to form a set of biological signals, including heart rate, blood pressure, and respiratory rate.

[0006] The support vector machine algorithm was used to perform multi-dimensional identification and pattern classification on the biological signal set, and to extract a multi-dimensional feature dataset containing cardiopulmonary function intensity, metabolic efficiency level and exercise stress response.

[0007] Input a multi-dimensional feature dataset into a pre-trained physical fitness assessment classifier and output a set of quantitative indicators that characterize the user's physical health status.

[0008] Based on the mapping relationship between the quantitative indicator set and the preset physical health standard library, the basic exercise parameters are matched using the preset exercise prescription rule library to generate an initial exercise prescription; the initial exercise prescription includes exercise type, intensity threshold, duration and interval cycle parameters.

[0009] In one embodiment, after generating the initial exercise prescription, the method further includes:

[0010] Acquire users' historical exercise biosignals, historical exercise data sequences, and individual health requirement parameters;

[0011] Based on historical motion data sequences, time-series alignment processing is performed on historical motion biosignals;

[0012] Feature extraction is performed on time-aligned historical motion biosignals to calculate motion intensity distribution index, duration stability index, and recovery cycle efficiency index.

[0013] Input individual health needs parameters into a preset needs analysis model to generate an exercise preference vector containing exercise type identifiers, tolerable intensity ranges, and health intervention priority parameters.

[0014] Based on exercise intensity distribution indicators, duration stability indicators, recovery cycle efficiency indicators, and exercise preference vectors, a multi-dimensional user exercise profile is constructed.

[0015] Based on the user's exercise profile, a multi-objective optimization algorithm is used to adjust the exercise type, intensity threshold, duration, and interval cycle parameters in the initial exercise prescription to generate the final exercise prescription.

[0016] In one embodiment, feature extraction is performed on the time-aligned historical motion biosignals, and motion intensity distribution index, duration stability index, and recovery cycle efficiency index are calculated. The specific steps are as follows:

[0017] The heart rate data in historical exercise biosignals are divided into intervals using a preset heart rate intensity grading threshold. The cumulative exercise duration is then statistically analyzed within the low-intensity interval (heart rate below 100 bpm), the medium-intensity interval (100-140 bpm), and the high-intensity interval (heart rate above 140 bpm), generating an exercise intensity distribution index.

[0018] Based on the standard deviation algorithm, the effective duration sequence of N consecutive movements in historical motion biosignals is processed, and the duration stability index is calculated according to the following formula:

[0019]

[0020] Among them, t i The duration of a single exercise session. σ is the average duration of the exercise, N is the number of consecutive exercises, and σ is the mean duration of the exercise.T The quantification value of the fluctuation in motion duration stability;

[0021] Based on time-series blood pressure data from historical exercise biosignals, the mean rate of blood pressure change within each recovery cycle is calculated using the following formula to obtain the recovery cycle efficiency index:

[0022]

[0023] Where R is the recovery cycle efficiency, K is the total number of data points within the recovery cycle, and P... k Let t be the blood pressure value at the k-th data point, and Δt be the preset time interval.

[0024] In one embodiment, a multi-dimensional user motion profile is constructed based on motion intensity distribution indicators, duration stability indicators, recovery cycle efficiency indicators, and motion preference vectors, including:

[0025] The exercise intensity distribution index, duration stability index, and recovery cycle efficiency index were standardized.

[0026] The standardized exercise intensity distribution index was mapped to the endurance dimension score, the standardized duration stability index was mapped to the exercise compliance dimension score, and the standardized recovery cycle efficiency index was mapped to the physical recovery dimension score.

[0027] Based on the health intervention priority parameter in the exercise preference vector, weight coefficients are assigned to the endurance dimension score, exercise compliance dimension score, and physical recovery dimension score. The weighted dimension scores are then fused with the exercise type identifier to generate a three-dimensional exercise profile vector as the user's exercise profile.

[0028] In one embodiment, based on the user's motion profile, a multi-objective optimization algorithm is used to adjust the motion parameters in the initial exercise prescription to generate a final exercise prescription, including:

[0029] Based on the motion type identifier in the user's motion profile, determine the set of motion types that the user can adapt to;

[0030] The first optimization objective is to maximize exercise adaptability based on the endurance dimension score in the user's exercise profile; the second optimization objective is to maximize the sustainability of the prescription based on the exercise compliance dimension score; and the third optimization objective is to minimize health risks based on the physical recovery dimension score.

[0031] A multi-objective optimization algorithm is used to optimize the first, second, and third optimization objectives to obtain an optimized set of motion parameters.

[0032] Based on the matching results between the adaptable set of exercise types and the optimized set of exercise parameters, the exercise type, exercise intensity, duration and interval cycle parameters in the initial exercise prescription are updated to generate the final exercise prescription.

[0033] In one embodiment, the method further includes:

[0034] Based on historical exercise data sequences, health trend prediction indicators are calculated according to the rate of change of endurance, exercise compliance, and physical recovery scores in user exercise profiles.

[0035] Input health trend prediction indicators into a pre-trained health risk prediction model, and output the probability of health risks and the offset of exercise capacity within a future preset period.

[0036] Based on the probability of health risks and the offset of exercise capacity, exercise guidance suggestions are generated, which include abnormal state warnings and exercise parameter adjustment strategies.

[0037] In one embodiment, the method further includes:

[0038] Based on the endurance, exercise compliance, and physical recovery scores in the user's exercise profile, a visual score map is generated using a radial basis function interpolation algorithm.

[0039] Based on the exercise type, intensity threshold, duration, and rest interval parameters in the final exercise prescription, an exercise prescription execution heatmap is generated.

[0040] Based on health trend prediction indicators and health risk probabilities, a time-series early warning dashboard for the evolution of future health status is constructed.

[0041] The interactive interface synchronously displays a visual scoring chart, a heatmap of exercise prescription execution, and a time-series early warning dashboard.

[0042] Secondly, this application also provides a device for generating personalized exercise prescriptions for physical health based on biofeedback, comprising:

[0043] The signal acquisition and processing module is used to acquire the user's biological signals during exercise and to filter them using a pre-established signal denoising model to form a set of biological signals, including heart rate, blood pressure, and respiratory rate.

[0044] The multi-dimensional feature recognition module is used to perform multi-dimensional recognition and pattern classification on biological signal sets using the support vector machine algorithm, and extract multi-dimensional feature datasets including cardiopulmonary function intensity, metabolic efficiency level and exercise stress response.

[0045] The physical fitness assessment quantification module is used to input multi-dimensional feature datasets into a pre-trained physical fitness assessment classifier and output a set of quantitative indicators that characterize the user's physical health status.

[0046] The exercise prescription generation module is used to generate an initial exercise prescription by matching basic exercise parameters with a preset exercise prescription rule library based on the mapping relationship between a set of quantitative indicators and a preset physical health standard library. The initial exercise prescription includes exercise type, intensity threshold, duration and interval cycle parameters.

[0047] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method for generating personalized exercise prescriptions for physical health based on biofeedback.

[0048] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described method for generating personalized exercise prescriptions for physical health based on biofeedback.

[0049] The aforementioned method, device, computer equipment, and storage medium for generating personalized exercise prescriptions based on biofeedback for physical health acquire biosignals (including heart rate, blood pressure, and respiratory rate) during user exercise and filter them using a signal denoising model to form a high-quality biosignal set. Addressing the issue of insufficient utilization of biosignals, the method integrates multi-dimensional signals to avoid inaccurate assessments caused by a single dimension. It utilizes a support vector machine algorithm to perform multi-dimensional recognition and pattern classification on the biosignal set, extracting feature datasets of cardiopulmonary function intensity, metabolic efficiency level, and exercise stress response, enhancing the comprehensiveness and objectivity of signal analysis and eliminating cognitive biases caused by reliance on physician subjective experience. The multi-dimensional feature dataset is input into a pre-trained physical fitness assessment classifier, outputting a quantitative index set to achieve an objective quantitative assessment of physical health status, overcoming the limitation of manual prescriptions in matching real-time changes in user physical fitness. Based on the mapping relationship between the quantitative index set and a preset physical health standard library, basic exercise parameters (such as exercise type, intensity threshold, duration, and interval parameters) are matched to generate an initial exercise prescription, achieving dynamic and quantitative generation of exercise prescriptions. This improves the accuracy of personalized health intervention and indirectly alleviates the problem of delayed risk warnings caused by a lack of dynamic adaptability. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a flowchart of a method for generating personalized exercise prescriptions for physical health based on biofeedback according to the present invention.

[0052] Figure 2 This is a structural diagram of a personalized exercise prescription generation device for physical health based on biofeedback according to the present invention. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0054] In one embodiment, such as Figure 1 As shown, a method for generating personalized exercise prescriptions for physical health based on biofeedback is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, or to a system including both a terminal and a server, and is implemented through the interaction between the terminal and the server. In the terminal implementation environment, the terminal device, such as a smartphone or smartwatch, is equipped with a signal acquisition module (e.g., an integrated sensor) to acquire real-time biosignals such as heart rate, blood pressure, and respiratory rate during the user's exercise process, and performs preliminary filtering to form a biosignal set. In the server implementation environment, the cloud server runs a support vector machine algorithm, a pre-trained physical fitness assessment classifier, and a multi-objective optimization algorithm to process the biosignal set or feature dataset received from the terminal, generating a quantitative index set and exercise prescription parameters. Under the user's exercise monitoring needs, the terminal continuously collects biosignals and transmits them to the server via a wireless network. After the server performs multi-dimensional feature extraction, physical fitness assessment, and prescription generation, it returns an initial or final exercise prescription containing parameters such as exercise type, intensity threshold, duration, and interval cycle. The terminal displays the prescription, a visual scoring graph, and a time-series warning dashboard through an interactive interface, realizing closed-loop feedback of personalized health intervention and real-time risk warning. In this embodiment, the method includes the following steps:

[0055] S01, acquire the user's biological signals during exercise, and use a pre-established signal denoising model for filtering to form a set of biological signals; the biological signals include heart rate, blood pressure and respiratory rate.

[0056] Among them, biosignals (motor physiological parameters collected in real time by wearable sensors, including heart rate (bpm), blood pressure (mmHg), and respiratory rate (breaths / minute), which constitute the original data source reflecting the user's exercise stress state); and a signal denoising model (a digital filter trained on historical noise samples, employing adaptive filtering algorithms (such as wavelet thresholding combined with Kalman filtering) to suppress motion artifacts, with its core filtering function satisfying...) Where S x (f) is the power spectrum of the effective physiological signal, S n (f) represents the power spectrum of environmental noise. Biosignals during exercise can be synchronously acquired using wearable sensors (such as an array of sensors consisting of ECG patches, photoplethysmography, and piezoelectric breathing bands). The heart rate data is filtered using a 0.5-5Hz bandpass filter to preserve sinus rhythm characteristics, and the respiratory signal is truncated in the 0.1-0.5Hz frequency band to eliminate motion interference. At the same time, a dynamic noise template is constructed based on accelerometer data to cancel motion artifacts, and finally a set of biosignals that meets the signal-to-noise ratio threshold (SNR≥15dB) is generated.

[0057] S02 utilizes the support vector machine algorithm to perform multi-dimensional identification and pattern classification on biological signal sets, extracting multi-dimensional feature datasets containing cardiopulmonary function strength, metabolic efficiency level, and exercise stress response.

[0058] Among them, the Support Vector Machine (SVM) algorithm (a supervised machine learning model based on statistical learning theory, which uses a predefined kernel function (such as radial basis function) to map the biological signal set to a high-dimensional feature space to handle nonlinear classification problems); the biological signal set (a structured time-series data set formed by filtering, including synchronously acquired heart rate, blood pressure, and respiratory rate sequences); cardiopulmonary function intensity (a comprehensive quantitative index reflecting cardiac output and pulmonary ventilation efficiency); metabolic efficiency level (characterizing the rate of energy conversion per unit time); and exercise stress response (a quantitative value of the degree of activation of the sympathetic nervous system in response to exercise load). The biological signal set can be input into a pre-trained SVM classifier, and pattern recognition of the time-series data can be performed through multi-class classification strategies (such as one-to-many or one-to-one methods). i x j )=exp(-γ||x i -x j || 2 Calculate feature similarity, where x iLet be the i-th biological signal sample vector, and γ be the kernel parameter. The signal is divided into cardiopulmonary function intensity levels (e.g., low, medium, high intensity ranges), metabolic efficiency level ranges (e.g., basal metabolic rate to peak metabolic rate), and exercise stress response patterns (e.g., acute stress or chronic adaptation) using hyperplane partitioning. Key feature vectors are extracted, including the peak heart rate variability (HRV) of cardiopulmonary function intensity. peak oxygen uptake slope at metabolic efficiency level Cort, the cortisol equivalent value in the exercise stress response eq This ultimately forms a multidimensional feature dataset FeatureSet = {F} cardio ,F meta ,F stress},in, Dimensions characterizing cardiopulmonary function strength (such as heart rate variability, HRV) peak ), Dimensions characterizing metabolic efficiency levels (such as oxygen uptake slope) ), Dimensions characterizing exercise stress response (such as cortisol equivalent Cort) eq ).

[0059] S03 inputs the multi-dimensional feature dataset into the pre-trained physical fitness assessment classifier and outputs a set of quantitative indicators representing the user's physical health status.

[0060] The pre-trained physical fitness assessment classifier (an ensemble learning model built on the XGBoost framework, trained using large-scale physical health data) can be input into the classifier after slicing the multidimensional feature dataset by time window. The contribution weight of each dimension is calculated through a feature importance weighting mechanism, with the central lung function intensity weight ω being a key component. c It can be dynamically adjusted by the difference between maximum heart rate and resting heart rate, with metabolic efficiency weight ω. m Positively correlated with the area under the oxygen uptake curve, the exercise stress response weight ω s Based on the calibration of the rate of blood pressure decrease during the recovery period, a set of quantitative indicators is output. in, Includes the equivalent value of maximum oxygen uptake (unit: mL / kg / min), Q MET Includes metabolic equivalent levels (1-12), Q HRR Includes heart rate recovery rate (unit: bpm / min), and this dataset can fully characterize the core parameters of a user's physical health status.

[0061] S04. Based on the mapping relationship between the quantitative indicator set and the preset physical health standard library, the basic exercise parameters are matched using the preset exercise prescription rule library to generate an initial exercise prescription. The initial exercise prescription includes exercise type, intensity threshold, duration and interval cycle parameters.

[0062] The system includes a pre-built physical fitness standard library (a pre-established database storing standardized parameters based on age, gender, and health grouping, such as cardiopulmonary function thresholds and metabolic baseline ranges; its mapping relationship is defined by the function Map(·), satisfying Map(QuantSet)→HealthLevel, where HealthLevel is a health level identifier, such as unhealthy, sub-healthy, and healthy). It also includes a pre-built exercise prescription rule library (a rule engine built using decision trees and fuzzy logic, containing basic exercise parameter matching rules; exercise types include aerobic, strength, or flexibility training; intensity thresholds are defined as heart rate ranges (e.g., 50-85% of maximum heart rate); duration is in minutes; and interval parameters refer to rest time (seconds) between sets). The QuantSet can be aligned with the physical fitness standard library using the mapping function Map(·) to generate the user's health level HealthLevel. Based on the pre-built exercise prescription rule library, exercise parameters are dynamically matched using HealthLevel, such as when... And Q MET When the intensity is ≤5, a moderate-intensity aerobic exercise prescription is triggered, with the intensity threshold set to 120-140 bpm, duration of 30 minutes, and rest interval of 60 seconds. An initial exercise prescription is generated, and through a rule-driven automatic matching mechanism, the subjective bias of traditional human experience is eliminated, and the objective quantitative generation of the exercise prescription is achieved.

[0063] The aforementioned method for generating personalized exercise prescriptions for physical health based on biofeedback acquires multidimensional biosignals such as heart rate, blood pressure, and respiratory rate during exercise. A pre-trained signal denoising model (e.g., wavelet thresholding combined with Kalman filtering) is used for filtering to eliminate motion artifacts and environmental noise, forming a biosignal set that meets the signal-to-noise ratio threshold (SNR≥15dB). By integrating multidimensional physiological parameters and improving data quality, it addresses the problem of inaccurate assessments caused by insufficient utilization of biosignals and avoids the deficiency of ignoring the correlation between blood pressure and respiratory rate when controlling intensity with a single heart rate threshold. Support vector machine algorithms (e.g., radial basis function-based classifiers) are used to perform multidimensional recognition and pattern classification on the biosignal set, extracting feature datasets of cardiopulmonary function intensity (e.g., peak heart rate coefficient of variation), metabolic efficiency level (e.g., oxygen uptake slope), and exercise stress response (e.g., cortisol equivalent value). This is then processed by an algorithm-driven approach. Nonlinear mapping enables multidimensional signal collaborative analysis, eliminating the subjective dependence and cognitive bias of traditional human experience. Multidimensional feature datasets are input into a pre-trained physical fitness assessment classifier (such as the XGBoost ensemble model), and a quantitative indicator set (including VO2 max equivalent, metabolic equivalent level, and heart rate recovery rate) is output through a feature importance weighting mechanism. This achieves an objective quantitative assessment of physical health status, overcoming the problem of lack of dynamic adaptability in matching real-time changes in physical fitness with manual prescriptions. Based on the mapping relationship between the quantitative indicator set and a preset physical health standard library (such as health level identifier matching), a preset exercise prescription rule library (such as decision trees and fuzzy logic engines) is used to match basic exercise parameters, generating an initial exercise prescription containing parameters such as exercise type, intensity threshold, duration, and rest interval. Through dynamic mapping and parameterized output, the personalization and accuracy of prescriptions are improved, indirectly alleviating the problem of delayed health risk warnings.

[0064] In one embodiment, after generating the initial exercise prescription, the method further includes:

[0065] S11, acquire the user's historical exercise biosignals, historical exercise data sequences and individual health requirement parameters;

[0066] S12, based on historical motion data sequences, performs time-series alignment processing on historical motion biological signals;

[0067] S13, extract features from the time-aligned historical motion biosignals, and calculate motion intensity distribution index, duration stability index and recovery cycle efficiency index;

[0068] S14. Input the individual health needs parameters into the preset needs analysis model to generate an exercise preference vector containing exercise type identifiers, tolerable intensity ranges and health intervention priority parameters.

[0069] S15 constructs a multi-dimensional user exercise profile based on exercise intensity distribution index, duration stability index, recovery cycle efficiency index, and exercise preference vector;

[0070] S16. Based on the user's motion profile, a multi-objective optimization algorithm is used to adjust the exercise type, intensity threshold, duration, and interval cycle parameters in the initial exercise prescription to generate the final exercise prescription.

[0071] Specifically, based on the user's historical exercise biosignals (including time-series data of heart rate, blood pressure, and respiratory rate), historical exercise data sequences (structured logs recording the start and end times and types of exercise), and individual health needs parameters (such as weight loss goals or cardiopulmonary function enhancement needs), the fragmented biosignals are time-aligned using the time-series markers of the historical exercise data sequences to eliminate time offsets caused by differences in sampling frequencies from multiple sensors. Multi-dimensional feature extraction is performed on the time-aligned biosignals: heart rate data can be divided into different intervals using a preset heart rate intensity grading threshold, and an exercise intensity distribution index is generated by statistically analyzing the cumulative duration percentage of each interval; based on the effective duration sequence of N consecutive exercise sessions, a duration stability index is calculated; and based on blood pressure time-series data, the mean blood pressure change rate is calculated to generate a recovery cycle efficiency index reflecting physical recovery ability. Individual health needs parameters are input into a pre-defined needs analysis model (such as a decision tree-based classifier) ​​to generate a structured exercise preference vector, which includes exercise type identifiers (such as aerobic training encoded as 01), tolerable intensity range (such as heart rate range of 120–150 bpm) and health intervention priority parameters (such as a weight coefficient of 0.7 for cardiopulmonary function optimization). Based on the above indicators, a user exercise profile is constructed: the exercise intensity distribution index, duration stability index, and recovery cycle efficiency index are standardized and mapped to endurance dimension score, exercise compliance dimension score, and physical recovery dimension score, respectively. Weight coefficients are assigned based on the priority parameters of the exercise preference vector, and a three-dimensional exercise profile vector is generated. Prescription parameters can be adjusted using a multi-objective optimization algorithm: a set of adaptable exercise types (e.g., jogging, swimming) is selected based on the exercise type identifier. The first optimization objective (optimizing intensity threshold) is established using the endurance dimension score to maximize exercise adaptability; the second optimization objective (optimizing duration) is established using the exercise compliance dimension score to maximize prescription sustainability; and the third optimization objective (optimizing interval cycle) is established using the physical recovery dimension score to minimize health risks. The initial prescription parameters are updated by solving for the Pareto optimal solution to generate the final exercise prescription.

[0072] In one embodiment, feature extraction is performed on the time-aligned historical motion biosignals, and motion intensity distribution index, duration stability index, and recovery cycle efficiency index are calculated. The specific steps are as follows:

[0073] S21. The heart rate data in the historical exercise biosignal is divided into intervals using a preset heart rate intensity grading threshold. The cumulative exercise duration is statistically analyzed in the low-intensity interval (heart rate value below 100 bpm), the medium-intensity interval (heart rate value between 100-140 bpm), and the high-intensity interval (heart rate value above 140 bpm), generating an exercise intensity distribution index.

[0074] S22, based on the standard deviation algorithm, processes the effective duration sequence of N consecutive movements in historical motion biosignals, and calculates the duration stability index according to the following formula:

[0075]

[0076] Among them, t i The duration of a single exercise session. σ is the average duration of the exercise, N is the number of consecutive exercises, and σ is the mean duration of the exercise. T The quantification value of the fluctuation in motion duration stability;

[0077] S23. Based on time-series blood pressure data from historical exercise biosignals, the mean rate of change of blood pressure within each recovery cycle is calculated using the following formula to obtain the recovery cycle efficiency index:

[0078]

[0079] Where R is the recovery cycle efficiency, K is the total number of data points within the recovery cycle, and P... k Let t be the blood pressure value at the k-th data point, and Δt be the preset time interval.

[0080] For example, a preset heart rate intensity grading threshold (based on the ACSM (American College of Sports Medicine) standard) can be used to divide heart rate data in historical exercise biosignals into intervals: intervals below 100 bpm are marked as low intensity, intervals between 100-140 bpm as medium intensity, and intervals above 140 bpm as high intensity; the cumulative exercise duration percentage within each interval is calculated to generate an exercise intensity distribution index. The effective duration sequence of N consecutive exercises (default value N=7, covering a one-week exercise cycle; if historical records are insufficient, the actual number of exercises is used) in historical exercise biosignals is processed using a standard deviation algorithm, and the duration stability index is calculated using a formula: Among them, t i The duration of a single exercise session (single exercise session duration ≥ 10 min, invalid short-duration activities are filtered out). σ is the average duration of the exercise, N is the number of consecutive exercises, and σ is the mean duration of the exercise. T This provides a quantification value for the fluctuation of exercise duration stability; based on time-series blood pressure data from historical exercise biosignals, the mean of the rate of change of blood pressure within each recovery cycle is calculated to generate a recovery cycle efficiency index. Where R is the recovery cycle efficiency, K is the total number of data points within the recovery cycle (K≥3, at least 3 data points are required to calculate the trend), and P... k Let t be the blood pressure value at the k-th data point, and Δt be the preset time interval.

[0081] In one embodiment, a multi-dimensional user motion profile is constructed based on motion intensity distribution indicators, duration stability indicators, recovery cycle efficiency indicators, and motion preference vectors, including:

[0082] S31, standardizes the exercise intensity distribution index, duration stability index and recovery cycle efficiency index;

[0083] S32 maps the standardized exercise intensity distribution index to the endurance dimension score, the standardized duration stability index to the exercise compliance dimension score, and the standardized recovery cycle efficiency index to the physical recovery dimension score.

[0084] S33 assigns weight coefficients to endurance dimension scores, exercise compliance dimension scores, and physical recovery dimension scores based on the health intervention priority parameter in the exercise preference vector, and combines the weighted dimension scores with the exercise type identifier to generate a three-dimensional exercise profile vector as the user's exercise profile.

[0085] Specifically, the exercise intensity distribution index, duration stability index, and recovery cycle efficiency index are standardized using the z-score method, converting each index into a standardized score with a unified dimension to eliminate dimensional differences and improve comparability. After standardization, the exercise intensity distribution index is mapped to an endurance dimension score (e.g., the standardized score is converted to a 0-100 score range using a linear mapping function, with higher scores indicating higher endurance levels). The standardized duration stability index is mapped to an exercise adherence dimension score (higher scores reflect the stability of users' adherence to exercise plans), and the standardized recovery cycle efficiency index is mapped to a physical recovery dimension score (higher scores indicate efficient recovery ability). Based on the health intervention priority parameter in the exercise preference vector (e.g., a weight coefficient of 0.7 for cardiopulmonary function optimization), the weight coefficients of the endurance dimension score, exercise adherence dimension score, and physical recovery dimension score are dynamically allocated (e.g., if the priority parameter emphasizes endurance, a higher weight is assigned to the endurance dimension). The weighted dimension scores are then fused using an exercise type identifier (extracted from the exercise preference vector) and a weighted average algorithm V = ω. e ·S e +ω c ·S c +ω r ·S r , where ω e ω c and ωr Weighting coefficient, S e S c and S r A three-dimensional motion profile vector (such as vector form [endurance score, compliance score, recovery score]) is generated for each dimension score. This vector is used as a complete representation of the user's endurance, exercise habits and recovery ability as a user motion profile, which is then used for multi-objective optimization of subsequent exercise prescriptions.

[0086] In one embodiment, based on the user's motion profile, a multi-objective optimization algorithm is used to adjust the motion parameters in the initial exercise prescription to generate a final exercise prescription, including:

[0087] S41, Determine the set of sports types that the user can adapt to based on the sports type identifier in the user's sports profile;

[0088] S42, based on the endurance dimension score in the user's exercise profile, establish the first optimization objective of maximizing exercise adaptability, based on the exercise compliance dimension score, establish the second optimization objective of maximizing prescription sustainability, and based on the physical recovery dimension score, establish the third optimization objective of minimizing health risks;

[0089] S43, use a multi-objective optimization algorithm to perform multi-objective optimization on the first optimization objective, the second optimization objective and the third optimization objective, and obtain the optimized set of motion parameters;

[0090] S44. Based on the matching results between the adaptable set of exercise types and the optimized set of exercise parameters, update the exercise type, exercise intensity, duration and interval cycle parameters in the initial exercise prescription to generate the final exercise prescription.

[0091] For example, the exercise type identifier (encoded parameters extracted from the exercise preference vector, such as 01 representing aerobic training, used to identify the user's preferred exercise category); the set of adaptable exercise types (compatible exercise types dynamically selected based on this identifier); and optimization objectives (including the first objective of maximizing exercise adaptability (based on endurance dimension scores, aiming to improve the match between exercise intensity and user physical condition), the second objective of maximizing prescription sustainability (based on exercise adherence dimension scores, optimizing duration to enhance the user's long-term adherence), and the third objective of minimizing health risks (based on physical recovery dimension scores, adjusting interval cycles to reduce health risks). (Low probability of sports injury); the optimized set of exercise parameters is a combination of parameters obtained through a multi-objective optimization algorithm (including core variables such as exercise intensity threshold, duration, and rest interval). Based on the exercise type identifier in the user's exercise profile (extracted from the exercise preference vector in the user's exercise profile), a rule matching engine (such as a decision tree-based classifier) ​​can determine the set of exercise types suitable for the user (e.g., if the identifier indicates an aerobic preference, then the set {jogging, cycling, swimming} is generated); a multi-objective optimization framework is established, where the first optimization objective is defined as a function that maximizes the endurance dimension score (e.g., objective function f1 = ω). e ·S e , where ω e S is the endurance weight. e For endurance scoring, the exercise intensity threshold is optimized to match the user's cardiorespiratory capacity. The second optimization objective is defined as a function that maximizes the exercise compliance dimension score (objective function f2 = ω). c ·S c , where ω c For compliance weights, S c To improve adherence scoring and duration of the program to enhance feasibility, the third optimization objective is defined as minimizing the negative risk of the physical recovery dimension score (objective function f3 = ω). r ·S r , where ω r To restore the weights, S rTo restore the score and optimize the rest interval to reduce the risk of blood pressure fluctuations, a multi-objective optimization algorithm (such as NSGA-II) is used to find the Pareto optimal solution for these three objectives. Through iterative calculation (setting the population size to 50 and iterating 100 times), the optimal parameter set (e.g., a parameter combination with an output intensity threshold of 120-150 bpm, a duration of 30 minutes, and a rest interval of 60 seconds) is searched in the solution space. Based on the matching results between the adaptable set of exercise types and the optimal parameter set (e.g., using a fuzzy logic engine to verify compatibility), the exercise type (selecting matching items from the set), exercise intensity, duration, and rest interval parameters in the initial exercise prescription are updated to generate the final exercise prescription (e.g., if the set is {jogging, swimming} and the parameter set indicates aerobic intensity, then jogging is output as the exercise type, and parameters such as the intensity threshold are integrated). Through multi-objective collaborative optimization, the problem of lack of dynamic adaptability in traditional methods is solved, achieving accurate generation of personalized prescriptions.

[0092] In one embodiment, the method further includes:

[0093] S51, based on historical exercise data sequences, calculates health trend prediction indicators according to the rate of change of endurance dimension scores, exercise compliance dimension scores, and physical recovery dimension scores in the user's exercise profile;

[0094] S52, input the health trend prediction index into the pre-trained health risk prediction model, and output the health risk probability and exercise capacity offset within the future preset period.

[0095] S53 generates exercise guidance suggestions that include abnormal state warnings and exercise parameter adjustment strategies based on the probability of health risks and the offset of exercise capacity.

[0096] Specifically, this includes: historical exercise data sequences (a time-series structured dataset of users' past exercise records, including timestamps indicating start and end times, type, and associated biosignals); health trend prediction indicators (which can be generated by extracting endurance score change rate, compliance score change rate, and recovery score change rate using time series differencing algorithms (e.g., S = ΔS / Δt, where S is the score value and Δt is a preset time window), combined with a weighted average formula (weights dynamically allocated by the user's health intervention priority parameters), used to capture the dynamic evolution trend of the user's physical condition); and a pre-trained health risk prediction model (based on an LSTM neural network). The system (built and trained using historical health big data) is used to process trend prediction indicators and output the probability of health risks (quantifying the likelihood of a user experiencing a cardiovascular event or sports injury, ranging from 0 to 1) and the exercise capacity offset (predicting the expected change in a user's maximum oxygen uptake or metabolic equivalent, expressed as a percentage) within a preset future period (e.g., the next 7 days); abnormal state warnings (risk alerts triggered when the probability of health risks exceeds a preset threshold (e.g., >0.6); and exercise parameter adjustment strategies (optimization schemes generated based on risk probability, offset, and user profile, such as lowering the intensity threshold or extending the rest interval to mitigate risks).

[0097] Based on historical exercise data sequences (such as a user's exercise logs over the past 30 days), and according to the time-varying rates of change of various dimensions of the user's exercise profile (endurance, compliance, recovery score), a linear regression algorithm can be used to calculate health trend prediction indicators. These indicators are then input into a pre-trained health risk prediction model (the model is trained using a historical dataset, where the input is the trend indicator and the output is labeled with actual risk events and ability changes). Forward propagation outputs the probability of health risks within a preset future period (e.g., probability values ​​generated using the Sigmoid function in the model's output layer) and the exercise ability offset (e.g., the percentage offset predicted through a regression layer). Based on the health risk probability and exercise ability offset, a rule engine (e.g., a decision tree-based logic system) is used to process the data. The system assesses health risks by comparing the probability of a health risk with a preset threshold (a quantitative boundary set by statistical analysis of historical health big data, such as a database containing user group exercise events and health outcomes; this threshold can be dynamically adjusted based on user group parameters such as age and gender; for example, the threshold for elderly users may be lowered to >0.5 to accommodate higher baseline risks). If the probability of a risk exceeds the threshold, an abnormal state warning is triggered (e.g., high risk: suggest reducing intensity). Combined with the offset of exercise ability, and based on exercise adjustment strategies (e.g., increasing the recovery period when the offset of exercise ability is negative; the specific adjustment value is dynamically adjusted by the dimension score of the user's exercise profile; for example, reducing the adjustment range when the physical recovery dimension score is high), exercise guidance suggestions containing warning text and specific parameter adjustment values ​​are generated.

[0098] In one embodiment, the method further includes:

[0099] S61, based on the endurance dimension score, exercise compliance dimension score and physical recovery dimension score in the user's exercise profile, generates a visual score map through radial basis function interpolation algorithm;

[0100] S62, Based on the exercise type, intensity threshold, duration and interval cycle parameters in the final exercise prescription, generate an exercise prescription execution heatmap;

[0101] S63, construct a time-series early warning dashboard for the evolution of future health status based on health trend prediction indicators and health risk probabilities;

[0102] S64 synchronously displays a visual scoring chart, an exercise prescription execution heatmap, and a time-series early warning dashboard through an interactive interface.

[0103] For example, based on the three-dimensional scores (endurance, compliance, recovery) of the user's motion profile, a radial basis function interpolation algorithm can be used to achieve continuous mapping of spatial data: the three-dimensional scores are used as discrete spatial coordinate points (e.g., the points (75, 80, 65) corresponding to endurance score = 75, compliance score = 80, and recovery score = 65), and the Gaussian kernel function φ(r) = exp(-εr) is applied. 2The algorithm calculates interpolation weights between neighboring points, fills coordinate gaps through matrix operations, and generates a continuous 3D surface model (endurance dimension mapped to the red channel, compliance to the green channel, and recovery to the blue channel; RGB fusion forms a full-color gradient map). It outputs a dynamic visualization model to intuitively display the user's overall physical condition (e.g., high endurance areas are presented in warm colors). Based on the structured parameter set of the final exercise prescription, a two-dimensional parameter matrix is ​​formed, with exercise type as the row index (e.g., constructing a set {jogging, cycling, swimming}) and intensity threshold and duration as the column index (e.g., intensity range 120-150 bpm, duration range 30-45 minutes). Matrix element values ​​are filled by the interval period (e.g., a 60-second interval is encoded as a high-density value of 1.0, and a 90-second interval as a low-density value of 0.5). A color mapping rule is applied (short intervals are represented by dark red for high intensity, and long intervals by light blue for low intensity) to render a heatmap. The parameter matrix is ​​then used to visualize the data. Visual encoding allows users to intuitively identify prescription intensity distribution (e.g., the intersection of swimming + 140 bpm + 40 minutes is displayed in dark red); the construction of a time-series early warning dashboard involves the integration of health trend prediction indicators and health risk probabilities: displaying trend indicators (e.g., endurance change rate curve) and risk probability curves on a time axis (X-axis for a future preset period, such as 7 days) (dual Y-axis design, left axis for change rate, right axis for probability value); when the risk probability exceeds a preset threshold (e.g., >0.6), a dynamic warning marker is triggered (e.g., a flashing red icon), and a prompt is generated based on the exercise parameter adjustment strategy (e.g., a text prompt indicating a 10% reduction in intensity); the above components are displayed synchronously through an interactive interface, which can adopt a split-screen layout (e.g., a scoring graph embedded on the left, a prescription heatmap embedded in the center, and an early warning dashboard embedded on the right), supporting dynamic interaction such as clicking on the heatmap to view parameter details or dragging the graph to rotate the view; the data linkage mechanism ensures that when the user triggers a time-series warning, the associated heatmap unit is automatically highlighted, forming an assessment-early warning-feedback closed loop.

[0104] The aforementioned method for generating personalized exercise prescriptions based on biofeedback acquires multi-dimensional biosignals such as heart rate, blood pressure, and respiratory rate during user exercise. A signal denoising model is used for filtering to form a high-quality biosignal set, overcoming the problems of insufficient utilization of biosignals and inaccurate assessments due to reliance on a single heart rate threshold in existing technologies. Support vector machine (SVM) algorithms are used to perform multi-dimensional recognition and pattern classification of the biosignal set, extracting multi-dimensional features such as cardiopulmonary function intensity, metabolic efficiency level, and exercise stress response. Combined with a pre-trained physical fitness assessment classifier, a quantitative indicator set is output, enabling an objective quantitative assessment of the user's physical health status. This avoids cognitive biases caused by reliance on physicians' subjective experience in traditional methods and can accurately match real-time changes in the user's physical condition. By acquiring the user's historical exercise data and individual health needs parameters, a multi-dimensional user exercise profile is constructed. A multi-objective optimization algorithm is used to adjust the initial prescription parameters to generate the final exercise prescription. Combined with a health trend prediction and early warning mechanism, this solves the problems of lack of dynamic adaptability in traditional methods, inability to optimize prescription parameters in real time based on the user's historical exercise characteristics and changes in health needs, and delayed health risk warnings, thus improving the accuracy and safety of personalized health interventions.

[0105] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0106] Based on the same inventive concept, this application also provides a biofeedback-based personalized exercise prescription generation device for realizing the above-mentioned generation of personalized exercise prescriptions for physical health. The solution provided by this device is similar to the solution described in the above-described method. Therefore, the specific limitations of one or more embodiments of a biofeedback-based personalized exercise prescription generation device provided below can be found in the limitations of the biofeedback-based personalized exercise prescription generation method described above, and will not be repeated here.

[0107] In one exemplary embodiment, such as Figure 2 As shown, a device for generating personalized exercise prescriptions for physical health based on biofeedback is provided, comprising:

[0108] The signal acquisition and processing module 101 is used to acquire the user's biological signals during exercise and to perform filtering processing using a pre-established signal denoising model to form a set of biological signals; the biological signals include heart rate, blood pressure and respiratory rate;

[0109] The multi-dimensional feature recognition module 102 is used to perform multi-dimensional recognition and pattern classification on the biological signal set using the support vector machine algorithm, and extract a multi-dimensional feature dataset containing cardiopulmonary function intensity, metabolic efficiency level and exercise stress response.

[0110] The physical fitness assessment quantification module 103 is used to input a multi-dimensional feature dataset into a pre-trained physical fitness assessment classifier and output a set of quantitative indicators representing the user's physical health status.

[0111] The exercise prescription generation module 104 is used to generate an initial exercise prescription by matching basic exercise parameters with a preset exercise prescription rule library based on the mapping relationship between a set of quantitative indicators and a preset physical health standard library. The initial exercise prescription includes exercise type, intensity threshold, duration and interval cycle parameters.

[0112] In one embodiment, the exercise prescription generation module 104 is further configured to:

[0113] Acquire users' historical exercise biosignals, historical exercise data sequences, and individual health requirement parameters;

[0114] Based on historical motion data sequences, time-series alignment processing is performed on historical motion biosignals;

[0115] Feature extraction is performed on time-aligned historical motion biosignals to calculate motion intensity distribution index, duration stability index, and recovery cycle efficiency index.

[0116] Input individual health needs parameters into a preset needs analysis model to generate an exercise preference vector containing exercise type identifiers, tolerable intensity ranges, and health intervention priority parameters.

[0117] Based on exercise intensity distribution indicators, duration stability indicators, recovery cycle efficiency indicators, and exercise preference vectors, a multi-dimensional user exercise profile is constructed.

[0118] Based on the user's exercise profile, a multi-objective optimization algorithm is used to adjust the exercise type, intensity threshold, duration, and interval cycle parameters in the initial exercise prescription to generate the final exercise prescription.

[0119] In one embodiment, the exercise prescription generation module 104 is further configured to:

[0120] Feature extraction is performed on the time-aligned historical motion biosignals, and motion intensity distribution, duration stability, and recovery cycle efficiency indices are calculated. The specific steps are as follows:

[0121] The heart rate data in historical exercise biosignals are divided into intervals using a preset heart rate intensity grading threshold. The cumulative exercise duration is then statistically analyzed within the low-intensity interval (heart rate below 100 bpm), the medium-intensity interval (100-140 bpm), and the high-intensity interval (heart rate above 140 bpm), generating an exercise intensity distribution index.

[0122] Based on the standard deviation algorithm, the effective duration sequence of N consecutive movements in historical motion biosignals is processed, and the duration stability index is calculated according to the following formula:

[0123]

[0124] Among them, t i The duration of a single exercise session. σ is the average duration of the exercise, N is the number of consecutive exercises, and σ is the mean duration of the exercise. T The quantification value of the fluctuation in motion duration stability;

[0125] Based on time-series blood pressure data from historical exercise biosignals, the mean rate of change of blood pressure within each recovery cycle is calculated using the following formula to obtain the recovery cycle efficiency index:

[0126]

[0127] Where R is the recovery cycle efficiency, K is the total number of data points within the recovery cycle, and P... k Let t be the blood pressure value at the k-th data point, and Δt be the preset time interval.

[0128] In one embodiment, the physical fitness assessment quantification module 103 is further configured to:

[0129] The exercise intensity distribution index, duration stability index, and recovery cycle efficiency index were standardized.

[0130] The standardized exercise intensity distribution index was mapped to the endurance dimension score, the standardized duration stability index was mapped to the exercise compliance dimension score, and the standardized recovery cycle efficiency index was mapped to the physical recovery dimension score.

[0131] Based on the health intervention priority parameter in the exercise preference vector, weight coefficients are assigned to the endurance dimension score, exercise compliance dimension score, and physical recovery dimension score. The weighted dimension scores are then fused with the exercise type identifier to generate a three-dimensional exercise profile vector as the user's exercise profile.

[0132] In one embodiment, the exercise prescription generation module 104 is further configured to:

[0133] Based on the motion type identifier in the user's motion profile, determine the set of motion types that the user can adapt to;

[0134] The first optimization objective is to maximize exercise adaptability based on the endurance dimension score in the user's exercise profile; the second optimization objective is to maximize the sustainability of the prescription based on the exercise compliance dimension score; and the third optimization objective is to minimize health risks based on the physical recovery dimension score.

[0135] A multi-objective optimization algorithm is used to optimize the first, second, and third optimization objectives to obtain an optimized set of motion parameters.

[0136] Based on the matching results between the adaptable set of exercise types and the optimized set of exercise parameters, the exercise type, exercise intensity, duration and interval cycle parameters in the initial exercise prescription are updated to generate the final exercise prescription.

[0137] In one embodiment, the physical fitness assessment quantification module 103 is further configured to:

[0138] Based on historical exercise data sequences, health trend prediction indicators are calculated according to the rate of change of endurance, exercise compliance, and physical recovery scores in user exercise profiles.

[0139] Input health trend prediction indicators into a pre-trained health risk prediction model, and output the probability of health risks and the offset of exercise capacity within a future preset period.

[0140] The exercise prescription generation module 104 is also used for:

[0141] Based on the probability of health risks and the offset of exercise capacity, exercise guidance suggestions are generated, which include abnormal state warnings and exercise parameter adjustment strategies.

[0142] In one embodiment, a visualization module is also included for:

[0143] Based on the endurance, exercise compliance, and physical recovery scores in the user's exercise profile, a visual score map is generated using a radial basis function interpolation algorithm.

[0144] Based on the exercise type, intensity threshold, duration, and rest interval parameters in the final exercise prescription, an exercise prescription execution heatmap is generated.

[0145] Based on health trend prediction indicators and health risk probabilities, a time-series early warning dashboard for the evolution of future health status is constructed.

[0146] The interactive interface synchronously displays a visual scoring chart, a heatmap of exercise prescription execution, and a time-series early warning dashboard.

[0147] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of a biofeedback-based personalized exercise prescription generation method for physical health as described above.

[0148] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0149] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0150] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A method for generating personalized exercise prescriptions for physical health based on biofeedback, characterized in that, The method includes: The system acquires the user's biological signals during exercise and filters them using a pre-established signal denoising model to form a set of biological signals, including heart rate, blood pressure, and respiratory rate. The biological signal set was subjected to multi-dimensional identification and pattern classification using the support vector machine algorithm, and a multi-dimensional feature dataset containing cardiopulmonary function intensity, metabolic efficiency level and exercise stress response was extracted. The multi-dimensional feature dataset is input into a pre-trained physical fitness assessment classifier, which outputs a set of quantitative indicators representing the user's physical health status. Based on the mapping relationship between the quantitative indicator set and the preset physical health standard library, the basic exercise parameters are matched using the preset exercise prescription rule library to generate an initial exercise prescription; the initial exercise prescription includes exercise type, intensity threshold, duration and interval cycle parameters.

2. The method according to claim 1, characterized in that, After generating the initial exercise prescription, the method further includes: Acquire users' historical exercise biosignals, historical exercise data sequences, and individual health requirement parameters; Based on the historical motion data sequence, the historical motion biosignal is subjected to time-series alignment processing; Feature extraction is performed on time-aligned historical motion biosignals to calculate motion intensity distribution index, duration stability index, and recovery cycle efficiency index. The individual health needs parameters are input into a preset needs analysis model to generate an exercise preference vector containing exercise type identifiers, tolerable intensity ranges, and health intervention priority parameters. Based on the aforementioned exercise intensity distribution index, duration stability index, recovery cycle efficiency index, and exercise preference vector, a multi-dimensional user exercise profile is constructed. Based on the user's motion profile, a multi-objective optimization algorithm is used to adjust the motion type, intensity threshold, duration, and interval parameters in the initial exercise prescription to generate the final exercise prescription.

3. The method according to claim 2, characterized in that, The specific steps for extracting features from the time-aligned historical motion biosignals and calculating motion intensity distribution, duration stability, and recovery cycle efficiency indices are as follows: The heart rate data in the historical exercise biosignal is divided into intervals using a preset heart rate intensity grading threshold. The cumulative exercise duration is then statistically analyzed within the low-intensity interval (heart rate below 100 bpm), the medium-intensity interval (100-140 bpm), and the high-intensity interval (heart rate above 140 bpm) to generate the exercise intensity distribution index. Based on the standard deviation algorithm, the effective duration sequence of N consecutive movements in the historical motion biosignal is processed, and the duration stability index is calculated according to the following formula: Among them, t i The duration of a single exercise session. σ is the average duration of the exercise, N is the number of consecutive exercises, and σ is the mean duration of the exercise. T The quantification value of the fluctuation in motion duration stability; Based on time-series blood pressure data from historical exercise biosignals, the mean rate of change of blood pressure within each recovery cycle is calculated according to the following formula to obtain the recovery cycle efficiency index: Where R is the recovery cycle efficiency, K is the total number of data points within the recovery cycle, and P... k Let t be the blood pressure value at the k-th data point, and Δt be the preset time interval.

4. The method according to claim 2, characterized in that, The process of constructing a multi-dimensional user motion profile based on the exercise intensity distribution index, duration stability index, recovery cycle efficiency index, and exercise preference vector includes: The exercise intensity distribution index, duration stability index, and recovery cycle efficiency index are standardized. The standardized exercise intensity distribution index was mapped to the endurance dimension score, the standardized duration stability index was mapped to the exercise compliance dimension score, and the standardized recovery cycle efficiency index was mapped to the physical recovery dimension score. Based on the health intervention priority parameter in the exercise preference vector, weight coefficients are assigned to the endurance dimension score, exercise compliance dimension score, and physical recovery dimension score. The weighted dimension scores are then fused with the exercise type identifier to generate a three-dimensional exercise profile vector as the user's exercise profile.

5. The method according to claim 4, characterized in that, The step of adjusting the exercise parameters in the initial exercise prescription based on the user's motion profile and generating the final exercise prescription using a multi-objective optimization algorithm includes: Based on the motion type identifier in the user's motion profile, determine the set of motion types that the user can adapt to; Based on the endurance dimension score in the user's exercise profile, a first optimization objective is established to maximize exercise adaptability; based on the exercise compliance dimension score, a second optimization objective is established to maximize prescription sustainability; and based on the physical recovery dimension score, a third optimization objective is established to minimize health risks. A multi-objective optimization algorithm is used to perform multi-objective optimization on the first, second, and third optimization objectives to obtain an optimized set of motion parameters. Based on the matching results between the set of adaptable exercise types and the optimized set of exercise parameters, the exercise type, exercise intensity, duration, and interval cycle parameters in the initial exercise prescription are updated to generate the final exercise prescription.

6. The method according to claim 4, characterized in that, The method further includes: Based on the historical exercise data sequence, and according to the rate of change of endurance dimension score, exercise compliance dimension score, and physical recovery dimension score in the user exercise profile, a health trend prediction index is calculated. The health trend prediction index is input into a pre-trained health risk prediction model, which outputs the probability of health risk and the offset of exercise ability within a future preset period. Based on the health risk probability and the offset of exercise ability, exercise guidance suggestions are generated, which include abnormal state warnings and exercise parameter adjustment strategies.

7. The method according to claim 6, characterized in that, The method further includes: Based on the endurance dimension score, exercise compliance dimension score, and physical recovery dimension score in the user's exercise profile, a visual score map is generated using a radial basis function interpolation algorithm; Based on the exercise type, intensity threshold, duration and interval cycle parameters in the final exercise prescription, an exercise prescription execution heatmap is generated. Based on health trend prediction indicators and health risk probabilities, a time-series early warning dashboard for the evolution of future health status is constructed. The interactive interface synchronously displays the visual scoring chart, exercise prescription execution heatmap, and time-series early warning dashboard.

8. A device for generating personalized exercise prescriptions for physical health based on biofeedback, characterized in that, The device includes: The signal acquisition and processing module is used to acquire the user's biological signals during exercise and to filter them using a pre-established signal denoising model to form a set of biological signals; the biological signals include heart rate, blood pressure and respiratory rate; The multi-dimensional feature recognition module is used to perform multi-dimensional recognition and pattern classification on the biological signal set using the support vector machine algorithm, and extract a multi-dimensional feature dataset containing cardiopulmonary function intensity, metabolic efficiency level and exercise stress response. The physical fitness assessment quantification module is used to input the multi-dimensional feature dataset into a pre-trained physical fitness assessment classifier and output a set of quantitative indicators representing the user's physical health status. The exercise prescription generation module is used to generate an initial exercise prescription based on the mapping relationship between the quantitative indicator set and the preset physical health standard library, and by matching basic exercise parameters using the preset exercise prescription rule library; the initial exercise prescription includes exercise type, intensity threshold, duration and interval cycle parameters.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images