A real-time monitoring system for teenagers' physical fitness training

Abstract

The present invention discloses a real-time monitoring system for teenagers' physical exercise, which relates to the field of intelligent monitoring technology. Through the spatiotemporal alignment processing of joint three-dimensional motion, electromyographic signals, and plantar pressure and the dynamic matching technology of bone age, an individualized health assessment model is constructed. Timing deviation is eliminated through unified clock marking and interpolation compensation, and parameters such as bone stress deviation, muscle synergy asymmetry, and plantar pressure imbalance are extracted, and weighted comprehensive health indicators are generated. Combined with a preset training path model, the health indicator trend is analyzed in real time and hierarchical adjustment instructions are triggered, the training intensity, interval and action plan are dynamically optimized, and the path model parameters are corrected through a closed-loop feedback mechanism to achieve adaptive updates. It solves the defects of large fusion deviation and strong lag of static evaluation of traditional teenagers' physical fitness monitoring data, realizes real-time and accurate monitoring of teenagers' physical exercise and personalized training optimization, and effectively reduces the risk of sports injuries.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent monitoring, and in particular to a real-time monitoring system for physical exercise of teenagers. Background Art

[0002] Existing technologies for adolescent fitness monitoring rely on single-dimensional motion parameters or static physiological indicators, making it difficult to fully reflect the biomechanical state in complex exercise scenarios. Traditional systems collect joint, muscle, and plantar pressure data, and then use static assessment models to determine skeletal stress and muscle synergy using fixed thresholds. This ignores the dynamic impact of adolescent bone development on athletic performance, resulting in deviations from actual physiological load.

[0003] Existing static assessment models are unable to adapt to the biomechanical characteristics of different bone age stages, resulting in inconsistent applicability of the same training program for adolescents during their developmental stages. These issues limit the practical value of monitoring systems and make it difficult to meet the needs of personalized scientific training.

[0004] In response to the above problems, this field urgently needs a monitoring system that can integrate multi-dimensional biomechanical parameters, dynamically adapt to individual developmental characteristics, and optimize training strategies in real time, so as to address the shortcomings of traditional methods in data synchronization, evaluation accuracy and decision-making real-time. Summary of the Invention

[0005] In view of the shortcomings of the existing technology in monitoring the physical fitness of teenagers, such as large data fusion deviation and strong lag in static evaluation, the present invention provides a real-time monitoring system for the physical fitness exercise of teenagers.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0007] A real-time monitoring system for physical fitness training of adolescents, which adjusts training parameters based on data sensed by a sensor device worn on a target person, includes:

[0008] The data acquisition module is used to receive the three-dimensional joint motion data, electromyographic signal data and plantar pressure distribution data of the current training stage and the bone age data of the physiological development stage;

[0009] A standardized feature library, for simultaneously acquiring a standardized biomechanical feature data set of a historical training phase, wherein the feature data set includes a joint motion amplitude threshold, a muscle synergy symmetry range, and a standard value of plantar pressure distribution that matches the bone age data;

[0010] a data processing module, configured to extract joint angle feature points from the three-dimensional joint motion data, the activation time series of the electromyographic signal data, and the pressure area of ​​the plantar pressure distribution data, and perform spatiotemporal alignment processing with the feature data set to generate feature data;

[0011] a health index generation module, which calculates a plurality of biomechanical difference parameters reflecting bone stress deviation, muscle synergy asymmetry, and plantar pressure imbalance based on the aligned characteristic data;

[0012] generating a comprehensive health index based on the biomechanical difference parameters and recording the index as time series data;

[0013] Dynamic path decision module, including:

[0014] Dynamically compare the change trend of the comprehensive health indicator in the time series data with the change trend of the target health indicator in the corresponding stage in the preset training path model to generate a deviation direction and amplitude; the deviation direction is divided into positive deviation and negative deviation, the positive deviation is used to increase the training amplitude, and the negative deviation is used to weaken the training amplitude;

[0015] generating an adjustment instruction according to the deviation direction and magnitude;

[0016] The execution module is used to generate a training parameter adjustment amount including training intensity adjustment, time interval correction and action replacement scheme according to the adjustment instruction through a preset rule mapping table.

[0017] Compared with the known public technology, the technical solution provided by the present invention has the following beneficial effects:

[0018] This invention addresses the unique needs of adolescent physical fitness monitoring by constructing a multi-dimensional, dynamic, real-time monitoring system. Its core innovation lies in the precise collection of biomechanical data and its deep integration with growth stages, effectively addressing the problem of traditional monitoring methods ignoring adolescent bone age differences and limiting evaluation dimensions. The system integrates three-dimensional joint motion, electromyographic signals, plantar pressure, and bone age data, relying on a standardized feature library for spatiotemporal alignment. Key parameters such as joint angle feature points and muscle activation sequences are extracted, and abnormal data is eliminated based on a threshold range matching bone age, ensuring scientific data processing. Furthermore, a comprehensive health index (Hindex) is generated through normalized weighted calculations. The weight coefficients of dimensions such as bone stress, muscle synergy, and plantar pressure are dynamically adjusted to ensure that the evaluation results are more aligned with the physiological characteristics of adolescents at different developmental stages, making them more targeted than traditional fixed-weight evaluation methods.

[0019] The dynamic optimization mechanism of this invention overcomes the limitations of the "one-way feedback" of traditional monitoring systems, achieving real-time intelligent optimization of training programs through a closed-loop control process of "monitoring-evaluation-adjustment-iteration." The dynamic path decision module dynamically compares the changing trends of health indicators over time series with the target trends of the preset training path model. It generates graded adjustment instructions for positive / negative deviations (such as shortening / extending training intervals, increasing intensity, or switching to alternative movements). The path model update module then corrects subsequent stage parameters and records version changes, forming a virtuous cycle of "data-driven, model self-learning, and benchmark updates." This mechanism not only promptly corrects training deviations but also continuously optimizes the training path based on actual results. Compared to traditional, experience-based adjustment methods, it is more timely and accurate, effectively reducing the risk of sports injuries and improving exercise efficiency.

[0020] This invention uses real-time data to generate a dynamic simulation model of the skeletal-muscular system. Combined with finite element analysis, it predicts stress distribution trends for future training cycles and displays these trends overlaid with target values. This system can generate adjustment instructions before potential risks occur (e.g., if stress imbalance exceeds a threshold), enabling a shift from post-intervention to proactive prevention. This simultaneous virtual-reality mapping not only enhances the system's intelligence but also provides safer, more scientifically sound support for adolescent physical training through preventative adjustments, offering a novel technical approach for developing personalized exercise intervention programs. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] To more clearly illustrate the technical solutions of the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort.

[0022] Figure 1 This is a system module connection diagram of the present invention;

[0023] Figure 2 This is a structural diagram of joint angle feature point extraction according to the present invention;

[0024] Figure 3 This is a structural diagram of the path model update module of the present invention;

[0025] Figure 4 This is the structural diagram of the biomechanical digital twin module of the present invention. DETAILED DESCRIPTION

[0026] It is easy to understand that according to the technical solution of the present invention, without changing the essential spirit of the present invention, a person skilled in the art can propose a variety of interchangeable structural modes and implementation modes. Therefore, the following specific embodiments and drawings are only exemplary descriptions of the technical solution of the present invention and should not be regarded as the entire invention or as a limitation or restriction of the technical solution of the present invention.

[0027] In existing technologies, adolescent physical fitness monitoring mostly relies on single sensor data or static assessment models, making it difficult to achieve collaborative analysis of multi-dimensional physiological parameters. When collecting joint movement, muscle activation, and plantar pressure data, traditional methods suffer from data fusion errors due to asynchronous sensor sampling timing, making it impossible to accurately capture the dynamic correlation characteristics of the motion chain biomechanics. Existing systems use fixed thresholds to assess bone stress and muscle coordination, ignoring the impact of adolescent bone age development differences on athletic ability, resulting in risk assessment results that deviate from actual physiological status. Especially in high-intensity scenarios such as strength training, the lack of real-time correlation analysis between plantar pressure distribution and joint load makes it difficult to provide timely warnings of potential sports injury risks.

[0028] To address the above issues, the inventors discovered that the spatiotemporal coupling of multimodal biomechanical data is significantly correlated with bone age developmental characteristics, and achieved personalized health assessment by establishing a dynamic path model. During the research, it was found that bone stress deviation, muscle synergy asymmetry, and plantar pressure imbalance exhibited differentiated weight distribution patterns at different bone age stages, thus proposing a dynamic weight allocation mechanism based on bone age parameters. Further experimental verification was conducted, and the time series trends of multidimensional health indicators were matched with the preset path model to construct a training parameter optimization system with adaptive adjustment capabilities.

[0029] Specifically, the monitoring system first synchronously collects the three-dimensional motion trajectory of the joint, surface electromyographic signals and plantar pressure distribution data. Through spatiotemporal alignment processing, the timing deviation between sensors is eliminated, and the key joint angle feature points, muscle activation time difference and pressure distribution ratio parameters are extracted. Combined with the biomechanical threshold library matched with bone age, quantitative indicators such as bone stress deviation and muscle synergy asymmetry are calculated, and a weighted comprehensive health index is generated. When it is detected that the health index trend continues to deviate from the preset path, the system automatically triggers a graded adjustment strategy: for positive deviations, the training interval is shortened and the intensity gradient is increased; for negative deviations, the recovery period is extended and a low-risk alternative action plan is switched. During the continuous monitoring process, the system dynamically updates the path model parameters based on real-time data, and optimizes the evaluation threshold and decision logic through a closed-loop feedback mechanism.

[0030] Compared with existing technologies, traditional methods are limited by single-dimensional data analysis and static assessment models, making them difficult to adapt to the dynamic changes in adolescent growth and development. This solution innovatively integrates multi-source biomechanical data with bone age development parameters to establish a personalized health assessment system through spatiotemporal coupling analysis. Unlike traditional fixed threshold judgment mechanisms, this solution can dynamically adjust training parameters based on real-time trends and continuously optimize decision-making accuracy through path model self-learning, significantly improving monitoring reliability and safety assurance capabilities in complex motion scenarios.

[0031] Through the above technical solutions, this application effectively solves the problems of multimodal data fusion bias and static assessment model lag, achieving accurate health risk warning while ensuring real-time performance. The dynamic path decision-making mechanism takes into account individual developmental differences and exercise load characteristics, and the adaptive adjustment function of model parameters ensures the accuracy of long-term monitoring. This method provides intelligent technical support for scientific physical training for adolescents, and is particularly suitable for scenarios such as improving athletic ability and preventing injuries during the critical period of growth and development.

[0032] After introducing the basic concept of the present invention, embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0033] Example

[0034] A real-time monitoring system for physical fitness training of teenagers in this embodiment is as follows: Figure 1 As shown, the training parameters are adjusted based on the sensing data of the sensor device worn on the target person, including:

[0035] The data acquisition module is used to receive the three-dimensional joint motion data, electromyographic signal data and plantar pressure distribution data of the current training stage and the bone age data of the physiological development stage;

[0036] A standardized feature library is used to simultaneously obtain a standardized biomechanical feature dataset from the historical training phase. The feature dataset includes joint motion range thresholds, muscle synergy symmetry ranges, and standard values ​​of plantar pressure distribution that match bone age data.

[0037] The data processing module is used to extract the joint angle feature points in the three-dimensional joint motion data, the activation time series of the electromyographic signal data, and the pressure area of ​​the plantar pressure distribution data, and perform spatiotemporal alignment processing with the feature data set to generate feature data;

[0038] The health index generation module calculates multiple biomechanical difference parameters reflecting bone stress deviation, muscle synergy asymmetry, and plantar pressure imbalance based on the aligned feature data;

[0039] A comprehensive health index is generated based on biomechanical difference parameters and recorded as time series data;

[0040] Dynamic path decision module, including:

[0041] Dynamically compare the change trend of the comprehensive health indicators in the time series data with the change trend of the target health indicators in the corresponding stage in the preset training path model to generate the deviation direction and amplitude; the deviation direction is divided into positive deviation and negative deviation, the positive deviation is used to increase the training amplitude, and the negative deviation is used to weaken the training amplitude;

[0042] Generate adjustment instructions based on the deviation direction and magnitude;

[0043] The execution module is used to generate training parameter adjustment amounts including training intensity adjustment, time interval correction and action replacement solutions according to the adjustment instruction through a preset rule mapping table.

[0044] 3D joint motion data refers to the 3D spatial coordinates and angular changes of joints (such as the knee, ankle, and hip) during adolescent training, acquired through an inertial measurement unit (IMU) or optical motion capture device. This data can be obtained using Xsens inertial sensors or Vicon optical cameras, and is used to record the dynamic trajectory of joints during exercise in real time. Electromyographic signal data refers to the bioelectrical signals of target muscle groups (such as the quadriceps and gastrocnemius) collected through surface electromyography (sEMG) sensors. This data can be obtained using the Delsys Trigno wireless electromyograph, and is used to reflect the temporal characteristics of muscle activation. Plantar pressure distribution data refers to the pressure distribution across various plantar regions (heel, forefoot, and arch) acquired through pressure-sensing insoles or force platforms. This data can be obtained using the Novel Pedar-X insole pressure system, and is used to quantify force balance in the foot. Bone age data refers to the physiological developmental stage of adolescents, as assessed by left wrist X-rays (such as the Greulich-Pyle method) or predicted by bioinformatics models, and is used to match personalized biomechanical assessment benchmarks.

[0045] Among them, the standardized biomechanical characteristic dataset refers to a benchmark library constructed based on a large sample of adolescent biomechanical experimental data, which is annotated by sports medicine experts or generated by machine learning model training. It contains joint motion amplitude thresholds that match each bone age stage (such as the safe range of knee flexion angle), muscle synergy symmetry range (such as the allowable value of the difference in myoelectric activation time between the left and right sides), and standard values ​​of plantar pressure distribution (such as the heel-forefoot pressure ratio benchmark value), which serve as a dynamic reference benchmark for the physical fitness assessment of adolescents.

[0046] Among them, spatiotemporal alignment processing refers to unifying the timestamps of joint three-dimensional motion data, the sampling frequency of electromyographic signals, and the acquisition period of plantar pressure data into the same time coordinate system, and eliminating the device space offset through three-dimensional coordinate alignment (such as SVD rigid transformation) to ensure the consistency of multi-source data in time and space dimensions, and provide a unified benchmark for subsequent feature extraction.

[0047] Biomechanical difference parameters mathematically quantify the degree of abnormal movement of bones, muscles, and soles, and are used to construct multidimensional health assessment indicators. Comprehensive health indicators simplify trend analysis of training effect evolution. Time series data refers to a collection of comprehensive health indicators recorded sequentially by training phase, stored in a time series database (such as InfluxDB) or CSV files, to reveal the temporal correlation and dynamic deviation characteristics of health indicator changes.

[0048] Among them, the preset training path model refers to a digital training plan that includes target health indicators for each bone age stage, training cycle time nodes and allowable deviation ranges. The adjustment instruction refers to generating a graded correction command based on the degree of deviation between the actual change rate of the comprehensive health indicator and the target change rate. It is specifically triggered by a rule engine (such as a positive deviation from the first threshold interval to trigger the shortening of the training interval) or a classification model (such as a gradient boosting tree classifier) ​​to automatically control the training intensity or action plan.

[0049] Among them, the training parameter adjustment refers to the specific correction value for training intensity, time interval and action substitution, which is obtained through table lookup method (such as the mapping table of deviation amplitude and intensity increment) or regression model (such as linear regression calculation of interval adjustment duration), and is used to optimize the execution strategy of subsequent training stages; path model update refers to the data iteration process of feeding back the adjustment amount to the preset training plan in real time, which is specifically implemented by version control (recording change time, effective stage identification) or parameter rewriting mechanism, which is used to maintain health indicators within the allowable deviation range of the model, forming a "monitoring-evaluation-adjustment-iteration" closed-loop control.

[0050] The core innovation of this application lies in constructing a closed-loop feedback control system based on biomechanical difference parameters and preset training paths. By comparing the evolution trend of comprehensive health indicators during adolescent training with the target path of the preset model in real time, it automatically generates training parameter adjustment instructions and updates the path model. This solves the technical bottleneck of traditional physical exercise monitoring technology that lacks a dynamic evaluation system and adaptive regulation capabilities, and realizes the leap from "experience-guided" to "data-driven".

[0051] The working process and principle of this application are as follows: first, the data acquisition module receives the joint three-dimensional motion data, electromyographic signal data, plantar pressure distribution data and bone age data of the current training stage; the standardized feature library synchronously obtains the historical standardized biomechanical feature data set that matches the current bone age (including joint motion amplitude threshold, muscle synergy symmetry range, etc.); the data processing module extracts joint angle feature points (such as knee flexion angle key value), electromyographic activation time series and plantar pressure area, and aligns them with the standardized data set in time and space to generate feature data; the health index generation module calculates the difference parameters such as bone stress deviation, muscle synergy asymmetry, plantar pressure imbalance, etc. based on the aligned features, and normalizes and dynamically calculates the difference parameters. Weighted comprehensive health indicators are generated and recorded as time series data; the dynamic path decision module applies sliding window analysis to the time series data, calculates the actual change rate and compares it with the target change rate of the preset training path model, and generates graded adjustment instructions according to the deviation direction and amplitude; the execution module converts the adjustment instructions into specific parameter adjustments for training intensity adjustment, time interval correction and action replacement schemes through the preset rule mapping table; finally, the path model update module corrects the parameters such as the time nodes of the subsequent stages and the allowable deviation range to maintain the comprehensive health indicators within the allowable range of the model, thereby realizing dynamic monitoring and intelligent optimization of the physical fitness training of adolescents, ensuring that the training process conforms to the laws of growth and development and reduces the risk of sports injuries.

[0052] like Figure 2 As shown, the present application further proposes that extracting joint angle feature points includes:

[0053] Based on the preset kinematic feature template, the key angle values ​​of knee flexion angle, ankle dorsiflexion angle and hip abduction angle are identified from the three-dimensional joint motion data;

[0054] Align the timestamps of key angle values. When the angle fluctuation of 5 consecutive frames of data is less than the fluctuation range, it is marked as a stable feature point.

[0055] Abnormal angle data that exceeds the bone age data matching amplitude threshold range is eliminated.

[0056] Specifically, the pre-set kinematic feature template utilizes a hierarchical kinematic feature recognition strategy. Based on a joint range of motion template library constructed according to the International Society of Biomechanics standards, it contains gold-standard angle curves for 12 basic movement patterns (such as walking, jumping, and squatting). When a 3D joint motion data stream is input, a dynamic time warping (DTW) algorithm is run in real time for pattern matching, simultaneously identifying key phase points (such as the moment of maximum knee flexion) in the three dimensions of knee flexion angle, ankle dorsiflexion angle, and hip abduction angle. The feature verification engine monitors the stability of the angle sequence using sliding window variance analysis (e.g., a window width of 200ms and a step size of 50ms). When the data fluctuation amplitude for five consecutive frames (corresponding to 100ms) is less than a preset bone age dynamic threshold (e.g., ±3° for 12 years old, ±5° for 15 years old), the feature point locking mechanism is triggered, and the joint angle value at that moment is written into the stable feature library. The abnormal data filtering module connects to the bone age-joint range of motion database in real time, uses Mahalanobis distance to calculate the degree of deviation between the current angle and the statistical distribution of the same bone age group, and implements hard elimination of abnormal values ​​beyond the 3σ range.

[0057] In the data processing pipeline, after candidate feature points are output based on a preset kinematic feature template, time series statistics are injected for secondary verification. The two exchange data in milliseconds via a shared memory buffer. When the fluctuation variance of a candidate feature point exceeds a threshold, the template parameter self-correction mechanism is immediately triggered, dynamically adjusting the matching similarity tolerance (from the default 0.8 to 0.6) based on the current motion pattern. The bone age adaptation anomaly filter establishes bidirectional communication with the feature library, not only eliminating anomalous data but also feeding valid features back to the bone age-range of motion database. Using online learning algorithms (such as stochastic gradient descent), it updates the normal angle range for each bone age segment in real time, forming a dynamically evolving human kinematic benchmark.

[0058] Compared with traditional solutions, existing joint angle analysis systems generally face the problem that traditional fixed threshold templates are unable to adapt to individual motion pattern variations, resulting in mismatching of walking and running movements. This solution improves the accuracy of cross-motion type recognition through DTW dynamic regularization and self-correction mechanism; traditional methods use single-frame threshold judgment and are easily interfered by motion artifacts (such as instantaneous displacement of sensors). This solution introduces multi-frame joint variance analysis and combines the dynamic fluctuation range of bone age to reduce the recognition rate of pseudo feature points; existing systems usually perform anomaly filtering at the end of data processing, resulting in invalid features contaminating the intermediate analysis process; this solution constructs a feedforward-feedback dual-path filtering mechanism to intercept abnormal data at the source of feature extraction.

[0059] This application constructs an intelligent extraction system for joint angle features for biomechanical evaluation, and realizes accurate capture and reliable analysis of joint motion features through three-stage collaborative processing of kinematic feature template matching, dynamic stability judgment and bone age adaptation anomaly filtering. In the feature extraction stage, a spatiotemporal consistency guarantee mechanism is established to improve the effective feature capture rate of joint angle data, while compressing the misjudgment rate caused by motion artifacts, providing a high-confidence input source for subsequent biomechanical modeling. Through the deep collaboration of kinematic template matching, dynamic stability verification and bone age intelligent filtering, this solution constructs a joint angle feature extraction system with self-evolution capability. This application not only achieves high-precision feature capture, but also enables the system to adapt to different motion scenarios and individual differences through real-time data verification and parameter optimization, establishing a new generation of feature engineering benchmarks for precise motion biomechanical analysis, and providing reliable technical support for real-time motion intervention decisions.

[0060] The present application further proposes that the biomechanical difference parameters include:

[0061] Bone stress deviation: The bone stress deviation is the relative difference between the real-time bone stress value and the safety threshold matching the bone age data;

[0062] Muscle synergy asymmetry, which is the product of the absolute value of the activation time difference between the left and right target muscle groups and the signal intensity ratio;

[0063] Plantar pressure imbalance: The plantar pressure imbalance is the absolute deviation of the heel-to-forefoot pressure ratio from the standard value.

[0064] Specifically, a comprehensive quantitative assessment of sports risk is achieved through a dynamic coupling analysis of skeletal stress deviation, muscle synergy asymmetry, and plantar pressure imbalance. Skeletal stress deviation reflects the physiological safety limit of skeletal system load, muscle synergy asymmetry characterizes neuromuscular control effectiveness, and plantar pressure imbalance indicates posture stability. These three factors are combined through a weighted fusion model to generate a comprehensive health index, forming a biomechanical assessment framework covering the trinity of "structure, control, and posture."

[0065] Bone stress deviation is calculated in real time using a finite element biomechanical simulation engine. Three-dimensional joint motion trajectory data is input into a personalized bone model adapted for bone age (e.g., a hierarchical mesh model for 12-year-olds and a trabecular microstructure model for 15-year-olds). A dynamic load mapping algorithm is used to determine the relative deviation between the peak stress (Speak) in the growth plate region and the safety threshold (Ssafe) (ΔS = (Speak - Ssafe) / Ssafe × 100%). Muscle synergy asymmetry is assessed using a dual-channel combined time-frequency analysis of electromyographic signals. The activation sequences of the left and right target muscle groups (e.g., the vastus lateralis and semitendinosus) are subjected to a Hilbert-Huang transform. The phase difference (Δt) and energy spectrum ratio (RE) within the 0-50ms time window are extracted to construct the asymmetry index (Masym) = Δt × |1- Ssafe|. The plantar pressure imbalance degree is calculated by integrating the pressure of the plantar zones. The pressure ratio Pratio of the unilateral heel zone (Zone 1-3) and the forefoot zone (Zone 4-6) in the mid-stance phase is calculated and dynamically compared with the standard value Pstd matched with bone age to generate the pressure imbalance degree Dp = |Pratio - Pstd|.

[0066] During the dynamic training cycle, a spatiotemporal feature fusion engine performs joint analysis, using bone stress deviation (ΔS) as a structural risk baseline. When ΔS exceeds 15%, the muscle synergy compensation mechanism is triggered, and the system automatically increases the Masym assessment weight to 0.6. This enhances the sensitivity of muscle synergy monitoring to prevent compensatory injuries. Plantar pressure imbalance (Dp) and ΔS form a composite warning signal. When Dp exceeds 0.2 and ΔS exceeds 10%, a high fall risk state is determined, and the plantar tactile feedback device is activated for immediate posture correction. The three parameters generate 120 feature vectors per second, which are input into the LSTM time series prediction network. The gating unit dynamically adjusts the contribution of each parameter to the comprehensive health indicator, enabling advanced prediction of risk evolution.

[0067] Compared with traditional solutions, existing biomechanical assessment systems mostly use a single-parameter threshold alarm mechanism, for example, only monitoring excessive bone stress or abnormal myoelectric signal amplitude, and cannot identify complex risk scenarios such as multi-system decompensation. However, this solution uses a three-parameter dynamic coupling model to accurately distinguish three typical risk modes:

[0068] Structural overload (ΔS-dominated): When ΔS>20% and Masym<0.4, it is determined to be pure mechanical overload, and a command to reduce training intensity is generated;

[0069] Control compensation (Masym-dominant): When Masym > 0.6 and ΔS < 15%, it is identified as neuromuscular control decompensation, triggering bilateral electromyographic biofeedback training;

[0070] Complex imbalance (three-parameter synchronous growth type): When ΔS, Masym, and Dp exceed the threshold for 5 consecutive seconds, it is judged as a precursor to system collapse, and emergency braking is implemented and an alternative training plan is promoted.

[0071] This solution reduces the misjudgment rate of compound sports injuries, shortens the response time for bone stress warnings, enables proactive intervention through Masym's early compensation signals, and improves the accuracy of identifying the association between plantar pressure imbalance and muscle coordination abnormalities. Through the spatiotemporal coupling and dynamic weight adjustment of three-level parameters, this solution constructs an intelligent assessment system with the ability to perceive the interaction of physiological systems. This enables sports risk identification to leap from traditional single-dimensional threshold judgment to multi-system collaborative state diagnosis, providing a new generation of biomechanical analysis paradigm for precise sports intervention.

[0072] This application further proposes that generating a comprehensive health indicator includes:

[0073] Normalize the skeletal stress deviation, muscle synergy asymmetry, and plantar pressure imbalance;

[0074] The weight coefficient is dynamically adjusted according to the bone age data, and a weighted sum is performed to generate a comprehensive health index:

[0075] Hindex=α×Sdev+β×Masym+λ×|Rp-γ|

[0076] Where Sdev represents skeletal stress deviation, Masym represents muscle synergy asymmetry, and |Rp-γ| represents plantar pressure imbalance. α, β, and λ represent corresponding weighting coefficients, which are dynamically adjusted based on bone age data. Rp represents the actual plantar pressure parameter, and γ is a reference value. The larger the absolute difference between the two, the more severe the plantar pressure imbalance.

[0077] Weight adjustment dynamically configures the fusion coefficient based on bone age data. For bone age <13 years, α=0.6, β=0.25, and λ=0.15 are set to emphasize skeletal development sensitivity. For bone age ≥16 years, the values ​​are adjusted to α=0.4, β=0.35, and λ=0.25, strengthening the weighting of muscle synergy monitoring. The fusion engine calculates the Hindex in real time using a sliding window (e.g., a window width of 2 seconds and a step size of 0.5 seconds). When consecutive windows exceed the threshold, an alert is triggered.

[0078] During data stream processing, the three parameters are precisely aligned using a hardware-level time synchronization framework: skeletal stress data (10ms period) and electromyographic signals (5ms period) are synchronized using the PTP protocol hardware clock, while plantar pressure data (20ms period) is resampled and aligned to a unified time axis via interpolation. A global timestamp generation module appends an absolute time stamp with μs-level accuracy to each data packet, ensuring that the phase deviation of the three parameters within the fusion window is less than 1ms. The dynamic weight regulator interacts with the timestamp engine in real time. When a timing deviation exceeding the tolerance is detected (e.g., a skeletal stress data delay >3ms), the weight coefficient of the affected parameter is automatically reduced (e.g., temporarily lowering α by 20%) to prevent asynchronous data from contaminating the fusion results.

[0079] Compared with traditional solutions, existing health assessment systems generally have three defects. The traditional fixed-weight model cannot adapt to the changes in physiological characteristics of adolescents during the rapid development period, resulting in assessment errors for adolescent users in the rapid development period. This solution improves the consistency of assessment across age groups through dynamic weight mapping of bone age; traditional software-level time synchronization has timing jitter, which causes phase misalignment between pressure peaks and joint angles; this solution combines hardware clock synchronization with interpolation compensation to compress multi-source data alignment errors; traditional Min-Max normalization ignores the distribution characteristics of bone age groups, and the parameters of 12-year-old and 15-year-old users are forced to be scaled to the same range. This solution uses bone age grouping Z-score processing to retain physiological development specificity and improve the sensitivity of abnormality detection.

[0080] This application constructs a comprehensive health assessment system based on the bone age response mechanism, and realizes the accurate quantification of sports health risks through dynamic normalization processing and adaptive weight fusion of multi-source biomechanical parameters. This solution establishes a spatiotemporal consistency guarantee mechanism in the indicator generation stage, so that the three elements of bone stress, muscle synergy and plantar pressure can complete feature fusion under a unified spatiotemporal benchmark, compress the comprehensive evaluation error of multimodal data, and provide a high-reliability decision-making basis for real-time sports intervention. Through the normalization processing of spatiotemporal alignment and bone age-responsive weight fusion, this solution constructs a health assessment engine with physiological adaptability. This application not only breaks through the limitations of traditional static models, but also realizes the accurate fusion of multi-source biomechanical data through the deep collaboration of hardware-level time synchronization and dynamic parameter adjustment, so that sports health risk assessment can leap from single parameter threshold judgment to multi-system joint state diagnosis, and establish a new generation of technical standards for intelligent sports management systems.

[0081] This application further proposes that the method for constructing the preset training path model includes:

[0082] Obtain at least 100,000 adolescent biomechanical experimental data to generate an experimental data set. Analyze the experimental data set using the random forest algorithm, group it, establish the target health indicator baseline value, and determine the allowable deviation Δ:

[0083] When the experimental data bone age is 12-14 years old: the allowable deviation Δ=0.3;

[0084] When the experimental data bone age is 15-18 years old: the allowable deviation Δ=0.2;

[0085] When the actual health indicator exceeds the allowable deviation Δ for five consecutive times, the model self-learning mechanism is triggered to update the allowable deviation Δ.

[0086] Specifically, through parameter normalization, a data set of at least 100,000 adolescent biomechanical experimental data was obtained. The group means and standard deviations of bone stress deviation, muscle synergy asymmetry and plantar pressure imbalance were calculated according to bone age groups. The Z-score algorithm was used to generate standardized parameters to eliminate the influence of bone age differences on parameter amplitudes.

[0087] In this embodiment, the bone age-weight mapping rule is established through random forest model analysis:

[0088] 12-14 years old: α=0.6, β=0.25, λ=0.15;

[0089] 15-18 years old: α=0.4, β=0.35, λ=0.25;

[0090] When the bone age is detected to be in the critical zone (such as 14.5-15.5 years old), the Sigmoid function is used to smooth the transition weight coefficient to avoid sudden changes in the evaluation results.

[0091] The initial allowable deviation Δ is set to 12-14 years old Δ = 0.3, 15-18 years old Δ = 0.2. index When the allowable deviation Δ is exceeded for five consecutive times, the gradient descent optimization algorithm is triggered, and the Δ value is dynamically updated based on the most recent 100 sets of data. The update amplitude is limited to ±0.05 / time to prevent overfitting.

[0092] Compared with traditional solutions, the traditional fixed Δ value (such as the uniform setting of Δ=0.25) cannot adapt to the nonlinear characteristics of bone age development. This solution uses a self-learning mechanism to dynamically adapt the Δ value to the individual development rhythm. Traditional artificial experience ignores the developmental synergy of the muscle-skeletal system. This solution is based on random forest feature importance analysis of 100,000 data points to ensure that the weight distribution is consistent with medical evidence-based basis. Traditional Min-Max normalization forces different bone age parameters to the same range, resulting in the compression of the plantar pressure characteristics of 12-year-old users. This solution uses bone age grouping Z-score processing to retain the true distribution of parameters in physiological development.

[0093] This application constructs a comprehensive health assessment system based on the bone age response mechanism, and realizes the accurate quantification and dynamic tracking of sports health risks through dynamic normalization processing of multi-source biomechanical parameters, adaptive weight fusion and deviation self-learning optimization. This solution establishes a strong coupling relationship between physiological development characteristics and dynamic weights in the indicator generation stage, and realizes the continuous evolution of the assessment model through the self-learning mechanism, which improves the consistency of cross-age assessment of health indicators and provides high-precision decision support for personalized sports management. Through the deep collaboration of multi-source parameter spatiotemporal alignment, bone age-driven weight fusion and deviation self-learning optimization, this solution constructs an intelligent assessment system with physiological adaptability.

[0094] This application further proposes that the dynamic comparison and generation of adjustment instructions include:

[0095] Apply sliding window analysis to the comprehensive health indicators in the time series data, calculate the actual change rate of the health indicators within the window period, and compare it with the target change rate of the corresponding stage in the preset training path model;

[0096] When the actual rate of change continues to exceed the allowable deviation range of the target rate of change, a graded adjustment instruction is generated according to the deviation direction.

[0097] This application builds an intelligent training path management system based on the evolution of real-time health indicators. Through the synergy of sliding window dynamic monitoring, deviation elasticity assessment, and model self-learning optimization, it achieves precise dynamic control of training plans. This solution breaks through the limitations of traditional static threshold management and establishes an adaptive decision-making system with evolutionary capabilities. This shortens the response time for training path deviation warnings, improves the accuracy of dynamic adjustments, and provides real-time closed-loop management support for scientific training for young people.

[0098] Specifically, after generating the comprehensive health index, a sliding window analysis is applied to the comprehensive health index data in the time series. During the sliding window analysis, the actual rate of change of the health index within the window period is calculated. The calculation of the actual rate of change is based on the change of the comprehensive health index within the window period. It is then compared with the target rate of change of the corresponding stage in the preset training path model. The preset training path model is obtained based on a large amount of health data training, where the target rate of change of the corresponding stage is the ideal reference value for the change of the health index in that stage.

[0099] When the actual rate of change consistently exceeds the allowable deviation range of the target rate of change, graded adjustment instructions are generated based on the direction of the deviation. These graded adjustment instructions are determined based on the direction and degree of deviation. Different levels of deviation and direction correspond to different levels of adjustment instructions. These adjustment instructions can be used for subsequent health interventions and rehabilitation planning, enabling effective monitoring and precise intervention of individual health status, ensuring that individuals' health status remains within or progresses toward the desired range. The system utilizes a dual-engine collaborative architecture and deploys a sliding window analyzer (e.g., with a window width of 30 seconds and a step size of 5 seconds) to continuously calculate the slope Kactual of the health indicator Hindex within the window period. A Kalman filter is used to eliminate motion artifacts. The graded adjustment mechanism is triggered when the actual rate of change consistently exceeds the allowable deviation range of the target rate of change, for example, when the absolute deviation |ΔK| between Kactual and the preset path target slope Ktarget for three consecutive windows exceeds Δ.

[0100] Compared with traditional solutions, the traditional batch processing mode (calculation once per minute) leads to deviation response delays; this solution uses streaming sliding window analysis to increase the monitoring frequency; traditional threshold adjustment relies on manual experience and has medical rationality risks. This solution constructs a clinical knowledge graph verification module to ensure that each △ value update complies with orthopedic safety rules.

[0101] Through real-time tracking with a sliding window, deep collaboration of elastic deviation thresholds and enhanced self-learning, dynamic adjustment of instructions improves training efficiency and activates the accuracy of the high-intensity action library. The elastic threshold regulator compresses the Δ value optimization cycle to adapt to the needs of the growth spurt period. The closed-loop feedback mechanism reduces the deviation rate of training plan execution and reduces the incidence of sports injuries. This application builds an intelligent training management center with evolutionary capabilities, breaking through the static management model of traditional solutions and realizing a full-link closed loop of "millisecond-level monitoring-intelligent decision-making-medical verification-model evolution".

[0102] This application further proposes that the triggering logic of the hierarchical adjustment instruction includes:

[0103] Obtaining the deviation direction and determining the deviation direction;

[0104] If it is a positive deviation (actual change rate > target change rate + Δ):

[0105] When the magnitude of the positive deviation is within a first threshold range, generating an instruction to shorten the training interval;

[0106] When the magnitude of the positive deviation is within a second threshold range, an instruction is generated to advance to the next training phase and increase the intensity;

[0107] If it is a negative deviation (actual change rate < target change rate - Δ):

[0108] When the magnitude of the negative deviation is within a first threshold range, generating an instruction to extend the training interval;

[0109] When the magnitude of the negative deviation is within the second threshold range, an instruction to forcibly switch to an alternative action library is generated, where the alternative action library is selected from historical low-risk actions.

[0110] This application builds an adaptive training management system based on real-time biomechanical state perception. Through a triple-coordinated mechanism of graded deviation identification, flexible command generation, and medical safety verification, it achieves precise dynamic control of the training process. This solution breaks through the traditional static threshold management paradigm and establishes an intelligent decision-making system with real-time response and self-learning capabilities. This improves the medical compliance of training adjustment commands and reduces the command execution response speed to 800ms, providing closed-loop management support for scientific training for young people.

[0111] Specifically, in one embodiment, real-time deviation analysis deploys a sliding window (window width 30 seconds, step length 5 seconds) to continuously calculate the change rate Kactual of the health indicator Hindex, and eliminates motion artifacts through a Kalman filter. When it is detected that Kactual exceeds the target change rate Ktarget±△ for three consecutive times, a hierarchical decision tree is triggered:

[0112] When the deviation is positive (Kactual >Ktarget +△):

[0113] Level 1 instruction (first threshold interval: △ < deviation ≤ 1.5△): Dynamically shorten the training interval Δt_new = Δt × [1-0.2 × (deviation / △)], and the minimum interval is not less than the safety threshold Δtmin = 45 seconds;

[0114] Secondary instruction (second threshold interval: deviation > 1.5△): Activate the next stage training protocol in advance, intensity increase coefficient β = 1 + 0.15 × (deviation / △), and simultaneously inject high-difficulty movement library (risk coefficient < 0.3);

[0115] When the deviation is negative (Kactual < Ktarget-△):

[0116] Level 1 instruction (first threshold interval: △ < deviation ≤ 1.5△): Extend the training interval Δt_new = Δt × [1 + 0.3 × (deviation / △)], with the maximum interval not exceeding Δt_max = 120 seconds;

[0117] Secondary instruction (second threshold interval: deviation >1.5△): Forced switching to the alternative action library, building an alternative solution pool based on historical low-risk actions (injury incidence <5%), and using cosine similarity to match the current movement pattern (similarity >85%).

[0118] Medical safety verification integrates a clinical knowledge graph (including orthopedic safety rules) to perform real-time compliance checks on generated adjustment instructions. If a risk is detected (e.g., the predicted bone stress after strength improvement exceeds Ssafe × 1.2), the instruction correction mechanism is immediately triggered, generating an alternative solution within 50ms using a reinforcement learning model.

[0119] During the operation cycle, the dual engines achieve deep collaboration through a high-speed data bus. The biomechanical sensor array transmits Hindex time series data to the deviation analysis engine at a frequency of 200Hz, and at the same time transmits real-time stress values, myoelectric coordination and other original parameters to the safety verification engine; the correction instructions of the safety verification engine are transmitted back to the deviation analysis engine through the priority interrupt channel, triggering dynamic adjustment of decision tree parameters (such as reducing the sensitivity of the △ value); after the daily training, the system automatically analyzes the instruction execution effect data (such as injury incidence, training efficiency improvement rate), and uses the Q-learning algorithm to optimize the classification threshold (△ value) and intensity adjustment coefficient. The update range is limited to ±0.05 / day to ensure that the evolution process complies with the medical safety boundary.

[0120] Through the aforementioned technical solution, which integrates hierarchical deviation identification, real-time safety verification, and model self-learning, this solution builds an intelligent training and control center with evolutionary capabilities. This breaks through the static management constraints of traditional solutions and achieves a fully closed loop of "millisecond-level perception - intelligent decision-making - medical support - dynamic evolution."

[0121] like Figure 3 As shown, the present application further proposes that the real-time monitoring system for physical fitness training of teenagers also includes a path model updating module, which is used to update the preset training path model according to the training parameter adjustment amount to maintain the comprehensive health index within the allowable deviation range of the training path model;

[0122] Among them, the preset training path model includes target health indicators, training cycle time nodes and allowable deviation ranges preset for each bone age stage, which serves as a dynamic benchmark for the actual training process.

[0123] The preset training path model includes target health indicators, training cycle time nodes, and tolerance ranges for each bone age stage. During actual training, target health indicators are set based on the characteristics of each bone age stage, and the training cycle time nodes define the training progress and status to be achieved at each stage. The tolerance range provides flexibility throughout the training process, ensuring that the system does not misjudge due to minor fluctuations.

[0124] During the training process, the preset training path model is updated in real time based on the training parameter adjustments. For example, when certain parameters change during actual training, the relevant parts of the preset training path model are adjusted based on the specific situation. For example, if it is found that the training progress of a certain bone age stage is ahead of or behind the preset training cycle time node, the subsequent time node arrangement and the setting of the target health indicator are adjusted accordingly to keep it within the allowable deviation range.

[0125] Furthermore, when generating the comprehensive health index, bone stress deviation, muscle synergy asymmetry, and plantar pressure imbalance are first normalized, making these different metrics comparable and compatible. Weight coefficients are then dynamically adjusted based on bone age data, and a weighted summation is performed to generate the comprehensive health index. The dynamically adjusted weight coefficients and changes in the comprehensive health index are fed back to the pathway model update module, providing important basis for updating the pre-set training pathway model.

[0126] Through the close interaction and collaborative work of the above-mentioned parts, the entire system can effectively and dynamically adjust the training path model according to the actual training situation, so that the comprehensive health indicators are always maintained within the allowable deviation range, thereby providing a high-precision and highly adaptable dynamic benchmark for training, significantly improving the effectiveness and quality of training.

[0127] The present application further proposes that updating the preset training path model includes:

[0128] Model parameter correction: adjust the time node interval and allowable deviation range of subsequent stages according to the actual health indicator trend;

[0129] Generate a version ID, record the model change time, effective stage ID, and adjusted deviation range, and associate it with historical training data;

[0130] Closed-loop control is executed, and the updated path model is written to the edge computing device as a real-time comparison benchmark for the next monitoring cycle.

[0131] Model parameter correction refers to the dynamic adjustment of the time node intervals and allowable deviation ranges for subsequent training phases based on real-time trends in actual health indicators (such as real-time risk coefficient, muscle synergy asymmetry, dynamic pressure load index, etc.). Specifically, this can be achieved by using online learning algorithms in machine learning (such as stochastic gradient descent). By calculating the deviation between the actual and target indicators (Δ = actual value - target value), combined with the statistical patterns of historical training data (such as the time series distribution of deviations), the time node intervals (such as shortening the original 3-day / phase to 2-day / phase) and the allowable deviation range (such as relaxing ±10% to ±15%) can be automatically optimized. For example, if the risk coefficient of an adolescent is monitored to be continuously below the target value for two consecutive weeks, the system will shorten the time node intervals for the next phase to accelerate training progress. If the risk coefficient frequently exceeds the original deviation range, the allowable deviation will be increased to improve the model's inclusiveness.

[0132] Version ID generation involves generating a unique version ID for each model change. This ID records the model change time, the effective phase identifier (e.g., "Phase 3_v2"), and the adjusted deviation range. This ID is then associated and stored with historical training data (e.g., risk coefficients and stress ratios for the previous 30 days) using a hashing algorithm. Specifically, the version ID includes a timestamp (accurate to milliseconds), the phase number (e.g., "T3" for the third phase), the adjustment type (e.g., "time node shortening"), and a checksum (to ensure data has not been tampered with). Blockchain technology is then used to bind this version information to the corresponding historical data block, forming a traceable chain of model changes.

[0133] Closed-loop control execution refers to writing the updated path model to edge computing devices (such as smart bracelets and treadmill controllers) via a low-latency communication protocol (such as MQTT) as a real-time comparison benchmark for the next monitoring cycle. After receiving the new model, the edge device immediately switches to the latest version of the time node and deviation range to dynamically evaluate the health indicators collected in real time. For example, when the edge device receives the "Phase 3_v2" model (the time node interval is shortened to 2 days and the deviation is allowed to be relaxed to ±15%), it will use the parameters of this model in subsequent monitoring to calculate risk factors and trigger interventions to ensure that the training path is synchronized with the actual development status.

[0134] Specifically, when the system detects that a teenager's actual risk factor has fallen below the original model's target value (deviation Δ = -0.2) for five consecutive days, the model parameter correction module, using an online learning algorithm, determines that the time interval between the next phase of training should be shortened from three days to two days, and the tolerance range should be adjusted from ±10% to ±12%. The version identifier generation module then generates a version number, "T4_v1," recording the change time, effective phase (Phase 4), and adjusted parameters. This version is then linked to the previous 30 days' risk factor data and stored on the blockchain using a hash algorithm. The closed-loop control execution module pushes the "T4_v1" model to the teenager's smart wristband via the MQTT protocol. Upon receiving the model, the wristband immediately updates its locally stored training path parameters. Subsequent monitoring uses the new time intervals (evaluated every two days) to compare in real time with the tolerance range (±12%) to ensure that training progress aligns with the teenager's actual fitness improvement rate.

[0135] Compared with existing technologies, traditional preset training path models use static parameters (such as a fixed 3 days / stage, ±10% deviation), which cannot be dynamically adjusted according to individual training results, leading to problems such as "training progress is too slow" or "risk assessment is too strict." This solution achieves adaptive optimization of the training path through model parameter correction (such as accelerating or slowing down progress based on the actual risk factor), ensures the traceability of model changes through version identification generation (avoiding evaluation confusion caused by incorrect parameter changes), and achieves seamless integration of model updates and monitoring through closed-loop control execution (edge ​​devices apply new models in real time). In existing technologies, model updates require manual intervention and have high latency (usually more than 24 hours). This solution shortens the model update cycle to minutes through online learning and low-latency communication, improving the response speed of parameter adjustments.

[0136] Through the above technical solutions, this application effectively solves the problem of disconnection between the preset training path model and the actual training effect. Model parameter correction dynamically optimizes the training progress according to real-time data, version identification generation ensures the traceability and security of model changes, and closed-loop control execution realizes the real-time implementation of model updates. The matching degree between the training path and the physical development of adolescents is improved, the misjudgment rate of risk assessment is reduced, and the scientificity and safety of adolescent physical training are significantly improved. Through real-time interaction between actual training data and the model, adaptive adjustment of the training path is achieved, solving the technical problem that traditional static models cannot adapt to individual developmental differences and dynamic training scenarios.

[0137] like Figure 4 As shown, this application further proposes that the real-time monitoring system for physical fitness training of teenagers also includes a biomechanical digital twin module:

[0138] Generate a dynamic simulation model of bones and muscles in a virtual environment based on real-time joint 3D motion data and plantar pressure distribution data;

[0139] The stress distribution trend of the next 10 training cycles is predicted through finite element analysis and superimposed with the target value of the preset path model;

[0140] When the predicted trend deviates from the target value by more than a preset deviation threshold, a pre-adjustment instruction is generated in advance and pushed to the processing terminal.

[0141] Specifically, the system collects real-time 3D joint motion data and plantar pressure distribution data, creating a dynamic bone-muscle simulation model in a virtual environment (e.g., a computer-simulated 3D scene) to visually demonstrate the dynamic changes in bones and muscles during exercise. Using finite element analysis (FEA), based on the current motion data and model, it predicts the stress distribution trends in bones and muscles over the next 10 training cycles (a training cycle can be a specific time period, such as physical training or rehabilitation training). The system also overlays the predicted stress distribution trends with the target values ​​(predetermined, desired stress distribution states) of a pre-set path model, visually comparing the differences between the predicted trends and the target values. The system also sets a preset deviation threshold (i.e., the maximum allowable deviation between the predicted trend and the target value). If the predicted stress distribution trend deviates from the target value by more than this threshold, it indicates that the current exercise state is unfavorable for training effectiveness or may cause damage to the body. In this case, the system generates pre-adjustment instructions (such as suggestions for adjusting training intensity or changing exercise posture) and pushes these instructions to processing terminals (such as computers, mobile phones, and smart wearable devices) for convenient access and action by relevant personnel.

[0142] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0143] By deploying a lightweight skeletal-muscle multibody dynamics model, the virtual skeletal system is driven in real time based on 3D joint motion data (200Hz) and plantar pressure distribution (500Hz). An improved inverse dynamics algorithm (InverseDynamics) is used to calculate skeletal joint torques. Muscle force-length relationship parameters are dynamically adjusted based on the activation state of electromyographic signals (1000Hz), achieving microsecond-level biomechanical state synchronization. GPU-accelerated finite element analysis (FEA) uses a time series extrapolation algorithm to generate stress evolution cloud maps for the next Y (default Y = 30) training cycles based on the current skeletal stress distribution and motion trajectory characteristics. Adaptive meshing technology (minimum mesh size 0.1mm) dynamically optimizes computing resource allocation, keeping the prediction time for a single 30-cycle prediction to less than 2 seconds. A pre-adjustment strategy library based on deep reinforcement learning (DRL) is built. When the deviation between the predicted stress trend and the target path exceeds a threshold (Δ>0.15), the Monte Carlo Tree Search (MCTS) algorithm is activated to generate an optimized instruction set including training intensity correction, action alternatives, and recovery cycle adjustment within 500ms, and push it to the training terminal in real time through the edge computing node.

[0144] During the operational cycle, a high-speed data bus enables deep interaction with the model interface. The biomechanical sensor array transmits six-degree-of-freedom joint positions, plantar pressure thermograms, and electromyographic activation waveforms to the twin engine with microsecond precision. Hardware-level clock synchronization (PTP protocol) ensures spatiotemporal alignment error of multi-source data within <0.5ms. The virtual skeletal system updates its dynamic state every 5ms, generating a holographic digital image of the biomechanical parameters. The finite element engine receives a snapshot of the current bone stress field every 30 seconds. It uses a spatiotemporal convolutional network (ST-Conv) to extrapolate future stress distribution trends and dynamically overlays and compares the predicted results with the target path model. When the predicted stress value in a critical growth plate region (such as the proximal tibia) exceeds a safety threshold (Ssafe × 1.25), a pre-decision process is immediately triggered, generating adjustment instructions three training cycles ahead. The pre-adjustment instruction set is directly connected to the edge computing node via the RDMA protocol, completing safety verification (including clinical rule verification and execution effect simulation) within 150ms. Verified instructions are injected into the training control system in the form of a priority queue to ensure that intervention measures are initiated two training cycles before the target deviation actually occurs.

[0145] Compared with traditional approaches, which only monitor the current biomechanical state and fail to predict future risks, resulting in a lack of early warning of bone stress injury events, this approach advances the risk identification window through finite element extrapolation, improving the success rate of early warnings. Traditional manual adjustment methods require intervention after injury indicators appear, while this approach implements automated pre-decision responses, enabling earlier intervention. Traditional inverse dynamics algorithms can lead to joint torque calculation errors; this approach reduces these errors through a muscle activation compensation mechanism driven by electromyographic signals.

[0146] By deeply integrating real-time digital twin construction, multi-physics field prediction and deduction, and intelligent preemptive decision-making, a training management system with future-forecasting capabilities has been constructed. This technology breaks through the technical boundaries of traditional monitoring systems, realizing a new management model of "real-time mirroring - future deduction - preemptive control." This elevates scientific training for youth from a passive response to a proactive prevention approach, setting a new benchmark for intelligent preemptive decision-making technology in the field of sports health, enabling proactive prevention and control of training risks and adaptive optimization of training plans.

[0147] The following is a complete embodiment of the real-time monitoring system for physical fitness training of teenagers of this application:

[0148] A teenager wore an inertial sensor array (sampling rate 200Hz) to accurately record their joint motion during basketball training. During a pull-up jump shot, the sensors captured real-time hip abduction (peak 58°±2°), knee flexion (maximum 112°±3°), and ankle dorsiflexion (dynamic range 25°-35°), fully capturing the characteristics of their movement. Surface electromyography electrodes (Noraxon DTS system, sampling rate 1500Hz) were attached to the rectus femoris and lateral head of the gastrocnemius muscles. During squat training, the system recorded activation latency of the left rectus femoris (Δt=18ms) and maximum contraction strength of the right gastrocnemius (RMS=1.2mV), revealing bilateral muscle force imbalances. A piezoresistive smart insole (Tekscan F-Scan system, 100 sensing points / cm²) monitored plantar pressure distribution during directional changes. Data showed a peak pressure of 450 kPa in the left heel area (380 kPa in the right foot), with a forefoot pressure ratio deviation ΔP of 0.23. Carpal bone development was assessed using the EOS low-dose 3D imaging system. The TW3 bone age scoring system determined BA = 14.3 years, and skeletal maturity parameters were determined using a physiological development database.

[0149] Multi-source data was synchronized at the hardware level (error <1ms) using the IEEE 1588 PTP protocol. Dynamic time warping (DTW) was applied to phase-align the knee angle curve (200Hz), the rectus femoris EMG envelope (downsampled from 1500Hz to 200Hz), and the heel pressure time series (100Hz). Based on a standard finite element model of a 14-year-old male skeleton (mesh accuracy 0.5mm), joint torque data were input to calculate dynamic stress distribution in the proximal tibial growth plate. During continuous jump training, peak stress reached 28 MPa (safety threshold Ssafe = 32 MPa). Hilbert-Huang transform (IMF component = 5) was performed on bilateral rectus femoris EMG signals to extract time-frequency features. The synergy asymmetry index (Masym) was calculated to be 0.18 (Δt = 22ms, intensity ratio R = 0.85). The comprehensive health index was calculated using a dynamic weighting formula: Hindex = 0.5×(28 / 32) + 0.3×0.18 + 0.2×0.23 = 0.63 (weighting coefficients for a bone age of 14 years: α=0.5, β=0.3, λ=0.2). The pre-set training path model indicated a target Hindex of 0.68±0.15 for week 3. The current value of 0.63 triggered a Level 2 deviation warning (Δ=0.05, persisting for three analysis windows). Adjustment instructions were generated: Training intensity adjustment: Increase squat load factor β=1.1 (from 1.0); Exercise substitution: Enable low-impact exercise library (lunge → wall squat, 89% similarity); Recovery period optimization: Increase inter-set interval to 90 seconds (from 60 seconds).

[0150] Finite element analysis predicted the cumulative tibial stress trend over the next five training sessions (10 cycles), indicating that the safety threshold (predicted value 34 MPa) would be exceeded by the eighth training session. The system issued adjustment recommendations two cycles in advance: introducing cushioning insoles (reducing heel pressure by 18%) and limiting the number of consecutive jumps to ≤ 15 per set.

[0151] When the actual Hindex deviates by >0.12 for three consecutive times, incremental learning is triggered with Δnew=0.15×[1+0.1×(0.63-0.68)]=0.1425. After the update, the allowed deviation Δ is adjusted from 0.15 to 0.1425.

[0152] This example demonstrates the complete application of the system in youth basketball training. Through the synergy of multi-dimensional biomechanical monitoring, intelligent dynamic adjustment, and advanced risk prediction, it achieves precise monitoring of errors in the simultaneous analysis of joints, muscles, and plantar structures. A bone age-driven model reduces assessment errors for 14-year-old users, advancing the window for identifying sports injury risks. Blockchain-encrypted transmission ensures zero data leakage.

[0153] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A real-time monitoring system for physical fitness training of teenagers, characterized by: include: The data acquisition module is used to receive the three-dimensional joint motion data, electromyographic signal data and plantar pressure distribution data of the current training stage and the bone age data of the physiological development stage; A standardized feature library, for simultaneously acquiring a standardized biomechanical feature data set of a historical training phase, wherein the feature data set includes a joint motion amplitude threshold, a muscle synergy symmetry range, and a standard value of plantar pressure distribution that matches the bone age data; a data processing module, configured to extract joint angle feature points from the three-dimensional joint motion data, the activation time series of the electromyographic signal data, and the pressure area of ​​the plantar pressure distribution data, and perform spatiotemporal alignment processing with the feature data set to generate feature data; a health index generation module, which calculates a plurality of biomechanical difference parameters reflecting bone stress deviation, muscle synergy asymmetry, and plantar pressure imbalance based on the aligned characteristic data; generating a comprehensive health index based on the biomechanical difference parameters and recording the index as time series data; Dynamic path decision module, including: Dynamically compare the change trend of the comprehensive health indicator in the time series data with the change trend of the target health indicator in the corresponding stage in the preset training path model to generate a deviation direction and amplitude; the deviation direction is divided into positive deviation and negative deviation, the positive deviation is used to increase the training amplitude, and the negative deviation is used to weaken the training amplitude; generating an adjustment instruction according to the deviation direction and magnitude; An execution module is used to generate, according to the adjustment instruction, a training parameter adjustment amount including training intensity adjustment, time interval correction and action replacement scheme through a preset rule mapping table; a path model updating module, configured to update the preset training path model according to the training parameter adjustment amount to maintain the comprehensive health indicator within an allowable deviation range of the training path model; The generating of comprehensive health indicators includes: Normalizing the bone stress deviation, the muscle synergy asymmetry, and the plantar pressure imbalance; The weight coefficient is dynamically adjusted according to the bone age data, and the weighted sum is performed to generate the comprehensive health index: H index =α×S dev +β×M asym +λ×|R p -c| Among them S dev Represents the bone stress deviation, M asym represents the muscle synergy asymmetry and |R p -γ| represents the plantar pressure imbalance, α, β, and λ represent corresponding weight coefficients, which are dynamically adjusted according to the bone age data; The method for constructing the preset training path model includes: Acquire experimental data to generate an experimental data set, analyze the experimental data set using a random forest algorithm, group the data set, establish target health indicator benchmark values, and determine the allowable deviation Δ: When the experimental data bone age is 12-14 years old: the allowable deviation Δ=0.3; When the experimental data bone age is 15-18 years old: the allowable deviation Δ=0.2; When the actual health indicator exceeds the allowable deviation Δ for five consecutive times, the model self-learning mechanism is triggered to update the allowable deviation Δ; The preset training path model includes target health indicators, training cycle time nodes and allowable deviation ranges preset for each bone age stage, which serve as a dynamic benchmark for the actual training process.

2. The real-time monitoring system for physical fitness training of teenagers according to claim 1 is characterized in that: Extracting the joint angle feature points includes: Based on a preset kinematic feature template, identifying key angle values ​​of the knee flexion angle, ankle dorsiflexion angle, and hip abduction angle from the three-dimensional joint motion data; Perform timestamp alignment on the key angle values, and mark them as stable feature points when the angle fluctuation of 5 consecutive frames of data is less than the fluctuation interval; Abnormal angle data that exceeds the bone age data matching amplitude threshold range is eliminated.

3. The real-time monitoring system for physical fitness training of teenagers according to claim 1 is characterized in that: The biomechanical difference parameters include: Bone stress deviation, which is the relative difference between the real-time bone stress value and the safety threshold matching the bone age data; Muscle synergy asymmetry, where the muscle synergy asymmetry is the product of the absolute value of the activation time difference between the left and right target muscle groups and the signal intensity ratio; Plantar pressure imbalance, where the plantar pressure imbalance is the absolute deviation of the lateral heel-to-forefoot pressure ratio from a standard value.

4. The real-time monitoring system for physical fitness training of teenagers according to claim 1 is characterized in that: Dynamically comparing and generating the adjustment instruction includes: Applying sliding window analysis to the comprehensive health indicators in the time series data, calculating the actual rate of change of the comprehensive health indicators within the window period, and comparing it with the target rate of change of the corresponding stage in the preset training path model to generate the deviation direction and magnitude; When the actual change rate continues to exceed the allowable deviation range of the target change rate, a graded adjustment instruction is generated according to the deviation direction.

5. The real-time monitoring system for physical fitness training of teenagers according to claim 4 is characterized in that: The triggering logic of the hierarchical adjustment instruction includes: Obtain the deviation direction and determine whether the deviation direction is the positive deviation. The positive deviation is when the actual change rate is greater than the sum of the target change rate and the allowable deviation Δ. If so, make the following decision based on the positive deviation amplitude: When the magnitude of the positive deviation is within a first threshold range, generating an instruction to shorten the training interval; When the magnitude of the positive deviation is within a second threshold range, generating an instruction to enter the next training phase in advance and increase the intensity; Otherwise, it is determined to be a negative deviation. The negative deviation is when the actual change rate is less than the difference between the target change rate and the allowable deviation Δ. The following decision is made based on the magnitude of the negative deviation: When the magnitude of the negative deviation is within a first threshold range, generating an instruction to extend the training interval; When the magnitude of the negative deviation is within a second threshold range, an instruction to forcibly switch to an alternative action library is generated, where the alternative action library is selected from historical low-risk actions.

6. The real-time monitoring system for physical fitness training of teenagers according to claim 1 is characterized in that: Updating the preset training path model includes: Model parameter correction, which adjusts the time node intervals and allowable deviation ranges of subsequent stages based on actual health indicator trends; Version identification generation, which records the model change time, effective stage identification and adjusted deviation range, and is associated with historical training data; The closed-loop control is executed, and the closed-loop control writes the updated path model to the edge computing device as a real-time comparison benchmark for the next monitoring cycle.

7. The real-time monitoring system for physical fitness training of teenagers according to claim 1 is characterized in that: The real-time monitoring system for physical fitness training of teenagers also includes a biomechanical digital twin module: generating a skeletal-muscle dynamic simulation model in a virtual environment based on the real-time three-dimensional joint motion data and the plantar pressure distribution data; The stress distribution trend of the next 10 training cycles is predicted through finite element analysis and superimposed with the target value of the preset path model; When the predicted trend deviates from the target value by more than a preset deviation threshold, a pre-adjustment instruction is generated in advance and pushed to the processing terminal.

Images