An intelligent sports big data analysis management method and system based on artificial intelligence

Abstract

This invention relates to the field of sports data analysis and discloses a smart sports big data analysis and management method and system based on artificial intelligence. The method includes: real-time collection of exercise data from multiple data sources on the physiological and / or physical aspects of sports users; processing and fusing the collected individual exercise data from each source to form fused data; calculating the correlation coefficients between different source data and analyzing the correlation between the source data; using multi-order differential or fast Fourier transform operations to calculate the dynamic changes of each source data and analyze the health status of sports users; forming a dataset from the fused data, correlated data, and dynamically changing data; comparing the historical data and current data of sports users to analyze relevant data on the physical fitness, training status, and competitive level of sports users, and assessing the physical fitness and competitive status of sports users; and providing health management suggestions. This achieves comprehensive analysis of health status from multiple source data.

Description

Technical Field

[0001] This invention relates to the field of sports data analysis technology, and in particular to a smart sports big data analysis and management method and system based on artificial intelligence. Background Technology

[0002] Competitive sports involve a variety of factors, including the athlete's physical abilities. Achieving optimal results requires a comprehensive consideration of these factors. Advances in data collection technology have led to an increasing variety of competitive data collected from athletes, and the analysis of this data has become more detailed, helping athletes achieve better performance. However, the integration, cleaning, and processing of multi-source data often rely on manual methods or single algorithms, lacking intelligent cross-platform fusion capabilities. This approach results in data redundancy, information loss, or low analytical accuracy, making it difficult to provide comprehensive real-time health and performance analysis.

[0003] Traditional systems often separate health management from performance analysis, lacking a comprehensive assessment. While health management recommendations are important, they are not adequately integrated with performance status and competition strategy development, failing to provide athletes with all-round intelligent support.

[0004] Existing sports data analysis systems typically focus on single-source data, such as heart rate and weight, failing to effectively uncover the correlations between different data points and derive assistance from these correlations. Traditional systems often overlook the deeper relationships between different sports data, making it difficult to identify potential health risks or opportunities for performance optimization.

[0005] Furthermore, many existing systems only provide static health advice or technical analysis, lacking dynamic decision support for competition. Athletes often lack real-time and accurate analysis of multiple factors such as physical fitness, strategy execution, and psychological state during competitions.

[0006] Therefore, how to conduct correlation analysis on the competitive data of sports users based on multi-source data is an urgent problem to be solved. Summary of the Invention

[0007] The purpose of this invention is to provide at least one intelligent sports big data analysis and management method and system based on artificial intelligence, which can at least solve the problem of correlation between multi-source data, and at least achieve the goal of analyzing the health status of sports users based on multi-source data, formulating competition strategies, and conducting personalized health management.

[0008] To address the aforementioned technical problems, at least one embodiment of this application provides an artificial intelligence-based smart sports big data analysis and management method, comprising: real-time collection of exercise data from multiple data acquisition sources of a sports user's physiological and / or exercise data; processing the collected single exercise data from each source to obtain standardized single data from each source; fusing the standardized single data from multiple sources to form fused data; calculating the correlation coefficients between different source data based on the different source data in the fused data, and analyzing the correlation between the source data; using multi-order differential or fast Fourier transform (FFT) operations to calculate the dynamic changes of each source data, analyzing the health status of the sports user, and identifying abnormal trends; forming a dataset from the fused data, correlated data, and dynamically changing data; comparing the sports user's historical data and current data; analyzing the relevant data of the sports user's physical fitness, training status, and competitive level based on the comparison results, and assessing the sports user's physical fitness and competitive status; and proposing health management suggestions and training suggestions based on the analysis results.

[0009] At least one embodiment of this application also provides an artificial intelligence-based smart sports big data analysis and management system, including: a data acquisition module, a data storage module, a data analysis module, a neural network module, and a management suggestion module; the data acquisition module includes various types of sensors, video data acquisition units, and environmental data acquisition units, and performs cleaning, noise reduction, correction, structured processing, and fusion on the acquired raw data to obtain fused data; the data storage module is used to store different data in different regions; the data analysis module is used to perform parameter correlation analysis, dynamic change analysis, and historical data analysis on the data to obtain the correlation between different data, and, combined with historical data, identify abnormal states and assess health status and competitive level; it also performs match performance analysis, real-time tactical execution analysis, physical fitness and strategy analysis, psychological state and technical action analysis, and opponent and environmental data analysis on the match data to assess physical fitness, competitive level, and match status; the neural network module is used to correlate input parameters with output competitive performance, analyze the psychological state and technical actions of the sports user, the match performance of the opponent, and environmental factors, and generate intelligent decisions; the management suggestion module is used to provide personalized health management suggestions, match and training suggestions.

[0010] At least one embodiment of this application also provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the above-described intelligent sports big data analysis and management method based on artificial intelligence.

[0011] At least one embodiment of this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described intelligent sports big data analysis and management method based on artificial intelligence.

[0012] The embodiments of this application provide an artificial intelligence-based smart sports big data analysis and management method and system, which processes and integrates multi-source data, analyzes the correlation coefficients between different source data, and integrates different data to provide a foundation for multi-source data analysis of sports data. It realizes the analysis of health status based on multi-source data and obtains health management suggestions based on multi-source data.

[0013] In some optional embodiments, an AI-based smart sports big data analysis and management method further includes: collecting competition data from multiple data acquisition sources of sports users, processing the data to obtain individual data from each source; storing the types of sports data and competition data in different regions, performing data analysis and / or neural network model analysis on the individual data from each source, obtaining data performance of sports users in the competition, the correlation between physical fitness and strategy execution, the influence of psychological state on technical actions, and the influence of opponents and venue on the competition results, generating intelligent decisions, and proposing health management suggestions, training suggestions, and competition response strategies; The system is divided into four zones: the first zone stores competition-related indicators to identify the competitive state and performance of athletes during the competition; the second zone stores athletes' physical fitness and strategy execution data to provide suggestions for energy allocation and strategy adjustment; the third zone stores athletes' psychological state data to analyze the impact of athletes' psychological state on technical movements and provide suggestions for psychological adjustment; and the fourth zone stores publicly available data on opponents and competition venue data to analyze the impact of referee bias and venue environment on the competition results.

[0014] It processes competition data and formulates strategies based on the data, providing health management and training / competition suggestions, thus combining health management with competitive performance and providing comprehensive analytical support for sports users.

[0015] By storing different data in different regions and performing different analyses based on the data in each region, the completeness and comprehensiveness of the analysis are improved.

[0016] In some optional embodiments, the plurality of data acquisition sources, including cameras, sensors, wearable devices, and physical examination devices, are used to acquire the physiological and exercise data and competition data of the exercise user, including: heart rate, respiratory volume, weight, body fat, stride, exercise speed, exercise intensity, real-time competition data, venue data, and biosignal data.

[0017] By obtaining data from different data sources and from different dimensions, the more data obtained, the more comprehensive the analysis can be, resulting in more general analytical results and improving the accuracy of the analysis.

[0018] In some optional embodiments, the data processing includes: data cleaning, denoising, correction, structuring, and fusion. The data cleaning and denoising are used to remove abnormal data from the dataset, and the correction is used to correct systematic errors during the data cleaning process. The structuring is used to unify and standardize the format of multi-source data. The fusion is used to merge the structured multi-source data according to a fusion mode to form fused data, wherein the fusion mode includes a regular mode and a cross mode.

[0019] The data processing achieved standardization and unification of multi-source data, providing a basis for subsequent data analysis.

[0020] In some optional embodiments, the analysis of the correlation between motion data from different sources includes: using the Pearson correlation coefficient or the Spearman correlation coefficient to calculate the correlation coefficient between the different source data and analyzing the correlation between the different source data.

[0021] By calculating the correlation coefficients between different source data, we can obtain the degree of correlation between the different source data, providing a basis for analysis.

[0022] In some optional embodiments, the step of using multi-order differential or fast Fourier transform (FFT) operations to calculate the dynamic changes of each source data, analyze the health status of the exercise user, and identify abnormal trends includes: using first-order differential on the position data to calculate the rate of change, and using second-order differential to calculate the acceleration of change; real-time monitoring of the rate of change and acceleration of change of the exercise user's multi-source data during training, and comparing the changing trends of different parameters; The third-order derivative is used to calculate the jerk, which represents the comfort and energy expenditure of the user during exercise. Fourth-order differential is used to calculate accelerometer, reflecting the smoothness of motion speed, the performance of the athlete, the professionalism of the athlete, and the physical wear and tear on the athlete. The fifth-order differential is used to calculate the acceleration, demonstrating ultra-precise muscle control and muscle data reconstruction; Fast Fourier Transform is used to establish the conversion between time-domain data and frequency-domain data, decompose motion into vibrations of different rhythms, analyze the rhythm of motion, convert displacement changes into vibration spectrum changes, and analyze the dynamic response of mechanical and control information to facilitate the diagnosis of anomalies and the identification of abnormal user jitter. The Laplace transform and Z-transform are used to map the physical domain to the discrete domain, distinguish the real part and conjugate, and evaluate the stability of the system. Based on the changing trends of multi-source data, determine whether sports users have unbalanced training loads, excessive fatigue, or health problems.

[0023] By employing multiple methods to process the data and analyzing its dynamic changes from different perspectives, we can provide a basis for subsequent analysis, expand the scope of analysis, and improve its accuracy.

[0024] Through intelligent data acquisition and analysis technologies, and by utilizing parameter correlation and variability analysis, the system can deeply explore the potential value of motion data. This addresses the problem of existing technologies where data analysis remains superficial, comprehensively improving data management and analysis capabilities.

[0025] By combining the different calculation results from multiple data sources with the assessment of the health and training status of sports users, the one-sidedness of single data analysis is avoided, and the comprehensiveness and completeness of the analysis results are improved.

[0026] In some optional embodiments, the data performance of the sports user in the game includes: scoring efficiency, defensive efficiency, and real-time tactical execution; the correlation between physical fitness and strategy execution includes: analyzing the sports user's physical fitness data and quantifying the impact of physical fitness on technical movements; the impact of psychological state on technical movements includes: adjusting psychology to maintain optimal competitive state based on the sports user's psychological state data; the impact of opponents and venue on the game result includes: analyzing the opponent's game performance to obtain optimized suggestions for defensive strategies, analyzing environmental data and referee penalty tendency data to assess the impact of external factors and provide support for tactical decision-making; the generation of intelligent decision-making includes: using a neural network model based on game data to perform parameter correlation analysis, dynamic change analysis, and historical data analysis, evaluating the sports user's physical fitness, strategy execution, and psychological state in the game, generating optimization suggestions, providing personalized health management suggestions, formulating personalized training plans and game strategies, and predicting potential sports injury risks based on the sports user's health data, competitive state, and historical data.

[0027] Based on competition data, the competition status of sports users is analyzed from different perspectives to obtain their health status and competitive status. Combined with historical data, personalized health management suggestions, personalized training plans and competition strategies are provided to sports users, thereby improving their competitive level.

[0028] By combining historical data with analysis, the analysis is extended longitudinally, further enhancing the comprehensiveness and completeness of the results. By introducing analytical methods based on correlation and historical changes, it not only helps identify health abnormalities in athletes but also provides optimization suggestions based on their training and competition status, avoiding the limitations of analyzing only single data sources.

[0029] In some optional embodiments, the neural network model includes at least one classical neural network model, with multiple classical neural network models working together to perform motion intelligence analysis and decision-making; the classical neural network model includes: a variational autoencoder for extracting motion latent variables and reconstructing nonlinear patterns; a physical information neural network for introducing physical constraints to improve the rationality and generalization ability of reconstruction; a long short-term memory network (LSTM) or a bidirectional LSTM for establishing temporal mapping relationships based on the training load data and training results of the sports users; a graph neural network (GNN) for establishing spatial interaction and tactical cooperation relationship models among multiple sports users; and deep reinforcement learning (DRL) for generating personalized training plans and real-time competition strategies.

[0030] By employing multiple classic neural network models, this study analyzes the correlation between different source data from multiple perspectives and examines the dynamic changes of each source data. This allows for more accurate analysis of various results for sports users, providing strategies for their subsequent training and competition, and improving their competitive level and health management.

[0031] Comprehensive analysis of multi-source competition data provides a complete assessment of athletes' condition, improving the accuracy of the evaluation. By collecting data in real time from multiple sensors and combining information on athletes' physical and mental states, as well as tactical execution, precise real-time decision support is provided to help coaches and athletes develop optimal competition strategies.

[0032] Intelligent analysis based on multi-source data leads to intelligent decision-making, which improves the accuracy of decision-making. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the embodiments of this application will be described in detail below. However, those skilled in the art will understand that many technical details are presented in the embodiments of this application to help readers better understand this application. However, even without these technical details and various changes and modifications based on the following embodiments, the technical solutions claimed in this application can be implemented. The division of the following embodiments is for the convenience of description and should not constitute any limitation on the specific implementation of this application. The various embodiments can be combined with and referenced by each other without contradiction.

[0034] To address the aforementioned technical problem of uncorrelated single data in sports analysis systems, this invention proposes an artificial intelligence-based intelligent sports big data analysis and management method. The following details the implementation of this embodiment of the artificial intelligence-based intelligent sports big data analysis and management method. These details are provided for ease of understanding and are not essential for implementing this solution. Example

[0035] This embodiment presents an artificial intelligence-based smart sports big data analysis and management method, which can be applied to electronic devices with communication, computing, and data storage capabilities, including: The data collection phase, data storage phase, data analysis phase, competition data analysis phase, intelligent analysis phase, and feedback and health management phase are all part of the data collection process.

[0036] During the data collection phase, data from different types of sensors and sources are collected, and the data is preprocessed to exclude outliers, standardize and unify the data, obtain single data from multiple sources, and then merge the single data from multiple sources to obtain fused data.

[0037] During the data storage stage, different data are stored in separate areas to facilitate data analysis.

[0038] The data analysis phase includes parameter correlation analysis, dynamic change analysis, and comparative analysis of historical and current data.

[0039] Specifically, parameter correlation analysis includes calculating the correlation coefficient between different sports parameters, discovering the correlation between different data, combining historical data to detect abnormal states, calculating the change ratio between historical data and real-time data, indicating the presence of anomalies when the change ratio exceeds the change threshold, and judging health status and competitive level.

[0040] Dynamic change analysis includes using multi-order differential or fast Fourier transform (FFT) operations to calculate the dynamic changes of each source data, calculate the rate of change and acceleration of change of the same source data, assess the health status of exercise users in real time, and identify abnormal change trends.

[0041] Data sets are formed by integrating data, related data, and dynamically changing data.

[0042] Historical data analysis includes comparing real-time data with historical data, analyzing relevant data on the physical fitness, training status and competitive level of athletes based on the comparison results, assessing the progress of athletes' physical fitness and competitive status, assessing whether athletes are at risk of overtraining or fatigue, and issuing warnings when abnormalities occur.

[0043] The competition data analysis phase includes analyzing competition performance, real-time tactical execution, physical fitness and strategy, psychological state and technical movements, and opponent and environment data based on the competition data.

[0044] In the intelligent analysis phase, a neural network model is established, trained, and optimized. Using the neural network model and competition data, the actual performance of athletes in the competition is analyzed, and key factors such as athletes' physical fitness, strategy execution, and psychological state are evaluated to generate intelligent strategies.

[0045] During the feedback and health management phase, personalized health management suggestions, competition and training recommendations are provided based on the sports user's health data, competitive status and historical data.

[0046] Data collection phase: The system collects real-time physiological and / or exercise data from multiple data acquisition sources of exercise users. It processes the collected multi-source exercise data to obtain individual data from each source, merges the individual data from multiple sources to form fused data, analyzes the correlation between different source data in the fused data, obtains the correlation coefficient between different source data, and uses multi-order differential or fast Fourier transform (FFT) operations to calculate the dynamic changes of each source data, analyze the health status of exercise users, and identify abnormal trends.

[0047] By comparing the current real-time data with historical data, the differences between the user's current real-time data and historical data under the current motion state are analyzed, and the difference results are obtained.

[0048] The health status of exercise users can be determined based on the correlation coefficient and the difference results.

[0049] Based on your health status, we will provide personalized health management and training recommendations.

[0050] We collect sports and competition data from users, process the data, and obtain standardized single datasets from multiple sources. We then store the data in different regions based on data type and perform data analysis on each source's single dataset.

[0051] A neural network model is established to analyze the competition status, opponent status, and environmental status of athletes based on standardized single data from multiple sources. This model generates intelligent decisions to formulate optimized training and competition strategies. Personalized health management, competition, and training recommendations are then provided based on the competition data.

[0052] Multiple data acquisition sources include cameras, sensors, wearable devices, and medical examination equipment, and each data acquisition source obtains a single data point.

[0053] The physiological and exercise data of sports users include: heart rate, respiratory volume, weight, body fat, pace, exercise speed, and exercise intensity.

[0054] Multi-source game data from sports users includes: real-time game data, venue data, biosignal data, scoring efficiency, defensive efficiency, and real-time tactical execution.

[0055] The multi-source competition data is categorized and stored separately.

[0056] Synchronize all data with time.

[0057] After processing the data, including cleaning, denoising, and removing outliers, the data is standardized and formatted to obtain standardized single data from each source.

[0058] Standardized single data from multiple sources are fused to obtain fused data. The fused data is then analyzed, including: calculating the correlation coefficients between different source data and analyzing the correlation between different sports data; using multi-order differential, fast Fourier transform (FFT), Laplace transform, and Z-transform to calculate the dynamic changes of each source data; forming a dataset from the fused data, correlated data, and dynamically changing data; comparing the current real-time data with historical data; and based on the comparison results, analyzing the relevant data of the sports user's physical fitness, training status, and competitive level to assess the sports user's physical fitness and competitive status. Based on the analysis results, health management suggestions and training suggestions are proposed.

[0059] The Pearson correlation coefficient method and the Spearman rank correlation coefficient method were used to calculate the correlation coefficient between different data, and the correlation between different data was found. Under a certain training intensity, the correlation of the changing trends of different data was analyzed.

[0060] First-order differential is used to calculate the rate of change; Second-order differential is used to calculate changing acceleration; Real-time monitoring of the rate and acceleration of change of multi-source data during training of sports users, and comparison of the changing trends of different parameters; The third-order differential is used to calculate jerk, which represents the comfort and energy expenditure of a user during exercise. For example, when riding a roller coaster, the magnitude of acceleration and jerk is related to comfort. The fourth-order differential is used to calculate acceleration, reflecting the smoothness of motion speed, the performance of the user, the level of professionalism of the user, and the physical wear and tear on the user. The fifth-order differential is used to calculate acceleration, demonstrating ultra-precise muscle control and muscle data reconstruction; Fast Fourier Transform (FFT) establishes the conversion between time-domain and frequency-domain data, decomposing motion into vibrations of different rhythms, analyzing the rhythm of motion, and converting displacement changes into changes in the vibration spectrum to analyze the dynamic response of mechanical and control information. This facilitates the diagnosis of problems such as insufficient resonance and the identification of abnormal user vibrations. The Laplace transform and Z-transform are used to map the physical domain to the discrete domain, distinguish the real part and conjugate, and evaluate the stability of the system to determine whether there is oscillation / overshoot / error / drift.

[0061] Based on the changing trends of multi-source motion data, it can be determined whether the exercise user has an unbalanced training load, excessive fatigue, or health problems.

[0062] Data sets are formed by integrating data, related data, and dynamically changing data.

[0063] By comparing real-time exercise data with historical exercise data, and determining that there is a significant difference between the current exercise user's exercise status and its historical data when the comparison error is greater than the comparison threshold, an abnormal situation of the exercise user can be identified.

[0064] Analyze real-time sports data to determine the health status and competitive level of sports users, and provide personalized health management suggestions.

[0065] Based on competition data, data analysis and neural network model analysis are conducted to obtain data on athletes' performance during the competition, the correlation between physical fitness and strategy execution, the impact of psychological state on technical movements, the influence of referees and the venue on the competition results, and intelligent decision-making. This results in the provision of health management suggestions, training suggestions, and competition response strategies. The analysis also identifies athletes' competitive state during the competition, analyzes the correlation between their physical fitness and strategy execution, and provides suggestions for energy allocation and strategy adjustment based on the analysis results. Furthermore, it analyzes the impact of athletes' psychological state on technical movements and provides psychological adjustment suggestions. Finally, it integrates publicly available data on opponents and competition venue data to analyze the influence of referee bias and venue environment on the competition results, providing decision support.

[0066] A neural network model is established, and motion latent variables and features are extracted. Nonlinear, non-commentary task-related motion patterns are reconstructed. A physical constraint loss function is applied, and training data is correlated with training results and competition data is correlated with competition results to generate intelligent decision-making strategies and provide personalized training plans and competition strategies.

[0067] In this application, one or more classic neural network models are selected. When multiple classic neural network models are used in combination, they can work together organically to form a hierarchical and complementary motion intelligence analysis and decision-making system.

[0068] In this application, several classic neural network models are used, including: variational autoencoder (VAE), physical information neural network (PINN), long short-term memory network (LSTM) or bidirectional LSTM, attention mechanism and graph neural network, and deep reinforcement learning (DRL).

[0069] Variational autoencoders (VAEs) and their temporal variants are core tools for extracting latent variables and reconstructing nonlinear patterns in motion. A VAE can compress high-dimensional motion data (such as joint angle sequences, ground reaction forces, and electromyographic signals) into low-dimensional probability distribution latent variables. These latent variables typically encode essential features of the movement, such as gait symmetry, the timing of force application, and joint coordination patterns. By sampling and decoding from the latent space, VAEs can reconstruct diverse and continuous motion trajectories, making them ideal for motion interpolation, anomaly detection, or generating personalized motion templates. To handle the inherent temporal dependencies of motion data, more commonly used variants are LSTM-VAE or GRU-VAE, where both the encoder and decoder employ recurrent neural network structures, enabling the capture of the dynamic evolution of periodic or quasi-periodic movements such as running, swimming, and weightlifting. Furthermore, conditional variational autoencoders (CVAEs) allow the addition of conditional information such as user identity, fatigue level, or skill level to the latent variables, making the extracted latent variables more individualized and facilitating the customization of analysis models for different users.

[0070] Physical Information Neural Networks (PINNs) represent a key breakthrough in enhancing the plausibility and generalization ability of reconstructed motion patterns by incorporating physical constraints. Purely data-driven models, when reconstructing motion patterns, may output statistically plausible but physically infeasible results, such as abnormal postures violating joint range of motion or momentum conservation. To address this issue, physical equation residual terms can be explicitly added to the loss function of the neural network, forming a Physical Information Neural Network (PINN). For motion analysis tasks, typical physical constraints include Newton's second law (the relationship between force and acceleration), conservation of angular momentum, the balance equation between ground reaction force and center-of-mass acceleration, and the feasibility boundary of joint torques. By simultaneously minimizing data fitting error and physical residuals, PINNs can reconstruct motion patterns that conform to fundamental biomechanical principles even with sparse or noisy training data. For example, from limited optical motion capture data, PINNs can infer complete joint torque sequences and ground force distributions, with results naturally satisfying dynamic consistency. This physically enhanced reconstruction capability provides a high-fidelity input foundation for subsequent task-related pattern recognition and policy generation.

[0071] In linking training / competition data with results and establishing predictive mappings, a combination of Long Short-Term Memory (LSTM) networks, attention mechanisms, and graph neural networks can be used. First, a temporal mapping relationship needs to be established between the training load data and training results of the athletes. Training load data includes running distance, heart rate zone duration, and strength training volume; training results include the next day's physical recovery score and injury occurrence. At this stage, LSTM networks or bidirectional LSTMs can effectively handle unequal time series and capture cumulative fatigue effects and delayed adaptation phenomena. Further introducing multi-head attention mechanisms allows the model to automatically focus on key time periods, such as the influence weight of training arrangements in the week before the competition on competition performance, thereby improving the interpretability of the prediction results. For team sports, it is also necessary to model the spatial interactions and tactical cooperation relationships between multiple athletes. In this case, graph neural networks (GNNs) become an ideal choice. Specifically, each athlete can be considered a node in a graph, with node features including its current position, speed, and heart rate, while the relative distances, passing relationships, or line-of-sight connections between athletes constitute the edges of the graph. Through graph convolution operations, GNNs can extract advanced tactical features such as formation stability, passing network density, and local numerical superiority, and use these features to predict possession rate, shot conversion rate, and even the final probability of victory or defeat. This multi-level modeling from the individual to the collective makes the correlation between data and results more comprehensive and accurate.

[0072] In generating personalized training plans and real-time match strategies, Deep Reinforcement Learning (DRL) is the ultimate decision-making core. Once the aforementioned modules—feature extraction, physical constraint reconstruction, and outcome prediction—are ready, they can be integrated into a deep reinforcement learning framework to train an intelligent agent capable of autonomous decision-making. The state space is defined by the athlete's current physical condition, technical indicators, mental fatigue level, and the opponent's tactical characteristics. The action space is defined by adjustments to training intensity (such as running distance and rest intervals in the next training session), nutritional supplementation plans, and tactical instruction selection (such as high pressing or defensive retreat in football). The reward function is a weighted average of multiple objectives, including improving win rate, reducing injury risk, and maintaining long-term competitive performance. Advanced algorithms such as Proximal Policy Optimization (PPO) or Flexible Action Evaluation (SAC) can be used for strategy optimization. Unlike traditional fixed plans, deep reinforcement learning-driven strategies can dynamically adjust recommended plans based on the athlete's real-time feedback (such as changes in heart rate variability and fluctuations in performance indicators), achieving true personalization. For example, the system might learn that for a football player prone to cramps in the second half of a match, the proportion of high-intensity interval training should be reduced three days before the game, and a specific electrolyte supplementation program should be increased. Additionally, when facing high temperatures on the field, it should be advised to reduce the frequency of forward sprints. These strategies often go beyond the scope of human coaches' experience and are highly data-driven and adaptable.

[0073] Ultimately, a complete sports intelligence analysis and decision-making system should integrate these models into a unified end-to-end framework. The front end uses LSTM-VAE or multimodal encoders to extract latent motion variables and physically plausible high-dimensional features from multi-source data such as wearable devices, optical capture, and video analysis. The mid-end uses attention mechanisms and GNNs to model these features in association with labels such as match results and injury records, outputting interpretable risk factors and win / loss predictions. The back end, centered on deep reinforcement learning, combines the current state of the sports user extracted in real-time by the front end with the future risks predicted by the mid-end to generate specific, executable, and time-evolving training and match strategies. During the training phase, this framework can utilize historical data for imitation learning or offline reinforcement learning, while during the deployment phase, it continuously fine-tunes the strategy through online interaction. Through this hierarchical design, the neural network not only completes the entire chain from raw signals to semantic features, from physical constraints to result prediction, and from data analysis to decision output, but also provides coaches, sports users, and medical teams with an interpretable, interventionable, and automatically optimized intelligent assistance platform.

[0074] The competition strategy is used to respond to situations during the competition, providing real-time decision-making suggestions and adjusting according to the athlete's competition status. For real-time competition strategy generation and dynamic decision-making suggestions, a closed-loop feedback system needs to be constructed, centered on deep reinforcement learning and integrating real-time state encoding and online inference. Unlike the macro-level training plan formulated before the competition, the strategy response during the competition requires the model to output executable tactical instructions or adjustment suggestions within milliseconds to seconds, based on the athlete's current physiological data (such as heart rate variability, blood oxygen saturation, electromyography fatigue index), athletic performance data (such as instantaneous speed, acceleration, cumulative running distance, shooting / hitting accuracy), and opponent / environment information (such as score, remaining time, venue temperature, and changes in opponent formation). To achieve this goal, the aforementioned latent variable extraction model (such as LSTM-VAE) needs to be deployed as a lightweight, real-time encoder at the edge, continuously compressing the raw sensor data stream into low-dimensional state vectors. Meanwhile, pre-trained deep reinforcement learning policy networks (such as PPO or SAC) can complete forward inference and output action decisions on mobile devices or wearable computing units with millisecond latency through techniques such as model quantization, knowledge distillation, or TensorRT acceleration.

[0075] To ensure the adaptability and robustness of the real-time strategy, the system needs to incorporate an online adaptive mechanism and a multi-gait risk prediction module. During the competition, the athlete's state is non-stationary (e.g., sudden fatigue, minor injury, psychological fluctuations), and pre-trained reinforcement learning strategies may not be able to cover all dynamic changes. Therefore, a lightweight online adaptation module can be added during the inference phase, for example, using a residual reinforcement learning framework, allowing the action output of the baseline strategy to be corrected by a small perturbation network. This perturbation network can perform rapid gradient updates based on the state-reward history of the past tens of seconds (e.g., using online meta-learning or model predictive control). In addition, the system should also integrate a short-term trajectory prediction module to predict changes in the athlete's key indicators (e.g., sprint ability decay curve, injury risk probability) within the next 5-10 seconds based on the current state and strategy actions. When a certain indicator is predicted to exceed a safety threshold, the policy network will automatically output forced adjustment instructions, such as reducing high-intensity confrontation, shifting to the sidelines to reduce sprints, or suggesting immediate hydration and cooling down, thereby minimizing the risk of injury while maintaining competitiveness.

[0076] In practical application architecture, real-time decision recommendations should be presented in a multimodal interactive manner and support the allocation of decision-making power in human-machine collaboration. The system can be deployed on coach tablets, smart wristbands for sports users, or AR glasses terminals, conveying strategy recommendations through various means such as vibration, voice, and visual cues. For example, when the system detects that a basketball player's shooting angle deviation exceeds a threshold three times in a row, it can prompt through the headset, "Prioritize driving and passing after a screen later, avoid jump shots" or "Request a timeout to adjust hand shape." Simultaneously, the system should not replace the coach's final judgment, but is designed as a three-in-one auxiliary tool of "recommendation-confidence-interpretability": each strategy recommendation is accompanied by a confidence score and key evidence (e.g., "Right ankle load abnormally increased by 23%, recommendation to reduce crossover drives, confidence level 87%)," which the coach can choose to adopt or ignore based on their experience. The system records these interactive feedbacks for subsequent offline retraining. Through this mechanism, the real-time game strategy system fully leverages the rapid response advantages of data-driven approaches while retaining the flexibility and trust foundation of human-machine collaboration.

[0077] Example 2: This embodiment is a detailed description of Embodiment 1.

[0078] Data collection phase: Simultaneously, data is collected from multiple sources to ensure data diversity and comprehensiveness. This includes using smart devices, sensors, and cameras to collect real-time exercise data from users, such as heart rate, weight, respiratory volume, competition records, strides, speed, movement type, technique execution, and opponent data. Biosensors, such as electroencephalograms (EEGs) and heart rate variability sensors, are used to monitor psychological states.

[0079] Collection of physiological movement data: Heart rate (HR): .

[0080] HR represents heart rate in bpm, and T represents heart rate cycle in seconds, which is the time interval between two consecutive heartbeats.

[0081] Respiratory volume (VR): is the respiratory rate per minute.

[0082] .

[0083] Where n represents the number of breaths per unit time, T represents the time period measured, and VR represents the respiratory volume, in bpm.

[0084] Weight and Body Fat Percentage: .

[0085] Fat Mass represents body fat mass, and Total Body Mass represents total body mass.

[0086] Data collection during exercise: Pace and speed of movement, including stride frequency and speed.

[0087] speed : .

[0088] Where S represents the distance traveled and t represents the time.

[0089] Exercise intensity or exercise load: estimated by the ratio of maximum heart rate to current heart rate.

[0090] .

[0091] in, Indicates the current heart rate. This represents the maximum heart rate, which is usually estimated as follows: .

[0092] Data collection during the match includes: real-time match data and field data.

[0093] The system uses a match recording system, sensors worn by athletes, cameras, and video analytics to collect data during the match, including statistics and tactical execution data. The statistics include points, assists, steals, etc.

[0094] Field data, including data that directly affects the performance of athletes, is obtained by using environmental sensors to monitor the condition of the competition field in real time, such as slipperiness, air temperature, and humidity.

[0095] The wetness index of the playing surface is calculated using temperature and humidity sensors. .

[0096] The slipperiness index is a function of the combined effects of ambient humidity and temperature.

[0097] Psychological state data, including brain waves and heart rate variability.

[0098] Using electroencephalogram (EEG) and heart rate variability (HRV) monitoring instruments, the psychological state of exercise users can be detected in real time. For example, heart rate variability can reflect psychological health data such as stress response and relaxation state of exercise users.

[0099] .

[0100] in, This represents the RR interval of each heartbeat. This represents the average RR interval, and n represents the number of RR intervals.

[0101] Data synchronization and transmission: All data collected by the sensors is synchronized to the central processing system or cloud platform via wireless communication protocols, ensuring rapid processing and storage of real-time data.

[0102] During data synchronization, clock synchronization technology is used to ensure that the timestamps of data from different devices and sensors are consistent, so as to enable correct correlation analysis.

[0103] The collected data undergoes preprocessing, including data cleaning, noise reduction, data structuring and fusion, to ensure the accuracy and reliability of the data, exclude outlier data, and perform data correction to guarantee the accuracy of subsequent analysis.

[0104] Data structuring and fusion converts raw data from different sources into a unified format, ensuring that the data can be fused and analyzed on the same platform.

[0105] By adopting data standardization and unified processing methods, differences between devices and platforms are eliminated, ensuring data consistency.

[0106] The data preprocessing stage is a crucial step in ensuring the accuracy and reliability of the system's analysis results. It comprises two main parts: data cleaning and denoising, and data structuring and fusion. These steps guarantee the quality of the system's input data, thereby improving the accuracy of subsequent analysis and decision support. The following is a detailed description of this stage, including relevant formulas and explanations of principles.

[0107] Data processing stage: This includes: data cleaning and denoising, data correction, and data structuring and fusion.

[0108] In actual data acquisition, data often contains noise and outliers due to environmental interference, equipment errors, sensor malfunctions, and other factors. Data cleaning and denoising aim to remove invalid data and correct erroneous data from the raw data, ensuring the reliability and consistency of the dataset. Specific methods include the following: Data cleaning involves removing outliers from a collected dataset. Outliers are data points that significantly deviate from normal values. These outliers often originate from sensor malfunctions, environmental interference, or data transmission problems. Certain criteria are needed to identify and remove these outliers.

[0109] Anomaly detection methods include the following: Method 1: Statistical approach: Data is analyzed using the mean and standard deviation. If a data point deviates from the mean by more than three times the standard deviation, it is considered outlier. Standardized values... The calculation method is as follows: .

[0110] in, Let represent the i-th data point, and represent the mean. Indicates standard deviation, if ,but This is an abnormal value.

[0111] Method 2, based on the quartile method: This method uses the quartiles (Q1, Q3) and interquartile range (IQR) of the data to detect outliers. If data points... Below or higher If it is, then it is an abnormal value.

[0112] .

[0113] Noise reduction methods address the issue of noise, which refers to meaningless random variations in data, potentially caused by measurement errors, environmental interference, etc. Commonly used noise reduction methods include the following: Method 1, Moving Average Method: This method smooths the data and eliminates noise. Commonly used moving average methods are Simple Moving Average (SMA) and Weighted Moving Average (WMA).

[0114] .

[0115] This represents the smoothed value at time t, where n represents the window size. This represents the original data points.

[0116] Method 2, median filtering: Replace each data point in the dataset with the median of its neighborhood, which can effectively remove isolated noise points.

[0117] .

[0118] in, Indicates the size of the neighborhood window.

[0119] Data calibration refers to the process of correcting systematic errors caused by equipment errors, offsets, or measurement errors during data cleaning. For example, the deviation of a temperature sensor can be corrected using known calibration standards.

[0120] .

[0121] in, This indicates uncorrected data. These are correction coefficients obtained through experiments. This indicates the corrected data.

[0122] Data structuring and fusion: During the data acquisition process, data from different devices and sensors often have inconsistent formats. The purpose of data structuring and fusion is to unify and standardize these data from different sources and eliminate differences between devices and platforms to ensure that the data can be analyzed uniformly on the same platform.

[0123] Data format standardization: Raw data from various sources, such as sensor data, video data, and text data, are converted into a unified structured format, typically in tabular or matrix form. This process mainly involves the following steps: Time synchronization: Since the data comes from different devices, there may be timestamp discrepancies, so it is necessary to synchronize the time of the data from different sensors.

[0124] .

[0125] in, Indicates the timestamp after synchronization. Represents the original timestamp. This indicates an adjustment deviation.

[0126] Data type conversion: Convert data from different data sources into a unified type format, such as integers, floating-point numbers, booleans, etc., and perform unit conversion as needed.

[0127] Data standardization is the process of converting data from different dimensions to the same dimension to facilitate comparison and analysis. Common standardization methods include the following: Standardization Method 1: Z-Score Standardization .

[0128] in, X Represents the original data. This represents the mean of the data. Standard deviation represents standardized data. It will have zero mean and unit variance.

[0129] Standardization Method Two: Min-Max Standardization .

[0130] in, This represents the maximum value of the data. This represents the minimum value of the data. The standardized data range is [0,1].

[0131] Data fusion is the process of combining data from different sources to form a comprehensive dataset. The goal of data fusion is to improve the integrity and accuracy of the data.

[0132] The merging methods include at least one of the weighted average method and the multimodal data fusion method.

[0133] Weighted average method: For data from different sources, different weights are assigned according to their credibility, and a weighted average is calculated.

[0134] .

[0135] in, This represents the weight of the i-th data source. This represents the value of the i-th data source. The merged data.

[0136] Multimodal data fusion methods require the use of appropriate algorithms to fuse data from different types of sensors, such as image data, audio data, and sensor data. These algorithms include Kalman filtering and particle filtering.

[0137] Data storage stage: Based on the data type, different data are stored in separate regions.

[0138] Within the first storage area, competition-related metrics are collected, such as match statistics, tactical execution data, and real-time on-field data. These are used to identify the competitive state and performance of athletes during the match based on real-time monitoring of the data, and to provide a reference for subsequent decision-making.

[0139] Game statistics include points, assists, rebounds, and field goal percentage; tactical execution data includes tactical execution success rate; and real-time on-court data includes the speed, position, and actions of the players.

[0140] In the second storage area, physical fitness and strategy execution data of sports users are collected to analyze the correlation between physical fitness and strategy execution. For example, the impact of physical fitness on technical actions such as shooting accuracy and sprint success rate is analyzed. Based on the analysis results, suggestions for physical fitness allocation and strategy adjustment are provided.

[0141] In the third storage area, data on the psychological state of athletes monitored by biosensors are collected to analyze the impact of psychological state on their technical movements, such as serving accuracy, and to provide psychological adjustment suggestions to help athletes maintain optimal mental state.

[0142] In the fourth storage area, publicly available data on opponents and data on the playing field, such as slipperiness and temperature, are collected to analyze the influence of refereeing biases and field conditions on the match results, providing decision support.

[0143] This phase involves managing and analyzing various data types through multiple dedicated storage areas, including competition-related metrics, athlete physical and strategic analysis, psychological state and technical movement analysis, and integrated analysis of opponent and environmental data. The goal is to monitor and analyze athlete performance in real time, helping them optimize their competition strategies and improve their competitive level.

[0144] Data analysis phase: This includes parameter correlation analysis, dynamic change analysis, and historical data analysis.

[0145] Parameter correlation analysis analyzes the correlation between different exercise parameters, such as heart rate and weight, respiratory volume and age, to identify key factors affecting the health status and competitive performance of exercise users, calculate the correlation coefficient between parameters, and reveal the potential correlation between parameters.

[0146] By combining historical data analysis, the system identifies the differences between the current motion state and historical data, and detects abnormal situations.

[0147] The correlation coefficient model is used to determine the current health status and competitive level of sports users.

[0148] Methods for calculating correlation coefficients include the Pearson correlation coefficient method and the Spearman rank correlation coefficient method.

[0149] Pearson Correlation Coefficient: The Pearson correlation coefficient is an indicator that measures the strength of the linear relationship between two variables. The value ranges from -1 to 1. The closer the value is to 1, the more positive the correlation between the two variables is; the closer the value is to -1, the more negative the correlation is; and a value of 0 indicates no correlation.

[0150] .

[0151] in, , These represent the i-th data points of the two variables, , Each of them represents its own mean. This represents the Pearson correlation coefficient.

[0152] Spearman Rank Correlation: Applicable to the analysis of nonlinear relationships, it measures the monotonic relationship between two variables.

[0153] .

[0154] in, This represents the difference in rank between corresponding values ​​of two variables. Indicates the number of samples. This represents the Spearman rank correlation coefficient.

[0155] By calculating correlation coefficients, we can discover the correlations between parameters such as heart rate and weight, and respiratory volume and age. For example, the relationship between heart rate and weight can be calculated using the Spearman rank correlation coefficient, reflecting whether changes in weight and heart rate exhibit the expected linear trend under a certain training intensity.

[0156] Historical data comparison and anomaly detection: By combining historical data analysis, we can determine whether there are significant differences between the current exercise user's exercise status and their historical data, thereby identifying abnormal situations. For example, if an exercise user's current heart rate rises significantly faster than historical data under the same training intensity, it may indicate that the exercise user is overloaded or has a health problem.

[0157] The formula for comparing historical data is shown below: .

[0158] In the formula, This indicates the ratio of the current heart rate to the historical heart rate. Indicates the current heart rate. This indicates the historical heart rate.

[0159] if If the data exceeds a preset threshold, the current data is marked as abnormal, indicating that the user may be overtraining, have health problems, or be fatigued.

[0160] Assessing health status and athletic performance: Based on the correlation coefficient calculation method and by comparing and analyzing current data with historical data, the current health status and competitive level of an athlete can be determined. Assuming the correlation coefficient between heart rate and weight is m, if the change in the correlation between current heart rate and weight exceeds a preset threshold, the system infers that the athlete may have health problems, such as rapid weight change leading to increased cardiac burden.

[0161] Dynamic change analysis: The system assesses the health status of exercise users in real time based on their rate of change and acceleration during training (such as rate of change in heart rate and rate of change in weight).

[0162] Identify abnormal trends, such as mismatches between heart rate and weight, and automatically generate warnings and health management suggestions.

[0163] Dynamic change analysis primarily assesses an athlete's health status by monitoring the rate and acceleration of changes in these parameters during training, such as the rate of change in heart rate and weight. This analysis helps to quickly identify health abnormalities or mismatched training data and issue timely warnings.

[0164] Rate of Change: The first derivative of a parameter, representing the magnitude of change of that parameter per unit time, is commonly used to measure the fluctuations in the health status of exercise users.

[0165] .

[0166] in, This represents the parameter value at the current moment. This represents the parameter value at the previous moment. Indicates time difference.

[0167] Rate of change: The second derivative of a parameter, representing the rate of change of that parameter per unit time.

[0168] .

[0169] In the formula, Indicates the rate of change at the current moment. It represents the rate of change at the previous moment.

[0170] By analyzing the rate and acceleration of change in parameters such as heart rate and weight of exercise users, their health status can be assessed in real time. For example, if an exercise user's rate of change is too fast while their weight change is small, it indicates that they are overtraining, and the system will automatically issue a warning.

[0171] Third-order differentials are used to calculate jerk, which represents the comfort and energy expenditure of the user during exercise. For example, when riding a roller coaster, the magnitude of acceleration and jerk is related to comfort.

[0172] Fourth-order differentials are used to calculate acceleration, reflecting the smoothness of motion speed, the superior performance of the user, the professionalism of the user, and the wear and tear on the user's body.

[0173] The fifth-order differential is used to calculate the acceleration, demonstrating ultra-precise muscle control and muscle data reconstruction.

[0174] By employing Fast Fourier Transform, a conversion between time-domain and frequency-domain data is established, decomposing motion into vibrations of different rhythms. The rhythm of the motion is analyzed, and the changes in displacement are converted into changes in the vibration spectrum. The dynamic response of the mechanical and control information is analyzed, which facilitates the diagnosis of problems such as insufficient resonance and identifies abnormal shaking of users.

[0175] The Laplace transform and Z-transform are used to map the physical domain to the discrete domain, distinguish the real part and conjugate, and evaluate the stability of the system to determine whether there is oscillation / overshoot / error / drift.

[0176] Based on the changing trends of multi-source data, determine whether sports users have unbalanced training loads, excessive fatigue, or health problems.

[0177] Anomaly trend identification: If the trends in heart rate and body weight do not match, it may indicate an unbalanced training load for the exercise user. For example, if body weight remains constant while heart rate continues to rise, it could be due to over-fatigue or other health problems. The system identifies these abnormal trends in real time through dynamic change analysis.

[0178] Historical data analysis: Historical data includes raw data collected from multiple data sources preceding the current data, standardized single data after data processing, fused data, correlated data, and dynamically changing data.

[0179] Historical data analysis assesses an athlete's progress in physical fitness and competitive condition by comparing their past training data with current data. It helps identify early warning signs of overtraining or fatigue.

[0180] By reviewing historical data of sports users and combining it with their current training status, we can assess their progress in physical fitness and competitive performance.

[0181] Assess the risk of overtraining or fatigue among sports users and provide early warnings.

[0182] Assess progress in physical fitness and competitive condition: By comparing historical data and current training data of athletes, such as heart rate, weight, and athletic performance, the progress trend of their physical fitness and competitive level can be analyzed. For example, if the current training intensity is the same as in the past, but the athlete's heart rate changes more slowly, it may mean that their physical fitness is progressing well.

[0183] The formula for physical fitness progress is as follows: .

[0184] In the formula, This represents the baseline heart rate of the exercise user, based on historical data. This indicates the current heart rate.

[0185] Warning signs of overtraining or fatigue: By evaluating training load and recovery status in historical data, we can assess whether there is a risk of overtraining. For example, if a user's fitness level does not improve significantly and heart rate changes more drastically, we can determine that the user may be entering a state of fatigue and issue an early warning.

[0186] The fatigue index is calculated using the following formula: .

[0187] If the fatigue index exceeds the set threshold, an overtraining warning will be automatically issued.

[0188] In the correlation analysis and data modeling phase, an in-depth analysis of the relationships between various physiological and exercise parameters of sports users is conducted to uncover their potential correlations, thereby constructing a model for judging health status and competitive level. This phase mainly includes three parts: parameter correlation analysis, dynamic change analysis, and historical data analysis. Through these analyses, the system can monitor the health status of sports users in real time and generate timely warnings and suggestions.

[0189] Match data analysis phase: This includes analyzing match performance based on match data, real-time tactical execution, physical and strategic performance, psychological state and technical movements, and opponent and environmental data.

[0190] Based on the collected game-related metrics, including statistics, tactical execution data, and real-time on-court data—including points, assists, rebounds, and field goal percentage; tactical execution data such as tactical execution success rate; and real-time on-court data such as the speed, position, and movements of the players—this data helps analyze the actual performance of the players during the game and provides a reference for subsequent decision-making.

[0191] Real-time monitoring of competition data allows for continuous tracking of each athlete's performance. For a specific athlete's real-time data, their performance during the competition is analyzed using the following formula: Scoring efficiency analysis: .

[0192] in, This indicates the score of the user in this sport. This indicates the total number of shots taken by the user in this sport. It reflects the scoring efficiency of sports users.

[0193] Defensive efficiency analysis: .

[0194] in, Indicates defensive rebound. Indicates inference, It indicates the time spent on the field and reflects the defensive efficiency of the sports players.

[0195] Real-time tactical execution analysis: Tactical execution data involves the success and failure of athletes in executing specific tactics during a match. By comparing this data, the effectiveness of tactical execution can be evaluated. For example, the formula for calculating the success rate of a certain tactic is as follows: .

[0196] in, Indicates the success rate of tactics. Indicates the number of successful executions. This indicates the total number of times the command was executed.

[0197] The above analysis can help teams and coaches adjust tactics and strategies in a timely manner to improve performance in the game.

[0198] Physical and strategic analysis, including analyzing the impact of physical condition on technical movements.

[0199] Athletes' physical fitness levels directly impact their performance in games. By analyzing athletes' physical fitness data, such as heart rate and fatigue index, the influence of physical fitness on technical movements can be quantified. For example, a decline in physical fitness may lead to reduced shooting accuracy and a lower sprint success rate.

[0200] The impact of physical fitness on shooting accuracy is shown in the following formula: .

[0201] in, The fatigue index indicates the level of physical exertion experienced by athletes. Indicates shooting percentage. Indicates the number of successful shots.

[0202] Through physical fitness and strategy analysis, dynamic suggestions for energy allocation and strategy adjustments can be provided during the competition. For example, if an athlete is in a low physical condition, it is recommended to reduce high-intensity sprints and instead focus on tactical execution.

[0203] Analysis of psychological state and technical movements: By monitoring the psychological state of athletes using biosensors (such as electroencephalograms and heart rate variability), and analyzing its impact on technical movements (such as serving accuracy and receiving reaction time), it becomes clear that psychological state has a significant influence on the accuracy of technical movements. Therefore, real-time analysis of psychological state can help athletes maintain their optimal competitive state.

[0204] The psychological state of athletes can be assessed using physiological parameters such as electroencephalograms (EEGs) and heart rate variability (HRV). For example, excessive anxiety or stress can lead to larger fluctuations in heart rate, longer reaction times, and affect the accuracy of technical movements.

[0205] Heart rate variability analysis is shown in the following formula: .

[0206] in, The heart rate variability (HRV) value represents the time interval between two consecutive heartbeats. It reflects the psychological stress and relaxation state of an exercise user; a lower HRV usually indicates higher psychological stress. At this time, the user is in a relaxed state.

[0207] The influence of mental state on serving accuracy is shown in the following formula: .

[0208] In the formula, It indicates the psychological state of the user. Indicates accuracy or precision. Indicates a success story. This represents the overall case study.

[0209] Based on real-time psychological state analysis, it can provide psychological adjustment suggestions, such as relaxation training, meditation, or deep breathing, to help athletes adjust their mental state and maintain their best competitive state.

[0210] Competitor and environmental data analysis: By integrating publicly available data about the opponent, such as match videos and statistics, as well as data about the playing field, such as slipperiness and temperature, and analyzing the influence of refereeing biases and field conditions on the match outcome, this data helps to comprehensively assess the external factors of the match and provides support for tactical decision-making.

[0211] Competitor data analysis: Publicly available data such as opponent statistics and game videos can provide targeted tactical guidance for the team. By analyzing the opponent's game performance, such as shooting habits and defensive strategies, it is possible to suggest how to adjust tactics to counter the opponent.

[0212] An analysis of an opponent's shooting habits is shown in the following formula: .

[0213] In the formula, Indicates the opponent's shooting percentage. Indicates the number of successful shots. This indicates the total number of attempts.

[0214] By analyzing the opponent's shooting percentage and shooting position, suggestions for optimizing defensive strategies are provided.

[0215] Environmental and referee analysis: Environmental factors at the competition venue and the referee's bias in officiating can also affect the outcome of the competition. For example, a slippery venue may cause athletes to slip and fall, and excessively high temperatures may affect athletes' physical performance. Collecting and analyzing environmental data can help assess the impact of external factors and provide support for tactical decisions.

[0216] The impact of venue environment on game performance As shown in the following formula: .

[0217] In the formula, Indicates humidity. Indicates temperature. This indicates the slipperiness of the playing surface. The impact of the playing surface environment on performance is related to factors such as temperature and slipperiness.

[0218] The correlation analysis between environmental data and sports user performance data is used to predict the impact of external factors on the competition.

[0219] Environmental data includes slipperiness and temperature, while user performance data includes the number of slips and falls and fatigue levels.

[0220] Referee penalty bias analysis involves analyzing historical referee penalty data, such as the number of fouls and technical fouls, to identify referee penalty biases and provide guidance for strategic adjustments during the game.

[0221] Intelligent analysis phase: Establishing neural network models for sports competition analysis can effectively capture the complex relationship between input parameters and output performance, and generate intelligent strategies.

[0222] A neural network model consists of an input layer, hidden layers, and an output layer.

[0223] The input parameters of the input layer are These represent parameters such as the physical fitness and mental state of the sports user, respectively, and the output layer represents the sports user's competitive performance. These represent the athletic performance of the user, such as shooting accuracy and sprint success rate. Their working principle is as follows: The input layer is used to receive data such as the user's physical fitness, strategy execution, and psychological state.

[0224] Hidden layers are used to perform weighted summation based on the weight matrix and input parameters, and to perform nonlinear transformations through activation functions to simulate complex input-output relationships. The activation functions include ReLU and Sigmoid functions.

[0225] The output layer is used to predict the competitive performance of sports users.

[0226] The basic mathematical expression for a neural network is shown below: .

[0227] in, For input parameters, This represents the weights from input to output. Indicates a bias value. This represents the activation function. This indicates the output parameters.

[0228] In this application, the function is used This indicates that the specific function in different formulas is different.

[0229] The training process of a neural network model includes the following steps: S1. Forward propagation: The input parameters are calculated through each layer of the neural network to obtain the prediction result.

[0230] S2. Error Calculation: Calculate the error of the neural network model based on the difference between the output result and the actual performance.

[0231] If the mean squared error loss function is used to calculate the difference, the calculation formula is as follows: .

[0232] in, This represents the output predicted by the neural network. Indicates actual competitive performance. Indicates the number of samples.

[0233] S3. Use the backpropagation algorithm to adjust the weights and biases in the neural network to minimize the error function.

[0234] In one specific embodiment of this application, a neural network model is used for analysis and prediction, as follows: The input parameters of the neural network model include physical condition, strategy execution, and psychological state. The predicted output is the shooting accuracy of the athlete. Based on neural network model analysis, the following correlation formula is given: .

[0235] in, This represents the output of a neural network model, i.e., the complex relationship between shooting accuracy and input parameters. Indicates physical fitness. Indicates psychological state. This indicates that the strategy has been executed.

[0236] Neural networks can generate optimization suggestions by learning from historical data. For example, when a person is in a bad mental state, they can predict that shooting accuracy will decline and suggest psychological adjustment measures, such as relaxation training and meditation.

[0237] During the real-time decision-making phase, based on the performance and current status of athletes in the competition, immediate feedback is provided to help coaches and athletes adjust tactics. Through the analysis of real-time data streams, the competitive status of athletes is judged, and tactical suggestions are provided according to changes.

[0238] Real-time monitoring of game data to assess key factors such as physical fitness, strategy execution, and mental state of athletes during the game, assuming a shooting percentage in the current sport. As input, the following factors are analyzed in real time: The effect of physical fitness is shown in the following formula: .

[0239] If a user's performance declines due to decreased physical fitness, the neural network model will suggest adjusting energy allocation and reducing high-intensity tasks for the user.

[0240] The influence of psychological state is shown in the following formula: .

[0241] If the psychological state is unstable, suggestions for psychological adjustment will be given.

[0242] The impact of strategy execution is shown in the following formula: .

[0243] If the current tactics are not being executed properly, suggestions for tactical adjustments will be provided.

[0244] Specifically, when a user experiences a decline in physical fitness, the coach is reminded to adjust the allocation of physical activity and reduce the user's high-intensity tasks.

[0245] If the analysis indicates that the user's psychological state is unstable, relaxation exercises such as meditation and deep breathing will be recommended. If the current tactics are not performing well, adjustment suggestions will be provided, such as changing the tactical arrangement or choosing a strategy more suitable for the current state.

[0246] Based on the analysis results of the neural network model and real-time data, real-time feedback is generated, and suggestions are promptly delivered to coaches and sports users.

[0247] If the analysis finds that the exercise user's heart rate is too high, it will be recommended to reduce the intensity of the exercise or take a rest.

[0248] Its decision support The formula is as follows: .

[0249] In the formula, Indicates heart rate, Indicates shooting percentage. It indicates a psychological state.

[0250] Feedback and Health Management Phase: Based on users' health data, competitive status, and historical data, personalized health management recommendations are provided. Users are reminded to pay attention to adjustments in areas such as physical recovery, rest, and nutritional intake to reduce the risk of sports injuries.

[0251] Based on the analysis results of the neural network model, personalized training plans and competition strategies are developed for athletes to improve their competitive performance. Optimization solutions are provided for aspects such as energy allocation, strategy adjustment, and psychological adjustment to ensure athletes achieve their best performance in competitions.

[0252] In the "Feedback and Health Management Phase," the system comprehensively analyzes the user's health data, competitive status, and historical data to provide personalized health management and training recommendations. This helps users maintain good physical and mental health, maximize competitive performance, and reduce the risk of sports injuries. The core objective of this phase is to provide scientific and personalized health management plans and training strategies to ensure users maintain peak performance during training and competition.

[0253] Health management recommendations are based on a comprehensive analysis of the athlete's physiological state, competitive performance, and historical data. The aim is to help athletes recover their physical fitness and improve their health through reasonable adjustments, thereby reducing the risk of sports injuries and promoting the maintenance of competitive performance over a long period of time.

[0254] The system collects health indicators from various physiological and psychological data sources, including heart rate, weight, sleep quality, and exercise load. Through data analysis, the system identifies fatigue accumulation, overtraining, and potential health risks during training.

[0255] Heart rate monitoring: Heart rate is an important indicator for measuring the physical condition and fatigue level of athletes. Based on heart rate changes, the following health management suggestions are provided: .

[0256] In the formula, Indicates resting heart rate. Indicates heart rate after exercise. This indicates that the heart rate has recovered.

[0257] If a user's heart rate recovers slowly during exercise, it is recommended to increase rest time and avoid overtraining.

[0258] The speed of physical recovery reflects the balance between training intensity and rest for an athlete. Assuming the athlete's training load is... Physical recovery coefficient is The recovery effect is evaluated as shown in the following formula: .

[0259] In the formula, K represents the ideal training load, and K is a constant.

[0260] If the recovery coefficient is lower than the recovery threshold, it is recommended to appropriately reduce the training intensity.

[0261] Sports injury warning: Based on the analysis of sports users' health and training data, potential sports injury risks are predicted.

[0262] Suppose the neural network model identifies the correlation between heart rate changes and exercise load. If a user's heart rate changes too quickly during high-intensity training, it indicates that the body may be overloaded and the risk of injury may be increased. In this case, a warning will be issued to remind the user to adjust the training intensity.

[0263] Probability of damage As shown in the following formula: .

[0264] In the formula, These are the weights associated with various health indicators. This represents each health data point. This represents the sigmoid function.

[0265] Personalized health management recommendations Based on health data and athletic performance, the system can provide personalized health management recommendations. For example, if an athlete has poor sleep quality, the system may suggest increasing sleep time or adjusting sleep quality, such as reducing strenuous activity after training or increasing relaxation techniques like meditation.

[0266] Training and match recommendations: Based on the athlete's historical data, current training status, and competition needs, personalized training and competition suggestions are provided, aiming to optimize energy allocation, tactical adjustments, and psychological conditioning, thereby improving the athlete's overall competitive performance.

[0267] Recommendations for energy distribution: Energy allocation is a key factor in improving the competitive performance of athletes, especially in high-intensity training and competition. Based on the athlete's energy data and competitive needs, an optimized energy allocation plan can be generated in real time.

[0268] Assuming the training time is Physical energy consumption is Based on the recovery status of the athletes, the following allocation suggestions are given: .

[0269] in, These represent adjustment coefficients, indicating the intensity of training and the strength of recovery, respectively. Indicates the optimal duration. It indicates the ability to recover.

[0270] Strategy adjustment suggestions: A sports user's ability to execute strategies is closely related to their physical fitness, mental state, and technical movements. By analyzing real-time data, suggestions for adjusting strategies during the competition can be provided.

[0271] If a player's physical condition declines during a game, leading to a decrease in shooting accuracy, it is recommended that the coach adjust tactics, such as increasing rest time or changing the offensive approach.

[0272] The strategy efficiency in the game is With physical fitness and tactical execution The relationship is as follows: .

[0273] If the strategy efficiency is lower than the strategy efficiency threshold, suggestions are given to adjust the training plan or tactics to improve the strategy execution efficiency.

[0274] Psychological adjustment suggestions: The psychological state of athletes has a significant impact on their competitive performance. By monitoring the psychological state of athletes, such as brain waves and heart rate variability, through biosensors, we can provide corresponding adjustment suggestions.

[0275] The psychological state of sports users is measured by the emotional index. The effect of psychological adjustment It can be calculated using the following formula: .

[0276] In the formula, A coefficient representing the effect of psychological adjustment.

[0277] like If the mood index is below the threshold, relaxation techniques such as meditation and deep breathing are recommended to help exercisers restore their optimal mental state.

[0278] Through the feedback and health management phase, personalized health management and training suggestions are provided to sports users, covering aspects such as physical recovery, training load, strategy adjustment, and psychological adjustment. By utilizing health data, competitive status, and historical data analysis, the system designs optimal training and competition plans for sports users, helps them reduce sports injuries, improve competitive performance, and ensures that they maintain their best condition during long-term training and competition. This effectively improves the competitive level of sports users and enhances their performance in competitions.

[0279] Based on the user's personal data, such as age, weight, and training history, personalized health management recommendations are provided to help users maintain optimal physical condition and improve their training and health status.

[0280] By employing precise energy allocation, strategy execution analysis, and psychological state optimization, we help sports users perform at their best in competitions, thereby improving their win rate and competitive level.

[0281] Through correlation analysis, the system can monitor information such as the physical load and training intensity of sports users in real time, promptly identify potential overtraining or sports injury risks, and help sports users avoid unnecessary injuries.

[0282] It can achieve intelligent fusion of multi-source data, simplify data management processes, improve the accuracy and efficiency of data analysis, and thus enhance the level of data analysis in the sports industry.

[0283] By integrating opponent data, match statistics, and venue data, the system can provide targeted match strategy support, helping coaching teams make more scientific and precise tactical decisions. Example 3: Another embodiment of this application relates to an artificial intelligence-based smart sports big data analysis and management system. The implementation details of this embodiment of an artificial intelligence-based smart sports big data analysis and management system are described in detail below. The following content is only for the convenience of understanding and is not necessary for implementing this solution. This embodiment of an artificial intelligence-based smart sports big data analysis and management system includes a data acquisition module, a data storage module, a data analysis module, a neural network module, and a management suggestion module.

[0284] The data acquisition module includes various types of sensors, video data acquisition units, and text data acquisition units. It cleans, denoises, corrects, structures, and fuses the acquired raw data to obtain a comprehensive dataset. The data storage module is used to store different types of data in separate regions; The data analysis module is used to perform parameter correlation analysis, dynamic change analysis, and historical data analysis on data to obtain the correlation between different data. By combining historical data, it can identify abnormal states and assess health status and competitive level. It can also perform match performance analysis, real-time tactical execution analysis, physical fitness and strategy analysis, psychological state and technical action analysis, and opponent and environment data analysis on match data to assess physical fitness, competitive level, and match status. The neural network module is used to correlate input parameters with output competitive performance, analyze the psychological state and technical movements of athletes, the performance of opponents and environmental factors, and generate intelligent decisions. The management advice module is used to provide personalized health management advice, competition and training recommendations.

[0285] It is worth mentioning that all modules involved in this embodiment are logical modules. In practical applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. Furthermore, to highlight the innovative aspects of this application, this embodiment does not introduce units that are not closely related to solving the technical problems proposed in this application; however, this does not mean that other units are absent in this embodiment.

[0286] Example 4: Another embodiment of this application relates to an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform an artificial intelligence-based smart sports big data analysis and management method according to the above embodiments.

[0287] The memory and processor are connected via a bus, which can include any number of interconnecting buses and bridges, connecting various circuits of one or more processors and memories. The bus can also connect various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and will not be described further herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by the processor is transmitted over the wireless medium via an antenna, which further receives data and transmits it to the processor.

[0288] The processor manages the bus and general processing, and also provides various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory is used to store data used by the processor during operation.

[0289] Example 5: Another embodiment of this application relates to a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the method embodiments described above.

[0290] That is, those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0291] Those skilled in the art will understand that the above embodiments are specific embodiments for implementing this application, and in practical applications, various changes can be made to them in form and detail without departing from the spirit and scope of this application.

Claims

1. A smart sports big data analysis and management method based on artificial intelligence, characterized in that, include: The system collects motion data from multiple data sources on the physiological and / or physical aspects of exercise users in real time. It processes the collected single motion data from each source to obtain standardized single data from each source. The standardized single data from multiple sources are then fused to form fused data. Based on the different source data in the fused data, the correlation coefficient between the different source data is calculated, and the correlation between the source data is analyzed. By employing multi-order differential or fast Fourier transform operations, the dynamic changes of each source data are calculated to analyze the health status of exercise users and identify abnormal trends. The dataset is formed by integrating data, correlated data, and dynamically changing data. The historical data and current data of sports users are compared. Based on the comparison results, the relevant data of sports users' physical fitness, training status, and competitive level are analyzed to assess the sports users' physical fitness and competitive status. Based on the analysis results, health management suggestions and training suggestions are proposed.

2. The intelligent sports big data analysis and management method based on artificial intelligence according to claim 1, characterized in that, Also includes: Collect competition data from multiple data acquisition sources for sports users, process the data to obtain individual data from each source; store sports data and competition data by type in different areas, perform data analysis and / or neural network model analysis on the individual data from each source, obtain data performance of sports users in the competition, the correlation between physical fitness and strategy execution, the influence of psychological state on technical movements, and the influence of opponents and venue on the competition results, generate intelligent decisions, and propose health management suggestions, training suggestions, and competition response strategies; The system is divided into four zones: the first zone stores competition-related indicators to identify the competitive state and performance of athletes during the competition; the second zone stores athletes' physical fitness and strategy execution data to provide suggestions for energy allocation and strategy adjustment; the third zone stores athletes' psychological state data to analyze the impact of athletes' psychological state on technical movements and provide suggestions for psychological adjustment; and the fourth zone stores publicly available data on opponents and competition venue data to analyze the impact of referee bias and venue environment on the competition results.

3. A smart sports big data analysis and management method based on artificial intelligence according to claim 1 or 2, characterized in that, The multiple data acquisition sources, including cameras, sensors, wearable devices, and physical examination equipment, are used to collect physiological and exercise data and competition data of sports users, including: heart rate, respiratory volume, weight, body fat, stride, exercise speed, exercise intensity, real-time competition data, venue data, and biosignal data.

4. A smart sports big data analysis and management method based on artificial intelligence according to claim 1 or 2, characterized in that, The data processing includes: data cleaning, denoising, correction, structuring, and fusion. Data cleaning and denoising are used to remove abnormal data from the dataset. Correction is used to correct systematic errors during data cleaning. Structuring is used to unify and standardize the format of multi-source data. Fusion is used to merge the structured multi-source data according to a fusion mode to form fused data. The fusion mode includes a regular mode and a cross mode.

5. The intelligent sports big data analysis and management method based on artificial intelligence according to claim 1, characterized in that, The analysis of the correlation between motion data from different sources includes: using the Pearson correlation coefficient or the Spearman correlation coefficient to calculate the correlation coefficient between the data from different sources, and analyzing the correlation between the data from different sources.

6. The intelligent sports big data analysis and management method based on artificial intelligence according to claim 1, characterized in that, The method of using multi-order differential or fast Fourier transform (FFT) operations to calculate the dynamic changes of each source data, analyze the health status of the exercise user, and identify abnormal trends includes: using first-order differential on the position data to calculate the rate of change, and using second-order differential to calculate the acceleration of change; real-time monitoring of the rate of change and acceleration of change of the exercise user's multi-source data during training, and comparing the changing trends of different parameters. The third-order derivative is used to calculate the jerk, which represents the comfort and energy expenditure of the user during exercise. Fourth-order differential is used to calculate accelerometer, reflecting the smoothness of motion speed, the performance of the athlete, the professionalism of the athlete, and the physical wear and tear on the athlete. The fifth-order differential is used to calculate the acceleration, demonstrating ultra-precise muscle control and muscle data reconstruction; Fast Fourier Transform is used to establish the conversion between time-domain data and frequency-domain data, decompose motion into vibrations of different rhythms, analyze the rhythm of motion, convert displacement changes into vibration spectrum changes, and analyze the dynamic response of mechanical and control information to facilitate the diagnosis of anomalies and the identification of abnormal user jitter. The Laplace transform and Z-transform are used to map the physical domain to the discrete domain, distinguish the real part and conjugate, and evaluate the stability of the system. Based on the changing trends of multi-source data, determine whether sports users have unbalanced training loads, excessive fatigue, or health problems.

7. The intelligent sports big data analysis and management method based on artificial intelligence according to claim 2, characterized in that, The data performance of the athletes in the game includes: scoring efficiency, defensive efficiency, and real-time tactical execution; the correlation between physical fitness and strategy execution includes: analyzing the athletes' physical fitness data and quantifying the impact of physical fitness on technical movements; the impact of psychological state on technical movements includes: adjusting psychology to maintain optimal competitive state based on the athletes' psychological state data; the impact of opponents and venue on the game result includes: analyzing the opponents' game performance to obtain optimized suggestions for defensive strategies, analyzing environmental data and referee penalty tendency data to assess the impact of external factors and provide support for tactical decision-making; the generation of intelligent decisions includes: using neural network models based on game data to conduct correlation analysis, dynamic change analysis, and historical data analysis of different source data, evaluating the athletes' physical fitness, strategy execution, and psychological state in the game, generating optimization suggestions, providing personalized health management suggestions, formulating personalized training plans and game strategies based on the athletes' health data, competitive state, and historical data, and predicting potential sports injury risks.

8. The intelligent sports big data analysis and management method based on artificial intelligence according to claim 7, characterized in that, The neural network model includes at least one classical neural network model, and multiple classical neural network models work together to perform motion intelligent analysis and decision-making. Classic neural network models include: variational autoencoders, used for motion latent variable extraction and nonlinear pattern reconstruction; Physical information neural networks are used to introduce physical constraints to improve reconstruction rationality and generalization ability; Long Short-Term Memory (LSTM) networks or bidirectional LSTM networks are used to establish a temporal mapping relationship based on the training load data and training results of sports users. Graph Neural Networks (GNNs) are used to build models of spatial interactions and tactical coordination among multiple sports users. Deep reinforcement learning (DRL) is used to generate personalized training plans and real-time competition strategies.

9. A smart sports big data analysis and management system based on artificial intelligence, characterized in that, include: The system comprises a data acquisition module, a data storage module, a data analysis module, a neural network module, and a management suggestion module. The data acquisition module includes various types of sensors, a video data acquisition unit, and an environmental data acquisition unit. It cleans, denoises, corrects, structures, and fuses the acquired raw data to obtain fused data. The data storage module stores different data in separate regions. The data analysis module performs parameter correlation analysis, dynamic change analysis, and historical data analysis on the data to obtain correlations between different data. Combined with historical data, it identifies abnormal states and assesses health status and competitive level. It also performs match performance analysis, real-time tactical execution analysis, physical fitness and strategy analysis, psychological state and technical movement analysis, and opponent and environmental data analysis on the match data to assess physical fitness, competitive level, and match status. The neural network module correlates input parameters with output competitive performance, analyzes the athlete's psychological state and technical movements, opponent match performance, and environmental factors, and generates intelligent decisions. The management suggestion module provides personalized health management suggestions, match and training recommendations.

10. An electronic device, characterized in that, include: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform an artificial intelligence-based smart sports big data analysis and management method as described in any one of claims 1 to 8.