AI body measurement-based motion data management method and system

Abstract

The invention discloses a motion data management method and system based on AI body measurement, and relates to the technical field of artificial intelligence and machine learning, and the method comprises the steps: collecting motion data of each user, and carrying out the preprocessing; dividing user intervals by using a K-means clustering algorithm, constructing a load prediction model by using a fuzzy BP neural network, and obtaining a final load prediction value by using an improved adaptive genetic algorithm; and constructing a regression model by using a QHD reinforcement learning algorithm, dynamically adjusting a discount factor, outputting a training load adjustment strategy, and adjusting the discount factor according to feedback. User training load intervals are accurately divided through a K-means clustering algorithm, distribution of personalized training loads is ensured, the load prediction precision is improved by using a fuzzy BP neural network and an IAGABP algorithm, the weight and bias of the network are optimized, a local optimal solution is avoided, the prediction accuracy is improved, a QHD reinforcement learning algorithm and a dynamic discount factor are combined, and the prediction accuracy is improved. And a strategy of dynamically adjusting the training load is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence and machine learning, and in particular to a method and system for managing sports data based on AI physical testing. Background Art

[0002] With the continuous advancement of science and technology, sports data management methods based on artificial intelligence have been widely used in the fields of sports science and health management. Especially driven by smart wearable devices, sports data collection technology has continued to mature. Various sensors and smart devices can monitor the user's exercise status in real time, such as heart rate, exercise intensity, cadence, exercise duration and other data. These data can provide users with scientific training feedback and help formulate personalized training plans. At the same time, with the rise of big data analysis and machine learning technology, the use of machine learning algorithms to conduct in-depth analysis of sports data and achieve more accurate load prediction and dynamic adjustment has become an important development direction for sports training management.

[0003] However, although the existing technology has made certain progress in motion data analysis and training load prediction, there are still some significant shortcomings. The existing motion data management methods mainly rely on rule-based empirical models or single algorithms, lacking personalization and dynamic adaptability. Traditional load prediction methods are usually based on static data models or empirical formulas. These methods are difficult to fully consider the individual differences of each user and the influence of environmental factors, resulting in a lack of flexibility in the formulation of training plans and difficulty in meeting the personalized needs of different users. Many existing methods often only rely on fixed training parameters or simple adjustment strategies when adjusting the load, and it is difficult to achieve dynamic adjustment based on real-time feedback and physiological status of users. How to combine advanced machine learning algorithms with motion data analysis to propose a personalized training plan that can accurately predict user load and flexibly adjust is a major challenge in current technology. Summary of the invention

[0004] The purpose of the present invention is to provide a motion data management method and system based on AI physical measurement to solve the above problems.

[0005] The present invention is achieved through the following technical solutions: In a first aspect, the present invention provides a method for managing sports data based on AI physical measurement, comprising: Collect and pre-process the motion data of each user; The K-means clustering algorithm is used to divide the user intervals, the fuzzy BP neural network is used to build the load forecasting model, and the improved adaptive genetic algorithm is used to obtain the final load forecast value; Use the QHD reinforcement learning algorithm to build a regression model and dynamically adjust the discount factor to output a training load adjustment strategy, and adjust the discount factor based on feedback; All data is stored and managed in a relational database.

[0006] As a preferred solution of the motion data management method based on AI physical measurement of the present invention, the method of collecting the motion data of each user for preprocessing and forming a feature vector includes: Collect the motion data of each user at each time point during the training cycle through the collection device; The motion data includes physiological data and environmental data; Preprocessing the collected motion data of each user; The preprocessing includes using a moving average method to denoise the motion data, using a linear interpolation method to fill in missing data, using a standard deviation method to detect and remove outliers, and using a Z-Score standardization method to standardize all motion data, and mapping the value of each motion data to between [0,1].

[0007] As a preferred solution of the sports data management method based on AI physical measurement of the present invention, wherein: the use of K-means clustering algorithm to divide user intervals includes: According to the motion data of all users, the number of clusters is set to 3, and the motion data of 3 users are randomly selected as the initial cluster centers. The Euclidean distance between each user's motion data and all initial cluster centers is calculated, and the user is assigned to the cluster center with the closest distance. The mean of the motion data of all users in each cluster is recalculated as the new cluster center. The above steps are repeated until the cluster centers converge. Each cluster corresponds to a training load interval. The training load intervals include a primary interval, an intermediate interval and an advanced interval; After clustering is completed, the silhouette coefficient is used to measure the similarity of each user to its cluster and to other clusters.

[0008] As a preferred solution of the motion data management method based on AI physical test described in the present invention, wherein: the use of fuzzy BP neural network to construct a load prediction model and the use of improved adaptive genetic algorithm to obtain the final load prediction value includes: Each data in each user's motion data is used as a feature, and the membership value of each feature is calculated using a Gaussian membership function; Combine all the membership values ​​of all the features of each user at each time point into a membership matrix, where each row represents a feature and each column represents the membership value of the feature in different fuzzy sets. The membership matrix of each user at each time point is taken as a sample. Select fuzzy BP neural network as load forecasting model and build the structure of load forecasting model; The weights and biases are initialized by random initialization method, and the membership matrix is ​​used as input to train the load forecasting model. According to the error of the output layer, the weights and biases in the network are optimized by back propagation algorithm until the error converges; The weights and biases of all nodes in the load forecasting model are organized into chromosomes, and each weight and bias is regarded as a gene; Set the size of the initial population, use uniform distribution to generate each chromosome in the initial population, the gene value of each chromosome is a random real number, and select the mean square error as the fitness function to calculate the initial fitness value of each chromosome. The formula is:

[0009]

[0010] in, Represents chromosome The initial fitness value of Represents chromosome The output error value is represents the number of samples, represents the load target value of the i-th sample, represents the load forecast value of the i-th sample; The load target value refers to the load prediction value based on the history of all users, and different load target values ​​are assigned to the three training load intervals through expert analysis; Construct an improved fitness function and calculate the final fitness value of each chromosome. The formula is:

[0011]

[0012] in, Represents chromosome The final fitness value of Represents the improved chromosome Output error value, represents the weighting coefficient of the i-th sample, represents the time point of the i-th sample, represents the maximum time point in the sample, represents the standard deviation of the Gaussian function, represents the dynamic discount factor at time point t, represents the load forecast value of the jth sample; The dynamic discount factor refers to a function that uses the Sigmoid function as the growth function of the dynamic discount factor, and the expression is:

[0013] in, represents the maximum value of the dynamic discount factor, represents the control coefficient, represents the central time point of the dynamic discount factor; According to the combination of weight and bias of each chromosome, the chromosome is evaluated by the improved fitness function, and the weight and bias are continuously optimized through the crossover and mutation operations of the genetic algorithm to minimize the prediction error; Calculate the probability of each chromosome being selected ; After calculating the probability based on the fitness value, the roulette wheel method is used to simulate the random selection of chromosomes for the next generation; By performing crossover and mutation operations on the selected chromosomes, the chromosomes in the population are optimized generation by generation. After each generation of selection, crossover and mutation operations, the prediction performance of each chromosome is evaluated using the fitness function, and the most optimized chromosome is selected for the next generation; Set an error threshold. When the value of the fitness function is less than the error threshold, stop the algorithm, set the current time point as the central time point of the dynamic discount factor and output the final load forecast value, indicating that the expected forecast accuracy has been achieved. Otherwise, repeat the iteration.

[0014] As a preferred solution of the sports data management method based on AI physical test described in the present invention, the training load adjustment strategy of using QHD reinforcement learning algorithm to build a regression model and dynamically adjust the discount factor output includes: Define the baseline discount factor ; The QHD reinforcement learning algorithm is used to build a regression model. The purpose of the regression model is to maximize the future cumulative return and obtain the final load forecast value and dynamic discount factor. and the baseline discount factor Substitute the corresponding membership matrix and use the membership matrix as the state space of the regression model And define the action space , the membership matrix is ​​mapped to the high-dimensional feature space through the QHD algorithm; The action space represents the actions that can be selected by the regression model at a certain point in time, including training intensity, training duration and training frequency. The action space is expressed as: ,in, Represented as the training intensity selected at time point t , training duration and training frequency Actions; The training intensity includes low intensity, medium intensity and high intensity; The training duration includes short, medium and long time; The training frequency includes low frequency, medium frequency and high frequency; Set the advantage function, the formula is:

[0015] in, is the advantage function, which represents the advantage value of the action selected based on the current strategy at time point t. is the discounted return, expressed in terms of the dynamic discount factor Next, the future cumulative return of the training strategy from the time point t after the action is executed to the end of the training cycle is, is the state value function, which means that in the state Next, according to the current strategy The expected return after The discount return The formula is:

[0016] in, is an immediate reward, indicating that at a certain point in time The rewards you get when It indicates the number of future time points starting from time point t; The instant reward refers to setting a heart rate target range for the user, collecting the user's heart rate in real time and comparing and determining the instant reward. If the user's heart rate is within the heart rate target range, it means that the training load is appropriate and the instant reward is a positive reward. Otherwise, the instant reward is a negative reward. The state value function The formula is:

[0017] in, is the expected value, indicating that in the state Next, according to the current strategy The expected return after execution; During the training process, the regression model calculates the advantage function, evaluates the current training load adjustment, and uses the regression model to optimize the training load, adjust the training intensity, training duration and training frequency to maximize the return of the current strategy; After each round of training, the discount factor is adjusted in real time according to the value of the advantage function: like , then the dynamic discount factor The value is high and needs to be reduced , reduce the impact of future cumulative returns on current decisions; like , then the dynamic discount factor Effective, need to increase , strengthen the impact of future cumulative returns on current decisions; like , then the dynamic discount factor At the optimal point, stop adjusting; If the dynamic discount factor At the optimum, the regression model outputs the optimal action as the training load adjustment strategy.

[0018] As a preferred solution of the motion data management method based on AI physical measurement of the present invention, wherein: the adjustment of the discount factor according to feedback includes: According to the obtained training load adjustment strategy, the user's training plan is adjusted. During the training load adjustment process, the discount factor is dynamically adjusted according to the real-time feedback data; If the user's recovery is good, increase the discount factor based on the feedback; If the user's recovery is poor, the discount factor is reduced based on the feedback.

[0019] As a preferred solution of the sports data management method based on AI physical test of the present invention, wherein: the storing and managing of all data in a relational database includes: Select a relational database to store and manage data and its analysis results, design the database table structure to store different types of data, set up regular backup tasks, back up all data in the database, manage the permissions of database users and encrypt and store data.

[0020] In a second aspect, the present invention provides a sports data management system based on AI physical measurement, comprising: Data collection module, used to collect and pre-process the motion data of each user; The load forecasting module is used to divide the user intervals using the K-means clustering algorithm, build the load forecasting model using the fuzzy BP neural network, and obtain the final load forecast value using the improved adaptive genetic algorithm; The strategy adjustment module is used to build a regression model using the QHD reinforcement learning algorithm and dynamically adjust the discount factor output training load adjustment strategy to adjust the discount factor based on feedback; The data storage module is used to store and manage all data in a relational database.

[0021] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the motion data management method based on AI body measurement as described in the first aspect of the present invention is implemented.

[0022] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the motion data management method based on AI body measurement as described in the first aspect of the present invention is implemented.

[0023] Compared with the prior art, the present invention has the following beneficial effects: accurately dividing the user training load interval through the K-means clustering algorithm to ensure the distribution of personalized training load, using the fuzzy BP neural network and IAGABP algorithm to improve the load prediction accuracy, optimizing the network weights and biases to avoid local optimal solutions, and improving the prediction accuracy, combining the QHD reinforcement learning algorithm and the dynamic discount factor to realize the strategy of dynamically adjusting the training load. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings: Figure 1 This is a flow chart of the motion data management method based on AI physical measurement in Example 1; Figure 2 This is a schematic diagram of the sports data management system based on AI physical measurement in Example 1. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical scheme and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments and drawings. The schematic implementation modes and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention. It should be noted that the present invention is already in the actual development and use stage. Example 1

[0026] like Figure 1 and Figure 2 As shown, this embodiment includes: S1, collect the motion data of each user and pre-process it; Specifically, the motion data of each user is collected for preprocessing and forming a feature vector, including: Collect the motion data of each user at each time point during the training cycle through the collection device; The motion data includes physiological data and environmental data; The physiological data includes the user's heart rate, cadence, exercise duration, lactate level, body temperature, etc.; The environmental data includes the temperature, humidity, air quality and light of the training environment; Preprocessing the collected motion data of each user; The preprocessing includes using a moving average method to denoise the motion data, using a linear interpolation method to fill in missing data, using a standard deviation method to detect and remove outliers, and using a Z-Score standardization method to standardize all motion data, and mapping the value of each motion data to between [0,1].

[0027] By preprocessing each user's motion data, including denoising, filling missing data, removing outliers and standardization, the quality and accuracy of the data can be effectively improved. Using the moving average method to denoise can reduce random fluctuations in motion data, linear interpolation to fill missing values ​​ensures data continuity, and standard deviation to remove outliers avoids the impact of extreme data on analysis results. All motion data are mapped to [0,1] through Z-Score standardization to ensure that different data dimensions are compared and analyzed on the same scale, thereby providing more stable and reliable input data for subsequent AI algorithm models. This series of preprocessing steps can effectively improve the reliability of the data and the prediction accuracy of the model.

[0028] S2, using K-means clustering algorithm to divide user intervals, using fuzzy BP neural network to build load forecasting model and using improved adaptive genetic algorithm to obtain the final load forecast value; Specifically, the K-means clustering algorithm is used to divide user intervals, including: According to the motion data of all users, the number of clusters is set to 3, and the motion data of 3 users are randomly selected as the initial cluster centers. The Euclidean distance between each user's motion data and all initial cluster centers is calculated, and the user is assigned to the cluster center with the closest distance. The mean of the motion data of all users in each cluster is recalculated as the new cluster center. The above steps are repeated until the cluster centers converge. Each cluster corresponds to a training load interval. The training load intervals include a primary interval, an intermediate interval and an advanced interval; After clustering is completed, the silhouette coefficient is used to measure the similarity of each user to the cluster in which he belongs and to other clusters. The value range of the silhouette coefficient is between [-1, 1]. The closer the value is to 1, the better the clustering effect is.

[0029] By using the K-means clustering algorithm to divide user motion data, the user's motion load can be divided into primary intervals, intermediate intervals and advanced intervals, thereby realizing personalized training load management. This method first calculates the Euclidean distance between each user and the initial cluster center, dynamically adjusts the cluster center to ensure that each user is assigned to the most appropriate load interval. After clustering is completed, the clustering effect is evaluated by the silhouette coefficient to ensure that each user's motion data has high similarity within the cluster and is highly different from other clusters, thereby improving the accuracy and effectiveness of training load division, providing a reliable basis for personalized motion data management, and optimizing the user's motion training effect.

[0030] Furthermore, the load forecasting model is constructed using fuzzy BP neural network and the final load forecasting value is obtained by using improved adaptive genetic algorithm, including: Each data in each user's motion data is used as a feature, and the membership value of each feature is calculated using a Gaussian membership function; The use of the Gaussian membership function to calculate the membership value of each feature includes setting a fuzzy set of "low", "medium" and "high" for each feature to represent different training load states, each fuzzy set has a corresponding Gaussian membership function, and the membership value of each feature in each fuzzy set is calculated by using the Gaussian membership function; Combine all the membership values ​​of all the features of each user at each time point into a membership matrix, where each row represents a feature and each column represents the membership value of the feature in different fuzzy sets. The membership matrix of each user at each time point is taken as a sample. Select fuzzy BP neural network as load forecasting model and build the structure of load forecasting model; The structure of constructing the load forecasting model includes adopting a single hidden layer neural network, including an input layer, a hidden layer and an output layer, the input of the input layer is the membership matrix of each user at all time points in the training cycle, the membership value of each fuzzy set corresponds to an input node, each node of the hidden layer is connected to all input nodes of the input layer, the hidden layer nodes receive the linear weighted combination signals of the input nodes and add bias values, and complete nonlinear mapping through the Sigmoid activation function, and the output layer summarizes the signals of the hidden layer nodes to generate the load forecast value of each user in the training cycle; The weights are used to adjust the impact of the input on the hidden layer activation; The bias is used to control the offset of the activation function, thereby improving the effect of nonlinear mapping; The weights and biases are initialized by random initialization method, and the membership matrix is ​​used as input to train the load forecasting model. According to the error of the output layer, the weights and biases in the network are optimized by back propagation algorithm until the error converges; The weights and biases of all nodes in the load forecasting model are combined into chromosomes. Each weight and bias is regarded as a gene. Assuming there are Q weights and W biases in total, each chromosome contains Q+W genes. Set the size of the initial population, use uniform distribution to generate each chromosome in the initial population, the gene value of each chromosome is a random real number, for example, the range is [-0.1, 0.1], determined according to the actual application, select the mean square error as the fitness function to calculate the initial fitness value of each chromosome, the formula is:

[0031]

[0032] in, Represents chromosome The initial fitness value of Represents chromosome The output error value is represents the number of samples, represents the load target value of the i-th sample, represents the load forecast value of the i-th sample; The load target value refers to the load prediction value based on the history of all users, and different load target values ​​are assigned to the three training load intervals through expert analysis; There may be some outliers or abnormal values ​​in the training data, which will have an adverse effect on model training. In order to reduce the impact of outliers on the model, we introduce a weighted Gaussian function to weight the error of the sample, especially for samples far away from the maximum timestamp (that is, the most recent training data). The expression is: ,in, represents the time point of the i-th sample, represents the maximum time point in the sample, Represents the standard deviation of the Gaussian function, which is used to control the range of weighting. A value of means that samples farther from the maximum time point will be given lower weights; Through the weighted Gaussian function, the model can reduce the impact of samples far away from the current time point on the training process, thereby reducing the interference of outliers; In the prediction of training load, as training progresses, the balance between short-term rewards (immediate feedback) and long-term rewards (future prediction) is crucial. The introduction of a dynamic discount factor can effectively adjust the model's focus on short-term and long-term rewards. The expression is: ,in, represents a dynamic discount factor, which is used to adjust the focus on long-term returns. represents the load forecast value of the jth sample; By introducing a dynamic discount factor, we can increase the focus on short-term returns through a lower discount factor in the early stage of training. In the later stage of training, as the model is optimized, the value of the discount factor is gradually increased, so that the model pays more attention to long-term returns, ensuring that the model can adapt to the needs of different training stages. Construct an improved fitness function and calculate the final fitness value of each chromosome. The formula is:

[0033]

[0034] in, Represents chromosome The final fitness value of Represents the improved chromosome Output error value, represents the weighting coefficient of the i-th sample, represents the time point of the i-th sample, represents the maximum time point in the sample, represents the standard deviation of the Gaussian function, represents the dynamic discount factor at time point t, represents the load forecast value of the jth sample; The dynamic discount factor refers to a function that uses the Sigmoid function as the growth function of the dynamic discount factor, and the expression is:

[0035] in, represents the maximum value of the dynamic discount factor, Represents the control coefficient, which is used to control the rate of increase. Indicates the central time point of the dynamic discount factor, which is pre-set according to actual needs; According to the combination of weight and bias of each chromosome, the chromosome is evaluated through the improved fitness function. The smaller the fitness value of the chromosome, the more accurate the prediction of the load forecasting model. Through the crossover and mutation operations of the genetic algorithm, the weight and bias are continuously optimized to minimize the prediction error and gradually improve the prediction accuracy of the training load. Calculate the probability of each chromosome being selected , the probability is inversely proportional to the fitness value, that is, the smaller the fitness value, the greater the probability of the chromosome being selected. The formula is:

[0036] in, represents the maximum fitness value in the initial population, represents the number of chromosomes in the initial population, Represents chromosome The final fitness value of After calculating the probability based on the fitness value, the roulette method is used to simulate the random selection of chromosomes for the next generation. The proportion of each chromosome in the total fitness determines its probability of being selected. The individuals with smaller fitness values ​​will have a greater probability of being selected. By performing crossover and mutation operations on the selected chromosomes, the chromosomes in the population are optimized generation by generation. After each generation of selection, crossover and mutation operations, the prediction performance of each chromosome is evaluated using the fitness function, and the most optimized chromosome is selected for the next generation; Set an error threshold. When the value of the fitness function is less than the error threshold, stop the algorithm, set the current time point as the central time point of the dynamic discount factor and output the final load forecast value, indicating that the expected forecast accuracy has been achieved. Otherwise, repeat the iteration.

[0037] This invention innovatively constructs an efficient load forecasting model by combining fuzzy BP neural network with improved adaptive genetic algorithm. It uses Gaussian membership function to fuzzy user motion data, takes the characteristic membership value of each user under different training load states as input, constructs a neural network structure with a single hidden layer, and optimizes weights and biases through back propagation algorithm. The improved adaptive genetic algorithm significantly improves the robustness of the model to outliers in the training data by introducing weighted Gaussian function and dynamic discount factor, and can effectively balance the concerns between short-term and long-term returns, thereby improving the prediction accuracy. The crossover and mutation operations in the genetic algorithm further optimize the network structure, minimize the prediction error, and ensure the high accuracy and adaptability of load forecasting. This method can provide high-precision prediction results in complex load forecasting tasks and adapt to dynamically changing load demands, and has strong application potential.

[0038] S3, using the QHD reinforcement learning algorithm to build a regression model and dynamically adjust the discount factor to output a training load adjustment strategy, and adjust the discount factor based on feedback; Specifically, the QHD reinforcement learning algorithm is used to build a regression model and dynamically adjust the discount factor output. The training load adjustment strategy includes: Define the baseline discount factor ; The QHD reinforcement learning algorithm is used to build a regression model. The purpose of the regression model is to maximize the future cumulative return and obtain the final load forecast value and dynamic discount factor. and the baseline discount factor Substitute the corresponding membership matrix and use the membership matrix as the state space of the regression model And define the action space , the membership matrix is ​​mapped to the high-dimensional feature space through the QHD algorithm; The action space represents the actions that can be selected by the regression model at a certain point in time, including training intensity, training duration and training frequency. The action space is expressed as: ,in, Represented as the training intensity selected at time point t , training duration and training frequency Actions; The training intensity includes low intensity, medium intensity and high intensity; The training duration includes short, medium and long time; The training frequency includes low frequency, medium frequency and high frequency; Set the advantage function, the formula is:

[0039] in, is the advantage function, which represents the advantage value of the action selected based on the current strategy at time point t. is the discounted return, expressed in terms of the dynamic discount factor Next, the future cumulative return of the training strategy from the time point t after the action is executed to the end of the training cycle is, Measures the impact of future rewards on current decisions in an action sequence. is the state value function, which means that in the state Next, according to the current strategy The expected return obtained after is approximated using a neural network, with the neural network input being the state , the output is the value of the state; The discount return The formula is:

[0040] in, is an immediate reward, indicating that at a certain point in time The rewards you get when It indicates the number of future time points starting from time point t, that is, Represents the reward at the current time point, Indicates the reward at the next time point; The instant reward refers to setting a heart rate target range for the user according to actual needs, and determining the instant reward by collecting the user's heart rate in real time and comparing it. If the user's heart rate is within the heart rate target range, it means that the training load is appropriate and the instant reward is a positive reward. Otherwise, the instant reward is a negative reward. The values ​​of the positive and negative rewards are determined according to actual conditions. The state value function The formula is:

[0041] in, is the expected value, indicating that in the state Next, according to the current strategy The expected reward after execution, which is estimated by training a neural network; During the training process, the regression model calculates the advantage function, evaluates the current training load adjustment, and uses the regression model to optimize the training load, adjust the training intensity, training duration and training frequency to maximize the return of the current strategy; After each round of training, the discount factor is adjusted in real time according to the value of the advantage function: like , then the dynamic discount factor The value is high and needs to be reduced , reduce the impact of future cumulative returns on current decisions; like , then the dynamic discount factor Effective, need to increase , strengthen the impact of future cumulative returns on current decisions; like , then the dynamic discount factor At the optimal point, stop adjusting; If the dynamic discount factor At the optimum, the regression model outputs the optimal action as the training load adjustment strategy.

[0042] By using the QHD reinforcement learning algorithm to build a regression model and dynamically adjust the discount factor, this method can intelligently optimize the training load adjustment strategy to ensure the maximization of the training effect. The baseline discount factor is set to 0.99, and its upper limit is limited to avoid excessive focus on future returns. During the training process, the regression model evaluates the pros and cons of the current strategy and the baseline strategy by calculating the advantage function, and dynamically adjusts the training intensity, duration and frequency according to the evaluation results to achieve the maximum future cumulative returns. By adjusting the dynamic discount factor in real time, the model can flexibly adjust the focus on future returns to ensure the effectiveness of the current decision and the optimality of the strategy. The regression model outputs the optimal training load adjustment strategy based on the adjusted discount factor to further improve the training effect and efficiency. This method is efficient and adaptive, and can continuously optimize the strategy based on the feedback information during the training process to improve the overall training quality and goal achievement.

[0043] Furthermore, the discount factor is adjusted based on the feedback including, According to the obtained training load adjustment strategy, the user's training plan is adjusted. During the training load adjustment process, the discount factor is dynamically adjusted according to the real-time feedback data so that the current training load can fully consider future returns and long-term goals.

[0044] By dynamically adjusting the discount factor, we can effectively balance short-term training load and long-term health goals. If the user's recovery is good, increasing the discount factor will make the training strategy focus more on future rewards, thereby optimizing long-term health benefits; on the contrary, if the recovery is not good, reducing the discount factor can strengthen the weight of current rewards, help adjust the training load, avoid excessive fatigue, and ensure that the user's health and training effects are fully guaranteed. This dynamic adjustment mechanism makes the training plan more personalized and flexible, and can respond to changes in the user's physical condition in real time to improve the overall effect.

[0045] S4. Store and manage all data in a relational database; All data is stored and managed in a relational database, including: Select a relational database to store and manage data and its analysis results, design the database table structure to store different types of data, set up regular backup tasks, back up all data in the database, manage the permissions of database users and encrypt and store data.

[0046] Storing and managing all data in a relational database can effectively improve the organization and queryability of data. By choosing a relational database to store and manage data and its analysis results, the integrity, consistency and efficient access of data can be ensured. A well-designed database table structure can classify and store different types of data, enhancing the flexibility and scalability of data management. Setting regular backup tasks can ensure data security and recoverability, avoiding data loss due to system failures or human errors. Through permission management, the reasonable allocation of operating permissions of different users in the database can be ensured to prevent unauthorized access and further ensure data security. Encrypted storage of data not only enhances the security of information, but also meets privacy protection and compliance requirements, thereby effectively improving the reliability, stability and data security of the system.

[0047] like Figure 2 As shown, this embodiment also provides a sports data management system based on AI physical measurement, including: Data collection module, used to collect and pre-process the motion data of each user; The load forecasting module is used to divide the user intervals using the K-means clustering algorithm, build the load forecasting model using the fuzzy BP neural network, and obtain the final load forecast value using the improved adaptive genetic algorithm; The strategy adjustment module is used to build a regression model using the QHD reinforcement learning algorithm and dynamically adjust the discount factor output training load adjustment strategy to adjust the discount factor based on feedback; The data storage module is used to store and manage all data in a relational database.

[0048] This embodiment also provides a computer device, which is suitable for the motion data management method based on AI physical measurement, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the motion data management method based on AI physical measurement as proposed in the above embodiment.

[0049] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0050] The present embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the motion data management method based on AI body measurement as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, disk or optical disk.

[0051] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0052] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for managing sports data based on AI physical measurement, characterized in that: include, Collect and pre-process the motion data of each user; The K-means clustering algorithm is used to divide the user intervals, the fuzzy BP neural network is used to build the load forecasting model, and the improved adaptive genetic algorithm is used to obtain the final load forecast value; Use the QHD reinforcement learning algorithm to build a regression model and dynamically adjust the discount factor to output a training load adjustment strategy, and adjust the discount factor based on feedback; All data is stored and managed in a relational database.

2. The method for managing sports data based on AI physical examination according to claim 1, characterized in that: The collecting of the motion data of each user for preprocessing and forming a feature vector comprises: Collect the motion data of each user at each time point during the training cycle through the collection device; The motion data includes physiological data and environmental data; Preprocessing the collected motion data of each user; The preprocessing includes using a moving average method to denoise the motion data, using a linear interpolation method to fill in missing data, using a standard deviation method to detect and remove outliers, and using a Z-Score standardization method to standardize all motion data, and mapping the value of each motion data to between [0,1].

3. The method for managing sports data based on AI physical examination according to claim 2, characterized in that: The use of the K-means clustering algorithm to divide the user intervals includes: According to the motion data of all users, the number of clusters is set to 3, and the motion data of 3 users are randomly selected as the initial cluster centers. The Euclidean distance between each user's motion data and all initial cluster centers is calculated, and the user is assigned to the cluster center with the closest distance. The mean of the motion data of all users in each cluster is recalculated as the new cluster center. The above steps are repeated until the cluster centers converge. Each cluster corresponds to a training load interval. The training load intervals include a primary interval, an intermediate interval and an advanced interval; After clustering is completed, the silhouette coefficient is used to measure the similarity of each user to its cluster and to other clusters.

4. The method for managing sports data based on AI physical examination according to claim 3, characterized in that: The method of using fuzzy BP neural network to construct a load forecasting model and using improved adaptive genetic algorithm to obtain the final load forecasting value includes: Each data in each user's motion data is used as a feature, and the membership value of each feature is calculated using a Gaussian membership function; Combine all the membership values ​​of all the features of each user at each time point into a membership matrix, where each row represents a feature and each column represents the membership value of the feature in different fuzzy sets. The membership matrix of each user at each time point is taken as a sample. Select fuzzy BP neural network as load forecasting model and build the structure of load forecasting model; The weights and biases are initialized by random initialization method, and the membership matrix is ​​used as input to train the load forecasting model. According to the error of the output layer, the weights and biases in the network are optimized by back propagation algorithm until the error converges; The weights and biases of all nodes in the load forecasting model are organized into chromosomes, and each weight and bias is regarded as a gene; Set the size of the initial population, use uniform distribution to generate each chromosome in the initial population, the gene value of each chromosome is a random real number, and select the mean square error as the fitness function to calculate the initial fitness value of each chromosome. The formula is: ； ； in, Represents chromosome The initial fitness value of Represents chromosome The output error value is represents the number of samples, represents the load target value of the i-th sample, represents the load forecast value of the i-th sample; The load target value refers to the load prediction value based on the history of all users, and different load target values ​​are assigned to the three training load intervals through expert analysis; Construct an improved fitness function and calculate the final fitness value of each chromosome. The formula is: ； ； in, Represents chromosome The final fitness value of Represents the improved chromosome Output error value, represents the weighting coefficient of the i-th sample, represents the time point of the i-th sample, represents the maximum time point in the sample, represents the standard deviation of the Gaussian function, represents the dynamic discount factor, represents the load forecast value of the jth sample; The dynamic discount factor refers to a function that uses the Sigmoid function as the growth function of the dynamic discount factor, and the expression is: ； in, represents the maximum value of the dynamic discount factor, represents the control coefficient, represents the central time point of the dynamic discount factor; According to the combination of weight and bias of each chromosome, the chromosome is evaluated by the improved fitness function, and the weight and bias are continuously optimized through the crossover and mutation operations of the genetic algorithm to minimize the prediction error; Calculate the probability of each chromosome being selected ; After calculating the probability based on the fitness value, the roulette wheel method is used to simulate the random selection of chromosomes for the next generation; By performing crossover and mutation operations on the selected chromosomes, the chromosomes in the population are optimized generation by generation. After each generation of selection, crossover and mutation operations, the prediction performance of each chromosome is evaluated using the fitness function, and the most optimized chromosome is selected for the next generation; Set an error threshold. When the value of the fitness function is less than the error threshold, stop the algorithm, set the current time point as the central time point of the dynamic discount factor and output the final load forecast value, indicating that the expected forecast accuracy has been achieved. Otherwise, repeat the iteration.

5. The method for managing sports data based on AI physical examination according to claim 4, characterized in that: The training load adjustment strategy of using the QHD reinforcement learning algorithm to build a regression model and dynamically adjust the discount factor output includes: Define the baseline discount factor ; The QHD reinforcement learning algorithm is used to build a regression model. The purpose of the regression model is to maximize the future cumulative return and obtain the final load forecast value and dynamic discount factor. and the baseline discount factor Substitute the corresponding membership matrix and use the membership matrix as the state space of the regression model And define the action space , the membership matrix is ​​mapped to the high-dimensional feature space through the QHD algorithm; The action space represents the actions that can be selected by the regression model at a certain point in time, including training intensity, training duration and training frequency. The action space is expressed as: ,in, Represented as the training intensity selected at time point t , training duration and training frequency Actions; The training intensity includes low intensity, medium intensity and high intensity; The training duration includes short, medium and long time; The training frequency includes low frequency, medium frequency and high frequency; Set the advantage function, the formula is: ； in, is the advantage function, which represents the advantage value of the action selected based on the current strategy at time point t. is the discounted return, expressed in terms of the dynamic discount factor Next, the future cumulative return of the training strategy from the time point t after the action is executed to the end of the training cycle is, is the state value function, which means that in the state Next, according to the current strategy The expected return after The discount return The formula is: ； in, is an immediate reward, indicating that at a certain point in time The rewards you get when It indicates the number of future time points starting from time point t; The instant reward refers to setting a heart rate target range for the user, collecting the user's heart rate in real time and comparing and determining the instant reward. If the user's heart rate is within the heart rate target range, it means that the training load is appropriate and the instant reward is a positive reward. Otherwise, the instant reward is a negative reward. The state value function The formula is: ； in, is the expected value, indicating that in the state Next, according to the current strategy The expected return after execution; During the training process, the regression model calculates the advantage function, evaluates the current training load adjustment, and uses the regression model to optimize the training load, adjust the training intensity, training duration and training frequency to maximize the return of the current strategy; After each round of training, the discount factor is adjusted in real time according to the value of the advantage function: like , then the dynamic discount factor The value is high and needs to be reduced , reduce the impact of future cumulative returns on current decisions; like , then the dynamic discount factor Effective, need to increase , strengthen the impact of future cumulative returns on current decisions; like , then the dynamic discount factor At the optimal point, stop adjusting; If the dynamic discount factor At the optimum, the regression model outputs the optimal action as the training load adjustment strategy.

6. The method for managing sports data based on AI physical examination according to claim 5, characterized in that: The adjusting the discount factor according to the feedback comprises: The user's training plan is adjusted according to the obtained training load adjustment strategy. During the training load adjustment process, the discount factor is dynamically adjusted according to the real-time feedback data.

7. The method for managing sports data based on AI physical examination according to claim 6, characterized in that: The storing and managing of all data in a relational database includes: Select a relational database to store and manage data and its analysis results, design the database table structure to store different types of data, set up regular backup tasks, back up all data in the database, manage the permissions of database users and encrypt and store data.

8. A sports data management system based on AI physical measurement based on the sports data management method based on AI physical measurement according to any one of claims 1 to 7, characterized in that: include, Data collection module, used to collect and pre-process the motion data of each user; The load forecasting module is used to divide the user intervals using the K-means clustering algorithm, build the load forecasting model using the fuzzy BP neural network, and obtain the final load forecast value using the improved adaptive genetic algorithm; The strategy adjustment module is used to build a regression model using the QHD reinforcement learning algorithm and dynamically adjust the discount factor output training load adjustment strategy to adjust the discount factor based on feedback; The data storage module is used to store and manage all data in a relational database.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the motion data management method based on AI body measurement are implemented as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the motion data management method based on AI body measurement are implemented as described in any one of claims 1 to 7.

Images