An intelligent pushing method and system for personalized precise fitness programs

Abstract

The application discloses a kind of individualized precision fitness scheme intelligent push method and system, it is related to intelligent fitness technical field, including: real-time acquisition user's exercise data and pre-processing;Collect the fitness goal of user and convert into target vector, according to the matching degree of user's exercise data and target vector, activate the task subnetwork in dendritic nerve network, the weight and importance of different tasks in task subnetwork are dynamically adjusted by attention mechanism, and the initial fitness scheme is generated and optimized by multi-agent. Real-time activation most relevant subnetwork and adjust task priority by attention mechanism, to adapt to environmental change and user state, optimize exercise intensity using fuzzy logic reasoning, automatically adjust training task according to user's real-time feedback, the balance between tasks is optimized by the synergistic evolution between agents, so that individualized fitness scheme can be continuously optimized, more accurate exercise recommendation is provided under dynamic environment and user demand change.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent fitness, and particularly to an intelligent pushing method and system for personalized and precise fitness programs. BACKGROUND

[0002] With the continuous development of technology, intelligent pushing of personalized fitness programs has become an important research direction in the fitness industry. Traditional fitness programs are mostly based on a one-size-fits-all approach, which is difficult to effectively consider individual differences and actual needs of each user. In recent years, with the popularity of wearable devices, smartphones and fitness applications, more and more data are used to monitor and analyze the health status and athletic performance of users in real time. In the prior art, methods for pushing fitness programs to users based on sensor data (such as heart rate, step count, exercise duration, etc.) have gradually become mainstream. Through these data, intelligent systems can help users better select appropriate exercise types and intensities to achieve fitness goals. However, existing technical solutions usually ignore the multi-dimensional needs of users (such as environmental conditions, psychological state, etc.), and often cannot make personalized dynamic adjustments according to real-time feedback of users, especially in cases where the environment changes greatly or the user's state is unstable. The existing system responds to user feedback with a lag, making it difficult to ensure long-term effectiveness and continuous optimization of user experience.

[0003] Existing personalized fitness pushing systems usually rely on traditional data processing methods and fixed exercise programs, lacking the ability to flexibly adjust to environmental factors, user states and the diversity of fitness goals. Some systems may fail to adjust exercise intensity or type in time when the user is in a high-temperature or low-temperature environment, resulting in user discomfort or poor results. Existing technologies are mostly based on simple linear models or rule engines, which are difficult to handle complex, multi-dimensional user data and task priorities, leading to difficulty in accurately matching fitness programs to actual user needs. Even some more complex models, such as deep learning models or reinforcement learning models, are applied in personalized recommendations, they usually have weak processing capabilities for user data and fail to fully explore the impact of factors such as environmental changes and psychological state on fitness programs. Due to the shortcomings of these technologies, user experience and fitness results are often affected, making it difficult to form truly personalized and intelligent pushing. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides an intelligent pushing method and system for personalized and precise fitness programs to solve the problem that existing personalized fitness pushing systems usually rely on traditional data processing methods and fixed exercise programs, lacking the ability to flexibly adjust to environmental factors, user states and the diversity of fitness goals.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In the first aspect, the present application provides an intelligent pushing method for personalized and precise fitness programs, which comprises:

[0008] Real-time collection of user exercise data and preprocessing;

[0009] The exercise data includes user heart rate, user blood pressure, energy consumption, ambient temperature, and ambient humidity;

[0010] Collection of user fitness goals and conversion into a target vector, activation of a task subnetwork in a dendritic neural network according to the matching degree of the user exercise data and the target vector, dynamic adjustment of the weight and importance of different tasks in the task subnetwork through an attention mechanism, and generation and optimization of a preliminary fitness program through multiple agents;

[0011] Prediction of exercise data based on the preliminary fitness program using an LSTM model and adjustment of exercise intensity through fuzzy logic reasoning, dynamic adjustment of task weight and routing according to context, and generation of a final fitness program;

[0012] Pushing of the final fitness program to the user through the device interface and recording of user adjustment content through a log, storage of all data in a database and management.

[0013] As a preferred solution of the intelligent pushing method for personalized and precise fitness programs, the user input fitness goals are converted into a target vector, the matching degree of exercise data tasks is calculated and the corresponding subnetwork of the dendritic neural network is activated, the importance is adjusted through an attention mechanism, the agent co-evolution optimization strategy is set, and the preliminary fitness program is generated, which comprises:

[0014] The user inputs the current fitness goal and converts it into a target vector H, maps the target vector into a multi-dimensional space, each dimension corresponds to a task F, constructs a task feature weight matrix based on the exercise data and evaluates the importance of different tasks, each task has a set of independent weight matrices W i , the inner product of each weight matrix W i and the target vector H is calculated as the matching degree of the task and the current fitness goal, the matching degrees of each task are summed to obtain the activation degree of each task;

[0015] Define multiple task subnetworks in the dendritic neural network, match the task subnetworks through the activation degree of each task, input the target vector H into the shared layer of the dendritic neural network, and the shared layer outputs a shared target vector Y s for all tasks;

[0016] Adjust the importance of the task using an attention mechanism, and the task subnetwork adjusts the target vector H and the shared target vector Ys generate a query matrix, a key matrix and a value matrix for each task, and calculate attention weights A by scaled dot-product attention i as a representation of the relevance of each task to the shared target vector;

[0017] define each task as an independent agent, each agent containing a state space s and an action space a;

[0018] Each agent initializes the target vector H according to the current task, and updates the attention weights of the task subnetwork, and sets the reward function R of each agent j ;

[0019] According to the reward value of each agent, the priority P of the task is calculated i Each agent optimizes its strategy through reinforcement learning, and each agent updates its strategy based on the reward and feedback information;

[0020] Agents evolve collaboratively by sharing experience, each agent shares its learned strategy with other agents, and updates its learning strategy;

[0021] In the process of collaborative evolution, the strategies of multiple agents are fused to obtain the final strategy by weighted summation of the agent strategies weighted by the task priority;

[0022] Use mean square error as loss function, in the training process, the loss function is transmitted to each layer of the dendritic neural network through the back propagation algorithm, and the weight matrix of the shared layer is updated by the Adam optimization algorithm to generate a preliminary fitness program.

[0023] As a preferred scheme of the personalized precise fitness program intelligent pushing method of the application, wherein: the LSTM model is used to predict the motion data based on the preliminary fitness program, comprising:

[0024] The user trains according to the generated personalized fitness program, converts the real-time collected motion data into time series data using a sliding window and extracts statistical features as input features;

[0025] Using principal component analysis, the first U features most affecting the prediction are selected from the input features, the selected input features are divided into a training set and a test set, an LSTM model is used for time-dependent modeling, mean square error is used as a loss function, Adam is used as an optimizer, a validation set is used to evaluate the model, root mean square error is calculated, the trained LSTM model is used to predict real-time motion data, and predicted motion data is output according to the regression output.

[0026] As a preferred scheme of the personalized precise fitness program intelligent pushing method, the adjusting exercise intensity by fuzzy logic reasoning comprises:

[0027] The predicted exercise data is divided into health data and environmental data and integrated into a context vector by splicing operation;

[0028] The membership value of the context vector is calculated using fuzzy logic, and each data point in the context vector is mapped into a fuzzy set using a Sigmoid function as a Gaussian membership function, and the matching degree between the health data ;

[0029] According to the membership of the health data and the environmental data, the rules for adjusting the preliminary fitness program are generated, and the fuzzy reasoning method is used to set health threshold and environmental threshold The membership values of the health data and the environmental data are converted into the adjustment amount of the exercise intensity:

[0030] If and , switch to the next task;

[0031] If and , reduce the exercise intensity;

[0032] If and , increase the exercise intensity;

[0033] If and , stop training;

[0034] Adjust the exercise intensity according to the result of fuzzy logic reasoning;

[0035] If the calculated exercise intensity exceeds the maximum recommended intensity, adjust the exercise program to a recovery exercise.

[0036] As a preferred scheme of the personalized precise fitness program intelligent pushing method, the adjusting exercise intensity by fuzzy logic reasoning comprises:

[0037] According to the importance of each feature, the priority P i of each task and the context vector are dynamically calculated to obtain the task weight of each task;

[0038] Based on the task weight of each task, the most suitable task routing path is selected;

[0039] When the state of the user changes, the task weight and the routing path are recalculated according to the new context vector.

[0040] According to the selected routing path, the task information flows to the corresponding task sub-network and is processed according to the priority of the task;

[0041] In the task execution process, the state of the user is monitored in real time, including heart rate, exercise intensity, fatigue, and the task routing and information flow path are dynamically adjusted according to the real-time feedback of the user, and the final fitness scheme is generated.

[0042] As a preferred scheme of the personalized precise fitness scheme intelligent pushing method, wherein: the device interface pushes the final fitness scheme to the user and records the user's adjustment content through the log includes:

[0043] According to the current environment adjustment, the generated final fitness scheme is displayed through the device interface, the user manually adjusts according to the final fitness scheme, and the user's adjustment content is recorded through the log management function;

[0044] The final fitness scheme includes specific exercise types, intensity and duration.

[0045] As a preferred scheme of the personalized precise fitness scheme intelligent pushing method, wherein: the all data is stored in the database and managed includes:

[0046] The relational database is selected to manage the data and the relationship analysis result, the database table structure is designed to store different types of data, the periodic backup task is set, all data in the database is backed up, the database user is managed, and the static data is stored by encryption.

[0047] In a second aspect, the application provides a personalized precise fitness scheme intelligent pushing system, comprising:

[0048] The data acquisition module is used for real-time acquisition and preprocessing of the user's exercise data;

[0049] The model construction module is used for converting the user input fitness target into a target vector, calculating the exercise data task matching degree, and activating the corresponding sub-network of the dendritic neural network;

[0050] The scheme generation module is used for adjusting importance through the attention mechanism, setting the agent collaborative evolution optimization strategy, and generating the preliminary fitness scheme;

[0051] The scheme adjustment module is used for predicting the exercise data based on the preliminary fitness scheme using the LSTM model, adjusting the exercise intensity through fuzzy logic reasoning, dynamically adjusting the task weight and routing according to the context, and generating the final fitness scheme;

[0052] A scheme pushing module is configured to push the final fitness scheme to the user through the device interface and record the user's adjustment content through a log, store all data to a database and manage the data.

[0053] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the computer program, when executed by the processor, implements any step of the intelligent pushing method of the personalized precise fitness scheme according to the first aspect of the present application.

[0054] In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program, and wherein the computer program, when executed by a processor, implements any step of the intelligent pushing method of the personalized precise fitness scheme according to the first aspect of the present application.

[0055] The present application has the following beneficial effects: the most relevant sub-network is activated in real time, the task priority is adjusted through the attention mechanism to adapt to the environmental changes and the user state, the exercise intensity is optimized by fuzzy logic reasoning, the training task is automatically adjusted according to the real-time feedback of the user, the balance between tasks is optimized through the collaborative evolution of agents, the personalized fitness scheme can be continuously optimized, and more precise exercise recommendations can be provided under dynamic environment and user demand changes. BRIEF DESCRIPTION OF DRAWINGS

[0056] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0057] Figure 1 The flowchart of the intelligent pushing method of the personalized precise fitness scheme in embodiment 1.

[0058] Figure 2 The schematic diagram of the intelligent pushing system of the personalized precise fitness scheme in embodiment 1. DETAILED DESCRIPTION

[0059] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail in combination with the drawings of the specification.

[0060] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the concept of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0061] Secondly, the "one embodiment" or "embodiment" referred to herein is intended to represent a specific feature, structure, characteristic, or combination of features and characteristics, which can be included in at least one implementation of the present application. The various appearances of "in one embodiment" or "in an embodiment" in the specification are not all referring to the same embodiment, nor are they mutually exclusive, although some embodiments can include more features than others.

[0062] Embodiment 1, Reference Figure 1 and Figure 2 As a first embodiment of the present application, the embodiment provides an intelligent pushing method for personalized precise fitness programs, including the following steps:

[0063] S1, real-time collection of user's exercise data and preprocessing;

[0064] Specifically, real-time collection of user's exercise data and preprocessing includes:

[0065] The user wears a smart collection device (such as a smart watch or a fitness bracelet), environmental sensors are deployed at the training site, the collection frequency is set, the user's exercise data is collected in real time, the collected exercise data is de-duplicated, the repeated data records are removed, the missing data is filled in by interpolation method, the abnormal values are corrected by statistical analysis and standardized processing;

[0066] The exercise data includes user's heart rate, user's blood pressure, energy consumption, environmental temperature, and environmental humidity.

[0067] Real-time collection and preprocessing of user's exercise data can significantly improve the accuracy and personalization of fitness programs. Through the combination of smart devices (such as smart watches or fitness bracelets) and environmental sensors, the system can accurately monitor the user's exercise state and environmental conditions. Data de-duplication, missing value filling, abnormal value correction and standardization processing can ensure the integrity and reliability of the data, providing high-quality input for subsequent analysis and decision-making. These preprocessing steps can effectively eliminate noise and errors, making the exercise data more accurate, optimizing the pushing effect of fitness programs, and improving training effectiveness and user experience.

[0068] S2, collection of user's fitness goals and conversion into target vectors, activation of task subnetworks in the dendritic neural network according to the matching degree of user's exercise data and target vectors, dynamic adjustment of the weights and importance of different tasks in the task subnetworks through attention mechanism, and generation and optimization of preliminary fitness programs through multi-agent;

[0069] Specifically, the user inputs the fitness goal into the target vector, calculates the exercise data task matching degree and activates the corresponding subnetwork of the dendritic neural network, adjusts the importance through the attention mechanism, sets the agent co-evolution optimization strategy and generates the preliminary fitness program, including:

[0070] The user inputs the current fitness goal and converts it into a goal vector H, which is mapped into a multi-dimensional space, each dimension corresponding to a task F, such as aerobic exercise, diet control, strength training, etc., each task representing a value between 0 and 1, 0 indicating no focus and 1 indicating complete focus, a task feature weight matrix is constructed based on exercise data and the importance of different tasks is evaluated, each task has a set of independent weight matrices W i , each weight matrix W i represents the relationship between the task and other tasks, the weight matrix W i has dimensions m x p, where m represents the task dimension and p represents the dimension of the exercise data, the inner product of each weight matrix W i and the goal vector H is calculated as the matching degree of the task and the current fitness goal, the matching degrees of each task are summed to obtain the activation degree of each task.

[0071] A plurality of task sub-networks in the dendritic neural network are defined, and each task sub-network is matched by the activation degree of each task, the task sub-network is a set of network modules in the dendritic neural network for processing specific tasks or sub-tasks, each sub-network focuses on processing information related to a specific task, and the system dynamically selects and activates the corresponding sub-network for task processing according to the user's demand or input data, for example, if the activation degree of the fat loss task is the highest, the fat loss sub-network is selected to generate a fitness plan;

[0072] The goal vector H is input into the shared layer of the dendritic neural network, the shared layer is responsible for processing features shared by all tasks, the weight matrix of the shared layer is initialized, and when the goal vector H passes through the shared layer, it is multiplied by the weight matrix of the shared layer as the output, the shared layer outputs a shared goal vector Y s for all tasks, the formula is:

[0073] Y s = W s H + b s ,

[0074] where Y s represents the output vector of the shared layer, W s represents the weight matrix of the shared layer, which is used to extract general features from the goal vector input by the user for multiple tasks, H represents the goal vector, and b s represents the bias term;

[0075] The importance of the task is adjusted using an attention mechanism, and the task sub-network generates a query matrix, a key matrix and a value matrix for each task based on the goal vector H and the shared goal vector Y s , the attention weight A i is calculated as the relevance of each task to the shared goal vector through scaled dot-product attention, the formula is:

[0076]

[0077] where A i represents the attention weight of the i-th task, Q i represents the query vector of the i-th task, K i represents the key vector of the i-th task, d k represents the scaling factor, usually the dimension of the key vector, V i represents the value matrix.

[0078] Each task is defined as an independent agent, each agent contains a state space s and an action space a, the state space includes the user's motion data, and the action space includes different fitness plans in the task (such as training type, intensity, duration, etc.);

[0079] Each agent initializes the target vector H according to the current task, and updates the attention weight of the task subnetwork, sets the reward function R j of each agent, the formula is:

[0080] R i = A i · G(H i ),

[0081] where R i represents the reward value of the i-th agent, G(H i ) represents the task completion degree of the i-th agent, taking a decimal value between 0 and 1, 0 represents not completed, and 1 represents completely completed task target;

[0082] According to the reward value of each agent, the priority P i of the task is calculated, the priority P i of each task is the ratio of the reward value of the task to the sum of the reward values of all tasks, and the fitness plan with high priority will get more resources and attention;

[0083] Each agent optimizes its strategy through reinforcement learning, so that the agent can choose the best action to maximize the long-term reward, each agent updates the strategy based on the reward and feedback information, the formula is:

[0084]

[0085] where, represents the optimal strategy of the i-th agent in the state space s, E[·] represents the expected value (weighted average) of all reward values from the initial state according to the strategy π i until the time step T, T represents the maximum time step, γ represents the discount factor, which represents the importance of future rewards, Rτ represents the reward value of the agent at time step τ;

[0086] The policy refers to a function or mapping that defines what action the agent should take in each state, for the i-th agent, the policy π i is a rule for selecting an action based on the current state;

[0087] In order to accelerate co-evolution, agents co-evolve by sharing experience (such as the effectiveness of training programs, user feedback, etc.), each agent shares its learned strategy with other agents, updates its learning strategy, and the formula is:

[0088]

[0089] where π shared represents the joint strategy shared by all agents;

[0090] In the process of co-evolution, the strategies of agents are weighted and summed by task priority to obtain the final strategy, and the formula is:

[0091]

[0092] where π final (s) represents the final strategy of all agents in state space s, and n represents the total number of agents;

[0093] In the process of co-evolution, the interaction between tasks is dynamically adjusted, and the strategy of each task will not only be optimized according to its own reward, but also be adjusted according to the strategy update of other tasks. Through the interaction between strategies, the system can find the optimal balance between tasks, thereby improving the effect of the overall fitness program;

[0094] The mean square error is used as the loss function, and in the training process, the loss function is transmitted to each layer of the dendritic neural network through the back propagation algorithm, and the weight matrix of the shared layer is updated through the Adam optimization algorithm to generate a preliminary fitness program.

[0095] By converting the user's fitness goal into a target vector and combining the task subnetwork of the dendritic neural network, the attention mechanism, and the multi-agent collaborative evolution mechanism of the ACE-D algorithm, the system can dynamically select and activate the most relevant task subnetwork according to the task matching degree, generate personalized and precise fitness programs for different fitness goals (such as weight loss, muscle gain, etc.), adjust the importance of tasks using the attention mechanism, and make each task fully consider the priority of the target when generating the fitness program, thereby optimizing the effectiveness of the training plan. Through reinforcement learning, the strategy of each agent is optimized, and collaborative evolution and task priority weighting adjustment further improve the intelligence and adaptability of the system, maximizing the training effect of each task and achieving efficient achievement of long-term fitness goals. The system updates the experience and strategy through sharing, ensuring continuous optimization of the fitness program in a multi-task and dynamic environment, greatly improving the accuracy of personalized recommendations and fitness effectiveness.

[0096] S3, based on the preliminary fitness program, using an LSTM model to predict exercise data and adjusting exercise intensity through fuzzy logic reasoning, dynamically adjusting task weights and routing according to the context and generating a final fitness program;

[0097] Specifically, using an LSTM model to predict exercise data based on the preliminary fitness program includes:

[0098] The user trains according to the generated personalized fitness program, converts the real-time collected exercise data into time series data using a sliding window and extracts statistical features as input features, including mean, standard deviation, maximum and minimum;

[0099] Using principal component analysis, the first U features that have the most influence on prediction are selected from the input features to reduce redundant features, the value of U is determined according to the actual situation, the selected input features are divided into training and test sets, time-dependent modeling is performed using an LSTM model, mean squared error is used as the loss function, Adam is used as the optimizer to speed up convergence and prevent overfitting, the model is evaluated using a validation set, the root mean square error is calculated, and the trained LSTM model is used to predict real-time exercise data. According to the regression output, the predicted exercise data is output.

[0100] By using the LSTM model to model the time dependence of the exercise data, combining the sliding window technique to convert real-time exercise data into time series data and extract statistical features (such as mean, standard deviation, maximum and minimum), the time sequence change rule of exercise data can be effectively captured. Through principal component analysis (PCA), the most predictive features are selected to reduce redundant information and improve the accuracy and efficiency of the model. In the training process of the LSTM model, mean squared error is used as the loss function, and Adam optimizer is used to speed up convergence and prevent overfitting. At the same time, the validation set is used for evaluation and calculation of the root mean square error, so as to ensure the stability and reliability of the prediction results. Using the trained LSTM model to accurately predict real-time exercise data ensures that the implementation effect of the personalized fitness program is more scientific and reasonable.

[0101] Further, adjusting the exercise intensity through fuzzy logic reasoning includes:

[0102] The predicted exercise data is divided into health data (such as heart rate, blood pressure, and other health indicators) and environmental data (such as environmental temperature, humidity, and other environmental indicators), and is integrated into a context vector through splicing operation;

[0103] The membership values of the context vector are calculated using fuzzy logic, and the Sigmoid function is used as a Gaussian membership function to map each data point in the context vector to a fuzzy set. The health data and environmental data are calculated respectively, and the membership values are calculated respectively.

[0104] The matching degree between the health data and the environmental data is calculated, and the formula is:

[0105]

[0106] Where and μ e (X′ e ) represent the membership of health data and environmental data, respectively, X′ h and X′ e represent health data and environmental data, respectively, a h and a e represent the parameters that control the shape of the membership function, c h and c e represent the center point or threshold value of health data and environmental data, respectively;

[0107] According to the membership of health data and environmental data, the rules for adjusting the preliminary fitness program are generated, and the fuzzy reasoning method is used to set the health threshold and the environmental threshold The membership values of health data and environmental data are converted into the adjustment amount of exercise intensity:

[0108] If and then switch to the next task;

[0109] if and then reduce the exercise intensity;

[0110] if and then increase the exercise intensity;

[0111] if and then stop the training;

[0112] According to the results of fuzzy logic reasoning, adjust the exercise intensity, the formula is:

[0113] I = I0 + ΔI = I0 + α · (μ h (X′ h )+ μ e (X′ e )-1),

[0114] where I represents the exercise intensity, I0 represents the initial recommended exercise intensity, usually the basic recommended value, ΔI represents the adjusted exercise intensity value, α represents the adjustment coefficient, usually set to 0.5;

[0115] If the calculated exercise intensity exceeds the maximum recommended intensity, adjust the exercise program to restorative exercise, such as walking, yoga, etc.

[0116] The method of adjusting exercise intensity through fuzzy logic reasoning can effectively adjust exercise intensity dynamically according to user's health data (such as heart rate, blood pressure, etc.) and environmental data (such as temperature and humidity, etc.). This method can accurately quantify the matching degree of user's exercise state (such as "healthy", "fatigue", "suitable for exercise") and the current environment by calculating the membership degree of health data and environmental data and mapping the data into fuzzy sets using Gaussian membership function. According to these membership values, the system generates rules for exercise intensity adjustment and optimizes exercise intensity adjustment through fuzzy reasoning, such as reducing, increasing or stopping training, etc., so as to ensure that the user's exercise intensity always remains within a reasonable range, avoiding overtraining or insufficient training. When the exercise intensity exceeds the maximum recommended intensity, the system can automatically adjust to restorative exercise (such as walking, yoga, etc.), ensuring the safety and effectiveness of the user's exercise. This intelligent adjustment method improves the personalization, precision and adaptability of fitness plans.

[0117] In addition, dynamically adjusting task weights and routing according to context and generating the final fitness plan includes:

[0118] According to the importance of each feature, weight each task by weighting the priority P iThe task weight of each task is dynamically calculated based on the context vector, and the formula is:

[0119] δ i (t) = f(C(t), P i ),

[0120] where δ i (t) represents the task weight of the i-th task at time point t, C(t) represents the context vector at time point t, f represents the weight function, and the task weight is adjusted according to the context vector, for example, when the user is in a high-temperature environment, the weight of the fat loss task is increased to adapt to the current environment;

[0121] Based on the task weight of each task, the most suitable task routing path is selected, and the formula is:

[0122]

[0123] where R path (t) represents the task routing path at time point t;

[0124] When the user's state changes, the task weight and routing path are recalculated according to the new context vector, for example, when the user enters a high-temperature environment, the system will adjust the routing path and preferentially select the path related to the fat loss task, and the formula is:

[0125]

[0126] where t+Δt represents the time point after time point t, the system recalculates the task weight and routing path to ensure that the priority of the task is correctly adjusted in the new environment;

[0127] According to the selected routing path, the task information flows to the corresponding task sub-network and is processed according to the priority of the task;

[0128] During the task execution process, the user's state is monitored in real time, including heart rate, exercise intensity, and fatigue degree, and according to the user's real-time feedback, the task routing and information flow path are dynamically adjusted to generate the final fitness plan, for example, when the user feedback shows high fatigue degree, the routing path of high-intensity task is reduced and low-intensity task is preferentially selected, and the goal of routing adjustment is to ensure the balance between the user's state and the task, and to optimize the training effect.

[0129] By dynamically adjusting task weights and routing based on the context, this solution can flexibly optimize task priorities based on the user's real-time status and environmental changes (such as high temperature, fatigue, etc.), ensuring the personalization and accuracy of the fitness plan. When the user's status or environment changes, the system recalculates task weights and routing paths, dynamically adjusts the execution order of tasks, and gives priority to tasks that best match the user's current status. This flexible adaptation mechanism can effectively avoid overtraining and improve training effects, while ensuring the safety and comfort of users in different environments and states, thereby maximizing fitness effects.

[0130] S4. Push the final fitness plan to the user on the device interface and record the user's adjustments in the log, and store all data in the database for management;

[0131] Specifically, the final fitness plan is pushed to the user on the device interface and the user's adjustments are recorded in the log, including:

[0132] Adjustments are made to the current environment, and the final fitness plan is displayed through the device interface (such as a smartwatch or mobile phone). The user manually adjusts the plan based on the final fitness plan, and the user's adjustments are recorded through the log management function;

[0133] The final fitness plan includes specific exercise type, intensity and duration.

[0134] This method pushes the final personalized fitness plan to the user through the device interface (such as a smart watch or mobile phone) and adjusts it according to the current environment, allowing the user to obtain the most suitable training plan in real time. At the same time, the user can manually adjust the exercise type, intensity and duration according to personal needs, and record these adjustments through the log management function, thereby achieving comprehensive monitoring and optimization of the fitness process. This solution not only enhances the user's sense of participation and flexibility, but also helps the system accumulate adjustment data to provide support for subsequent personalized recommendations and improvements.

[0135] Furthermore, storing and managing all data in a database involves:

[0136] Select a relational database to manage data and relationship 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 database user permissions, and encrypt and store static data.

[0137] By storing all data into a relational database and effectively managing it, the structured storage and efficient querying of data can be ensured. Designing a reasonable database table structure helps to organize and classify different types of data and analysis results, improves the access speed and management efficiency of data, and sets up regular backup tasks to ensure data security and recovery capabilities, effectively preventing data loss. Database user permission management can ensure the security of the system and prevent unauthorized access or modification. The encryption storage of static data can further enhance the security of data and ensure the protection and compliance of sensitive information, thereby providing an efficient, secure, and reliable data management solution for the overall system.

[0138] The embodiment also provides an intelligent pushing system for personalized precise fitness programs, comprising:

[0139] A data acquisition module is configured to acquire and pre-process the motion data of a user in real time.

[0140] A model construction module is configured to convert the fitness goal input by the user into a target vector, calculate the task matching degree of the motion data, and activate the corresponding subnetwork of the dendritic neural network.

[0141] A program generation module is configured to adjust the importance through an attention mechanism, set an agent co-evolution optimization strategy, and generate a preliminary fitness program.

[0142] A program adjustment module is configured to predict the motion data based on the preliminary fitness program using an LSTM model, adjust the exercise intensity through fuzzy logic reasoning, dynamically adjust the task weight and routing according to the context, and generate a final fitness program.

[0143] A program pushing module is configured to push the final fitness program to the user on the device interface, record the user's adjustment content through a log, store all data into a database, and manage it.

[0144] The embodiment also provides a computer device suitable for the intelligent pushing method for personalized precise fitness programs, comprising a memory and a processor. The memory is configured to store computer executable instructions, and the processor is configured to execute the computer executable instructions to implement the intelligent pushing method for personalized precise fitness programs as described in the above embodiment.

[0145] The computer device can be a terminal, which includes a processor, a memory, a communication interface, a display screen and an input device connected by a system bus. 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 operating system and the computer program in the non-volatile storage medium to run. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be achieved by WIFI, operator network, NFC (Near Field Communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad or mouse, etc.

[0146] The embodiment also provides a storage medium having a computer program stored thereon, the program being executed by a processor to implement the method for intelligently pushing an individualized precise fitness scheme 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 a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic storage, a flash memory, a magnetic disk or an optical disk.

[0147] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application but not limit the present application. Although the present application 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 application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application, and all of them should be covered in the scope of the claims of the present application.

Claims

1. A method for intelligently pushing personalized and precise fitness plans, characterized by: include: Collect user's motion data in real time and perform pre-processing; The exercise data includes user heart rate, user blood pressure, energy consumption, ambient temperature, and ambient humidity; The user's fitness goals are collected and converted into target vectors. The task sub-network in the dendritic neural network is activated based on the matching degree between the user's motion data and the target vector. The weights and importance of different tasks in the task sub-network are dynamically adjusted through the attention mechanism, and a preliminary fitness plan is generated and optimized through multi-agents. Based on the preliminary fitness plan, the LSTM model is used to predict exercise data and adjust the exercise intensity through fuzzy logic reasoning. The task weights and routes are dynamically adjusted according to the context to generate the final fitness plan. The final fitness plan is pushed to the user on the device interface and the user's adjustments are recorded through the log. All data is stored in the database and managed.

2. The personalized precision fitness program intelligent push method according to claim 1, characterized in that: The method collects the user's fitness goals and converts them into target vectors, activates the task sub-network in the dendritic neural network according to the matching degree between the user's motion data and the target vector, dynamically adjusts the weights and importance of different tasks in the task sub-network through the attention mechanism, and generates and optimizes the preliminary fitness plan through multi-agents, including: The user enters the current fitness goal and converts it into a goal vector , mapping the target vector into a multidimensional space, where each dimension corresponds to a task , based on the motion data, the task feature weight matrix is ​​constructed and the importance of different tasks is evaluated. Each task has an independent set of weight matrices , calculate each weight matrix With the target vector The inner product of is taken as the matching degree between the task and the current fitness goal, and the matching degree of each task is summed up to obtain the activation degree of each task; Define multiple task sub-networks in the dendritic neural network, match the task sub-networks by the activation of each task, and transform the target vector Input to the shared layer of the dendritic neural network, the shared layer outputs a shared target vector for all tasks ; The attention mechanism is used to adjust the importance of the task, and the task sub-network is based on the target vector of each task. and the shared target vector , generate the query matrix, key matrix and value matrix for each task, calculate the attention weight by scaling the dot product attention As a representation of the relevance of each task to the shared target vector; Define each task as an independent agent, each agent contains a state space and action space ; Each agent initializes the target vector according to the current task , and update the attention weight of the task sub-network, setting the reward function of each agent ; Calculate the priority of tasks based on the reward value of each agent , each agent optimizes its strategy through reinforcement learning, and each agent updates its strategy based on reward and feedback information; Agents collaborate and evolve through shared experience. Each agent shares its learned strategy with other agents and updates its learning strategy. In the co-evolution process, the strategies of the agents are weighted and summed by task priorities, and the strategies of multiple agents are integrated to obtain the final strategy; The mean square error is used as the loss function. During the training process, the loss function is passed to each layer of the dendritic neural network through the back propagation algorithm, and the weight matrix of the shared layer is updated through the Adam optimization algorithm to generate a preliminary fitness plan.

3. The personalized precision fitness program intelligent push method according to claim 2, characterized in that: The method of using the LSTM model to predict motion data based on the preliminary fitness program includes: Users train according to the generated personalized fitness plan, and the real-time collected motion data is converted into time series data using a sliding window and statistical features are extracted as input features; Use principal component analysis to filter out the top U features that have the greatest impact on prediction from the input features. Divide the filtered input features into training and test sets. Use the LSTM model for time-dependency modeling, use mean square error as the loss function, and adopt Adam as the optimizer. Use the validation set to evaluate the model and calculate the root mean square error. Use the trained LSTM model to predict real-time motion data, and output the predicted motion data based on the regression output.

4. The personalized precision fitness program intelligent push method according to claim 3, characterized in that: The method of adjusting the exercise intensity by fuzzy logic reasoning includes: The predicted motion data is divided into health data and environmental data and integrated into a context vector through a splicing operation; The membership value of the context vector is calculated using fuzzy logic, and the Sigmoid function is used as the Gaussian membership function to map each data point in the context vector to a fuzzy set. and Quantify the matching degree between the user's current motion state and the fuzzy category; According to the membership of health data and environmental data, the rules for adjusting the preliminary fitness program are generated, and the health threshold is set using fuzzy reasoning method. and environmental thresholds , converting the membership values ​​of health data and environmental data into the adjustment amount of exercise intensity: like and , then switch to the next task; like and , then reduce the intensity of exercise; like and , then increase the intensity of exercise; like and , then stop training; Adjust the exercise intensity according to the results of fuzzy logic reasoning; If the calculated exercise intensity exceeds the maximum recommended intensity, the exercise program will be adjusted to recovery exercise.

5. The personalized precision fitness program intelligent push method according to claim 4, characterized in that: The method of dynamically adjusting task weights and routing according to the context and generating a final fitness solution includes: Weighted according to the importance of each feature, by prioritizing each task The task weight of each task is obtained by dynamically calculating the context vector; Select the most appropriate task routing path based on the task weight of each task; When the user's status changes, the task weight and routing path are recalculated based on the new context vector; According to the selected routing path, the task information is directed to the corresponding task sub-network and processed according to the priority of the task; During the task execution process, the user's status is monitored in real time, including heart rate, exercise intensity, and fatigue. Based on the user's real-time feedback, the task routing and information flow path are dynamically adjusted to generate the final fitness plan.

6. The personalized precision fitness program intelligent push method according to claim 5, characterized in that: The method of pushing the final fitness plan to the user on the device interface and recording the user's adjustment content in a log includes: Adjust according to the current environment, and display the final fitness plan generated through the device interface. Users can manually adjust according to the final fitness plan, and the user's adjustment content is recorded through the log management function; The final fitness plan includes specific exercise type, intensity and duration.

7. The method for intelligently pushing a personalized, precise fitness plan according to claim 6, wherein: Said storing and managing all data in the database includes: Select a relational database to manage data and relationship 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 database user permissions, and encrypt and store static data.

8. A personalized precision fitness program intelligent push system, based on the personalized precision fitness program intelligent push method according to any one of claims 1 to 7, characterized in that: include: Data acquisition module, used to collect user's motion data in real time and perform pre-processing; A model building module is used to convert user input fitness goals into target vectors, calculate the task matching degree of motion data, and activate the corresponding sub-network of the dendritic neural network; The solution generation module is used to adjust the importance through the attention mechanism, set the agent co-evolution optimization strategy and generate the preliminary fitness solution; The plan adjustment module is used to use the LSTM model to predict exercise data based on the preliminary fitness plan and adjust the exercise intensity through fuzzy logic reasoning, dynamically adjust the task weights and routes according to the context, and generate the final fitness plan; The program push module is used to push the final fitness program to the user on the device interface and record the user's adjustment content through the log, and store all data in the database for management.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the personalized precision fitness program intelligent push method according to any one of claims 1 to 7 are implemented.

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 method for intelligently pushing a personalized, precise fitness plan are implemented as described in any one of claims 1 to 7.

Images