An artificial intelligence-based personalized whole-body vibration health index regulation method

Abstract

The present application relates to the technical field of digital health, intelligent hardware and artificial intelligence, in particular to a personalized whole-body vibration health index regulation method based on artificial intelligence, comprising the following steps: S1: based on questionnaire, basic physiology, disease subtype subdivision and real-time environmental multi-dimensional data, the vibration intervention effect is associated through an AI model, and an initial vibration intervention scheme with high adaptability is generated; the present application integrates questionnaire, basic physiology, disease subtype subdivision and real-time environmental multi-dimensional data, and constructs a population subtype-individual feature double-layer evolution model fused with reinforcement learning, not only realizes the accurate personalized generation of the initial vibration intervention scheme, but also relies on real-time monitoring data of wearable devices as dynamic reward signals to drive the model to continuously iterate and optimize, thereby significantly improving the regulation efficiency of health indicators such as blood pressure and blood glucose and the safety of the training process, and effectively reducing the risk of abnormal events.

Description

Technical Field

[0001] This invention relates to the fields of digital health, smart hardware and artificial intelligence, specifically to a personalized whole-body vibration health index regulation method based on artificial intelligence. Background Technology

[0002] Abnormal indicators of metabolic syndrome, such as hypertension, hyperglycemia, and hyperuricemia, have become a global public health problem, seriously threatening human physical and mental health and imposing a heavy health burden and medical pressure on individuals, families, and society. Besides drug treatment and dietary control, regular physical exercise is widely recognized as a safe and effective non-pharmacological intervention that can effectively improve metabolic levels and help regulate abnormal health indicators. Among these, whole-body vibration training, as a novel passive exercise therapy, does not require users to actively engage in high-intensity exercise. Related studies have confirmed that it can positively regulate indicators of metabolic syndrome such as hypertension, hyperglycemia, and hyperuricemia through mechanisms such as activating neuromuscular systems, improving blood circulation, and regulating the endocrine system, and is gradually becoming a research hotspot and important development direction in the field of non-pharmacological intervention.

[0003] However, in current practical applications, existing whole-body vibration training technology still has many obvious limitations, making it difficult to fully leverage its advantages as a non-drug intervention and meet users' personalized and routine health management needs. Specifically, existing whole-body vibration training technologies suffer from limitations such as homogeneous programs that cannot adapt to individual user differences, static interventions that cannot be dynamically adjusted based on body feedback, and the difficulty in optimizing and evolving programs by treating the equipment as an isolated system. This results in varying effects of vibration training programs from person to person, making it difficult to achieve the ideal health indicator regulation effect and potentially leading to training risks due to insufficient program adaptability. At the same time, the isolated equipment design also prevents the effective collection, analysis, and utilization of the massive amounts of user response data generated during training, further limiting the promotion and application of whole-body vibration training technology.

[0004] Based on this, the present invention provides a personalized whole-body vibration health index regulation method based on artificial intelligence to solve the above-mentioned technical problems. Summary of the Invention

[0005] The purpose of this invention is to provide a personalized whole-body vibration health indicator regulation method based on artificial intelligence. This invention integrates multi-dimensional data from questionnaires, basic physiology, disease subtype subdivision, and real-time environment, and constructs a two-layer evolutionary model of population subtype and individual characteristics that integrates reinforcement learning. This not only achieves accurate and personalized generation of initial vibration intervention plans, but also relies on real-time monitoring data from wearable devices as dynamic reward signals to drive continuous iterative optimization of the model. This significantly improves the regulation efficiency of health indicators such as blood pressure and blood sugar, as well as the safety of the training process, and effectively reduces the risk of abnormal events.

[0006] To achieve the above objectives, the present invention provides the following technical solution: This invention provides a personalized whole-body vibration health index regulation method based on artificial intelligence, comprising the following steps: S1: Based on questionnaires, basic physiology, disease subtype subdivision and real-time environmental multi-dimensional data, use AI models to learn the mapping relationship between multi-dimensional features and intervention effects in historical data, and output personalized initial plans for new users including frequency, amplitude, waveform and duration. S2: Construct a two-layer evolutionary model that integrates reinforcement learning with population subtypes and individual characteristics. First, refine the population subtypes and train the subtype-specific sub-models. Then, use the health indicator regulation effect monitored by S3 as a reward to upload the individual optimization data to the central model in the cloud. The cloud aggregates the data under the premise of privacy protection, cross-validates it with the data of users of the same subtype, identifies the intervention pattern of the "gold standard response group" and extracts efficient parameter weights. The optimized model and parameters are then fed back to the terminals of users of the same subtype. S3: Real-time monitoring of dynamic changes in multi-dimensional health indicators through wearable devices, implementation of graded intervention on vibration parameters based on risk level, and simultaneous push of corresponding health advice; S4: Adjust vibration parameters based on the characteristics of simple vibration equipment in different scenarios, and generate fragmented training schemes by combining user daily life data.

[0007] The specific steps of S1 are as follows: S1.1: Integrate user questionnaire information, basic physiological parameters, disease subtype subcategories and real-time environmental variables to construct a comprehensive initial feature input set; S1.2: Standardize, encode, and fuse heterogeneous multidimensional data to generate structured, computable user health status vectors; S1.3: Utilize AI models to learn the mapping relationship between multidimensional features in historical user data and the effects of vibration intervention; S1.4: Based on the current user's fusion characteristics and the trained AI model, output a personalized initial vibration intervention plan containing parameters such as frequency, amplitude, waveform, and duration.

[0008] The disease subtype subclass data includes primary / secondary hypertension and insulin-resistant / insufficient hyperglycemia.

[0009] The specific steps of S2 are as follows: S2.1: Based on multidimensional health and behavioral characteristics, users are divided into fine-grained subgroups with similar response patterns using unsupervised or semi-supervised learning methods; S2.2: Build and train an independent reinforcement learning sub-model for each sub-group to learn the optimal vibration intervention strategy for that sub-group; S2.3: Using the health indicator regulation effect monitored by S3 as a reward signal, the individual interaction data is driven to update the central model synchronously under the reinforcement learning framework and cross-validate with the data of users of the same subtype. The cloud extracts the gold standard intervention pattern and efficient parameter weights by aggregating the data of the same subtype, and sends the validated high-weight parameters and optimized model to the terminals of users of the same subtype. S2.4: Based on a two-layer model of continuous evolution of subtype commonalities and individual characteristics, it dynamically outputs more accurate and adaptive personalized vibration intervention solutions.

[0010] In step S2.2, an independent reinforcement learning sub-model is constructed and trained for each sub-group to learn the optimal vibration intervention strategy for that sub-group. The specific steps are as follows: S2.2.1: Define the state space The user's multidimensional health feature vector at time t includes blood pressure, blood sugar, uric acid, and recent training compliance; S2.2.2: Define the action space Optional combinations of vibration intervention parameters, including frequency, amplitude, and waveform type; S2.2.3: Calculate the reward value based on the real-time feedback of health indicators from S3 and the group coordination factor; the group coordination factor is used to measure the consistency between the individual response strategy and the cloud-based subtype model prediction strategy, guiding the individual plan to quickly converge to the subtype's optimal intervention range. The specific formula is as follows: ; in, Let be the instantaneous reward value at time t. As a weighting coefficient for the effect of blood pressure regulation, Let be the change in blood pressure at time t. This is the weighting coefficient for blood glucose stability. Let be an indicator of blood glucose stability at time t. This represents the penalty coefficient for abnormal events. For abnormal event indicator variables, For group collaboration weighting coefficient, As a group synergy factor; S2.2.4: Employ a deep Q-network to maximize cumulative discount rewards. To achieve the goal, iteratively update the policy network parameters of this subtype sub-model until the model converges to a preset accuracy, where, is the discount factor, and T is the total training cycle duration.

[0011] In S2.3, the health indicator regulation effect monitored in S3 is used as a reward signal to drive individual interaction data to synchronously update the central model under the reinforcement learning framework, and cross-validate it with user data of the same subtype. The cloud extracts the gold standard intervention pattern and efficient parameter weights by aggregating data of the same subtype, and distributes the validated high-weight parameters and optimized model to the terminals of users of the same subtype. The specific operation steps are as follows: S2.3.1: Obtain the regulatory effect of the health indicators monitored by S3, quantify it into a reward signal for reinforcement learning, and calculate the reward value using the following formula: ; in, Let i be the reward value for the i-th user. For changes in health indicators, For the stability of health indicators, This represents the abnormal risk coefficient. , , These are the weighting coefficients; S2.3.2: The cloud uses privacy-preserving data aggregation technology to input individual users' interaction data and corresponding reward signals into the reinforcement learning framework, and synchronously update the policy parameters of the central model; S2.3.3: Extract historical interaction data and reward signals from users of the same subtype, and perform cross-validation on the updated central model to verify the model's generalization ability in this subtype group; S2.3.4: If the cross-validation accuracy does not reach the preset threshold, the central model parameters are iteratively adjusted until the model meets the generalization requirements.

[0012] The specific steps of S3 are as follows: S3.1: Continuously acquire dynamic data streams of key health indicators such as blood pressure, blood sugar, heart rate, and uric acid during the user's vibration training process through wearable devices; S3.2: Utilize AI models to analyze real-time monitoring data and automatically detect risk events that exceed safety thresholds or exhibit abnormal fluctuations; S3.3: Based on the risk identification results, implement graded intervention: for mild abnormalities, fine-tune vibration parameters; for severe abnormalities, immediately trigger training shutdown. S3.4: Synchronously generate and push non-diagnostic matching suggestions that match the current physiological state.

[0013] In step S3.2, an AI model is used to analyze real-time monitoring data and automatically detect risk events that exceed safety thresholds or exhibit abnormal fluctuations. The specific operation steps are as follows: S3.2.1: Preprocess the real-time health indicator data collected in S3.1 to remove noise and fill in missing values; S3.2.2: Compare the preprocessed data with the preset safety threshold to detect whether there are any risk events where the indicators exceed the threshold; S3.2.3: Calculate the fluctuation range of health indicators within a preset time window, and identify abnormal risks through abnormal fluctuation detection; S3.2.4: Integrate the detection results of threshold exceeding and abnormal fluctuations, and output the risk event type and trigger level.

[0014] S3.2.3 identifies abnormal risks through abnormal fluctuation detection, and the specific formula is as follows: ; in, Let be the health indicator value at time t. This represents the average of the indicators over the previous w time windows. The standard deviation of the index for the first w time windows. This is the fluctuation sensitivity coefficient.

[0015] The specific steps of S4 are as follows: S4.1: Based on the physical characteristics of the stiffness and motor torque of simple vibration equipment used in home, office and other scenarios, the energy value output by the AI ​​model is automatically converted into the gear parameters that the equipment can execute, and the vibration intervention parameters are automatically mapped and adjusted. S4.2: Based on users' daily routine data, intelligently plan personalized vibration training programs with short durations and high frequencies.

[0016] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention integrates multi-dimensional data from questionnaires, basic physiology, disease subtype subdivision, and real-time environment, and constructs a two-layer evolutionary model of population subtype and individual characteristics that incorporates reinforcement learning. This not only enables the precise and personalized generation of initial vibration intervention plans, but also relies on real-time monitoring data from wearable devices as dynamic reward signals to drive continuous iterative optimization of the model. This significantly improves the regulation efficiency of health indicators such as blood pressure and blood sugar, as well as the safety of the training process, and effectively reduces the risk of abnormal events.

[0017] 2. This invention provides an efficient, convenient, and feasible technical approach for routine health management in home and office settings by adapting the characteristics of simple vibration devices to various scenarios such as home and office, and by combining user lifestyle data to intelligently plan fragmented training programs. Attached Figure Description

[0018] Figure 1 This is a system diagram of a personalized whole-body vibration health index regulation method based on artificial intelligence according to the present invention.

[0019] Figure 2This is a flowchart of the real-time safety control closed loop in a personalized whole-body vibration health index control method based on artificial intelligence according to the present invention. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0021] Example: like Figures 1-2 As shown, this embodiment provides a personalized whole-body vibration health indicator regulation method based on artificial intelligence, including the following steps: S1: Based on questionnaires, basic physiology, disease subtype subdivision, and real-time environmental multi-dimensional data, the AI ​​model learns the mapping relationship between multi-dimensional features and intervention effects in historical data, and outputs a personalized initial plan including frequency, amplitude, waveform, and duration for new users; S2: Construct a two-layer evolutionary model of population subtype and individual characteristics that integrates reinforcement learning. First, finely subdivide the population subtype and train the sub-model for each subtype. Then, using the health indicator regulation effect monitored in S3 as a reward, the individual optimized data is uploaded to the central model in the cloud. The cloud aggregates data under the premise of privacy protection, cross-validates with data of users of the same subtype, identifies the intervention pattern of the "gold standard response group", extracts efficient parameter weights, and feeds back the optimized model and parameters to the terminals of users of the same subtype; S3: Monitor the dynamic changes of multi-dimensional health indicators in real time through wearable devices, perform graded intervention on vibration parameters based on risk level, and push corresponding health suggestions simultaneously; S4: Adjust vibration parameters based on the characteristics of simple vibration devices in different scenarios, and generate fragmented training plans by combining user's daily life data.

[0022] The specific steps of S1 are as follows: S1.1: Integrate user questionnaire information, basic physiological parameters, disease subtype subcategories, and real-time environmental variables to construct a comprehensive initial feature input set; disease subtype subcategories include primary / secondary hypertension and insulin resistance / insufficient secretion hyperglycemia. S1.2: Standardize, encode, and fuse heterogeneous multidimensional data to generate a structured, computable user health status vector; S1.3: Use an AI model to learn the mapping relationship between multidimensional features and vibration intervention effects in historical user data; S1.4: Based on the current user's fused features and the trained AI model, output a personalized initial vibration intervention plan containing parameters such as frequency, amplitude, waveform, and duration.

[0023] Furthermore, it should be noted that in S1.2, standardization can adopt Z-score or Min-Max normalization; encoding can adopt one-hot encoding (for disease subtype labels) or time encoding (for real-time environmental variables); fusion can adopt attention mechanism or splicing method, and clarify how to integrate data from different dimensions into a health status vector.

[0024] AI model type in S1.3: Gradient boosting tree (XGBoost) or deep neural network (DNN), and mean squared error (MSE) is used to measure the prediction bias of intervention effect.

[0025] In this embodiment, it should also be noted that the specific steps of S2 are as follows: S2.1: Based on multidimensional health and behavioral characteristics, users are divided into fine-grained subtype groups with similar response patterns using unsupervised or semi-supervised learning methods; S2.2: For each subtype, an independent reinforcement learning sub-model is constructed and trained to learn the optimal vibration intervention strategy for that subtype; the specific operation steps are as follows: S2.2.1: Define the state space. S2.2.2: Define the action space for the user at time t, which includes blood pressure, blood sugar, uric acid, and recent training compliance; The optional vibration intervention parameter combinations include frequency, amplitude, and waveform type; S2.2.3: Calculate the reward value based on the real-time feedback of health indicators from S3 and combined with the group synergy factor; the group synergy factor is used to measure the consistency between the individual response strategy and the cloud-based subtype model prediction strategy, guiding the individual plan to quickly converge to the subtype's optimal intervention range. The specific formula is as follows: ; in, Let be the instantaneous reward value at time t. As a weighting coefficient for the effect of blood pressure regulation, Let be the change in blood pressure at time t. This is the weighting coefficient for blood glucose stability. Let be an indicator of blood glucose stability at time t. This represents the penalty coefficient for abnormal events. For abnormal event indicator variables, For group collaboration weighting coefficient, S2.2.4: Employ a deep Q-network to maximize cumulative discount rewards. To achieve the goal, iteratively update the policy network parameters of this subtype sub-model until the model converges to a preset accuracy, where, , where T is the total training period. S2.3: Using the health indicator regulation effect monitored in S3 as a reward signal, the individual interaction data is used to synchronously update the central model within the reinforcement learning framework, and cross-validated with data from users of the same subtype. The cloud extracts the gold standard intervention pattern and efficient parameter weights by aggregating data from the same subtype, and distributes the validated high-weight parameters and optimized model to the terminals of users of the same subtype. The specific operation steps are as follows: S2.3.1: Obtain the health indicator regulation effect monitored in S3 and quantify it as a reward signal for reinforcement learning. The reward value calculation formula is: ; in, Let i be the reward value for the i-th user. For changes in health indicators, For the stability of health indicators, This represents the abnormal risk coefficient. , , S2.3.2: The cloud uses privacy-preserving data aggregation technology to input individual users' interaction data and corresponding reward signals into the reinforcement learning framework, and synchronously updates the policy parameters of the central model; S2.3.3: Historical interaction data and reward signals of users of the same subtype are extracted and cross-validated on the updated central model to verify the model's generalization ability in this subtype group; S2.3.4: If the cross-validation accuracy does not reach the preset threshold, the central model parameters are iteratively adjusted until the model meets the generalization requirements. S2.4: Based on a continuously evolving two-layer model of subtype commonality + individual characteristics, more accurate and adaptive personalized vibration intervention solutions are dynamically output.

[0026] Furthermore, it should be noted that S2.2.3 is set according to the priority of clinical indicators. , , In S2.2.4, the experience replay pool capacity for deep Q-networks is set to... The target network is updated every 100 steps, improving feasibility.

[0027] In this embodiment, it should also be noted that the specific steps of S3 are as follows: S3.1: Continuously acquire dynamic data streams of key health indicators such as blood pressure, blood sugar, heart rate, and uric acid during the user's vibration training process through wearable devices; S3.2: Analyze the real-time monitoring data using an AI model to automatically detect risk events that exceed safety thresholds or exhibit abnormal fluctuations; the specific operation steps are as follows: S3.2.1: Preprocess the real-time health indicator data collected in S3.1 to remove noise and fill in missing values; S3.2.2: Compare the preprocessed data with preset safety thresholds to detect whether there are risk events where indicators exceed the thresholds; S3.2.3: Calculate the fluctuation range of health indicators within a preset time window and identify abnormal risks through abnormal fluctuation detection; S3.2.4: Integrate the detection results of threshold exceeding and abnormal fluctuations, and output the risk event type and trigger level. The specific formula for identifying abnormal risks through abnormal fluctuation detection in S3.2.3 is as follows: ; in, Let be the health indicator value at time t. This represents the average of the indicators over the previous w time windows. The standard deviation of the index for the first w time windows. S3.3: Based on the risk identification results, implement graded intervention: fine-tune vibration parameters for mild abnormalities, and immediately trigger training shutdown for severe abnormalities; S3.4: synchronously generate and push non-diagnostic matching suggestions that match the current physiological state.

[0028] Furthermore, it should be noted that the rules for setting the safety threshold in S3.2.2 are: based on clinical guidelines (such as setting the blood pressure threshold to 180 / 110 mmHg) or dynamic thresholds based on user baseline data (the current indicator is considered abnormal if it exceeds the baseline by 30%).

[0029] In S3.2.3, the typical value of the time window w is 5-10 minutes, and the fluctuation sensitivity coefficient... The value ranges from 1.5 to 2.5. =2 corresponds to the anomaly detection within the 95% confidence interval.

[0030] S3.3 Intervention Triggering Rules: For mild abnormalities, only the vibration frequency is adjusted; for severe abnormalities, training is directly suspended.

[0031] In this embodiment, it should also be noted that the specific steps of S4 are as follows: S4.1: Based on the physical characteristics of the stiffness and motor torque of the simple vibration equipment used in home, office and other scenarios, the energy value output by the AI ​​model is automatically converted into the gear parameters that the equipment can execute, and the vibration intervention parameters are automatically mapped and adjusted; S4.2: Combined with the user's daily life data, a personalized vibration training plan with short duration and high frequency is intelligently planned.

[0032] Furthermore, it should be noted that the equipment characteristics and parameter mapping rules in S4.1 are as follows: the amplitude range of the vibration seat is 2-5mm, and the amplitude parameters output by the model (such as 3mm) need to be mapped to the gear that the equipment can execute (such as medium gear).

[0033] The time planning logic for fragmented training in S4.2 is as follows: Combine the user's free time in their schedule (such as commuting and lunch break) to generate a training plan of 5-10 minutes each time, 3-4 times a day.

[0034] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0035] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A personalized whole-body vibration health index regulation method based on artificial intelligence, characterized in that, Includes the following steps: S1: Based on questionnaires, basic physiology, disease subtype subdivision and real-time environmental multi-dimensional data, use AI models to learn the mapping relationship between multi-dimensional features and intervention effects in historical data, and output personalized initial plans for new users including frequency, amplitude, waveform and duration. S2: Construct a two-layer evolutionary model that integrates reinforcement learning with population subtypes and individual characteristics. First, refine the population subtypes and train the subtype-specific sub-models. Then, use the health indicator regulation effect monitored by S3 as a reward to upload the individual optimization data to the central model in the cloud. The cloud aggregates the data under the premise of privacy protection, cross-validates it with the data of users of the same subtype, identifies the intervention pattern of the "gold standard response group" and extracts efficient parameter weights. The optimized model and parameters are then fed back to the terminals of users of the same subtype. S3: Real-time monitoring of dynamic changes in multi-dimensional health indicators through wearable devices, implementation of graded intervention on vibration parameters based on risk level, and simultaneous push of corresponding health advice; S4: Adjust vibration parameters based on the characteristics of simple vibration equipment in different scenarios, and generate fragmented training schemes by combining user daily life data.

2. The method for personalized whole-body vibration health index regulation based on artificial intelligence according to claim 1, characterized in that, The specific steps of S1 are as follows: S1.1: Integrate user questionnaire information, basic physiological parameters, disease subtype subcategories and real-time environmental variables to construct a comprehensive initial feature input set; S1.2: Standardize, encode, and fuse heterogeneous multidimensional data to generate structured, computable user health status vectors; S1.3: Utilize AI models to learn the mapping relationship between multidimensional features in historical user data and the effects of vibration intervention; S1.4: Based on the current user's fusion characteristics and the trained AI model, output a personalized initial vibration intervention plan containing parameters such as frequency, amplitude, waveform, and duration.

3. The method for personalized whole-body vibration health index regulation based on artificial intelligence according to claim 2, characterized in that, The disease subtype subclass data includes primary / secondary hypertension and insulin-resistant / insufficient hyperglycemia.

4. The method for personalized whole-body vibration health index regulation based on artificial intelligence according to claim 1, characterized in that, The specific steps of S2 are as follows: S2.1: Based on multidimensional health and behavioral characteristics, users are divided into fine-grained subgroups with similar response patterns using unsupervised or semi-supervised learning methods; S2.2: Build and train an independent reinforcement learning sub-model for each sub-group to learn the optimal vibration intervention strategy for that sub-group; S2.3: Using the health indicator regulation effect monitored by S3 as a reward signal, the individual interaction data is driven to update the central model synchronously under the reinforcement learning framework and cross-validate with the data of users of the same subtype. The cloud extracts the gold standard intervention pattern and efficient parameter weights by aggregating the data of the same subtype, and sends the validated high-weight parameters and optimized model to the terminals of users of the same subtype. S2.4: Based on a two-layer model of continuous evolution of subtype commonalities and individual characteristics, it dynamically outputs more accurate and adaptive personalized vibration intervention solutions.

5. The method for personalized whole-body vibration health index regulation based on artificial intelligence according to claim 4, characterized in that, In step S2.2, an independent reinforcement learning sub-model is constructed and trained for each sub-group to learn the optimal vibration intervention strategy for that sub-group. The specific steps are as follows: S2.2.1: Define the state space The user's multidimensional health feature vector at time t includes blood pressure, blood sugar, uric acid, and recent training compliance; S2.2.2: Define the action space Optional combinations of vibration intervention parameters, including frequency, amplitude, and waveform type; S2.2.3: Calculate the reward value based on the real-time feedback of health indicators from S3 and the group coordination factor; the group coordination factor is used to measure the consistency between the individual response strategy and the cloud-based subtype model prediction strategy, guiding the individual plan to quickly converge to the subtype's optimal intervention range. The specific formula is as follows: ； in, Let be the instantaneous reward value at time t. As a weighting coefficient for the effect of blood pressure regulation, Let be the change in blood pressure at time t. This is the weighting coefficient for blood glucose stability. Let be an indicator of blood glucose stability at time t. This represents the penalty coefficient for abnormal events. For abnormal event indicator variables, For group collaboration weighting coefficient, As a group synergy factor; S2.2.4: Employ a deep Q-network to maximize cumulative discount rewards. To achieve the goal, iteratively update the policy network parameters of this subtype sub-model until the model converges to a preset accuracy, where, is the discount factor, and T is the total training cycle duration.

6. The method for personalized whole-body vibration health index regulation based on artificial intelligence according to claim 4, characterized in that, In S2.3, the health indicator regulation effect monitored in S3 is used as a reward signal to drive individual interaction data to synchronously update the central model under the reinforcement learning framework, and cross-validate it with user data of the same subtype. The cloud extracts the gold standard intervention pattern and efficient parameter weights by aggregating data of the same subtype, and distributes the validated high-weight parameters and optimized model to the terminals of users of the same subtype. The specific operation steps are as follows: S2.3.1: Obtain the regulatory effect of the health indicators monitored by S3, quantify it into a reward signal for reinforcement learning, and calculate the reward value using the following formula: ； in, Let i be the reward value for the i-th user. For changes in health indicators, For the stability of health indicators, This represents the abnormal risk coefficient. , , These are the weighting coefficients; S2.3.2: The cloud uses privacy-preserving data aggregation technology to input individual users' interaction data and corresponding reward signals into the reinforcement learning framework, and synchronously update the policy parameters of the central model; S2.3.3: Extract historical interaction data and reward signals from users of the same subtype, and perform cross-validation on the updated central model to verify the model's generalization ability in this subtype group; S2.3.4: If the cross-validation accuracy does not reach the preset threshold, the central model parameters are iteratively adjusted until the model meets the generalization requirements.

7. The method for personalized whole-body vibration health index regulation based on artificial intelligence according to claim 1, characterized in that, The specific steps of S3 are as follows: S3.1: Continuously acquire dynamic data streams of key health indicators such as blood pressure, blood sugar, heart rate, and uric acid during the user's vibration training process through wearable devices; S3.2: Utilize AI models to analyze real-time monitoring data and automatically detect risk events that exceed safety thresholds or exhibit abnormal fluctuations; S3.3: Based on the risk identification results, implement graded intervention: for mild abnormalities, fine-tune vibration parameters; for severe abnormalities, immediately trigger training shutdown. S3.4: Synchronously generate and push non-diagnostic matching suggestions that match the current physiological state.

8. The method for personalized whole-body vibration health index regulation based on artificial intelligence according to claim 7, characterized in that, In step S3.2, an AI model is used to analyze real-time monitoring data and automatically detect risk events that exceed safety thresholds or exhibit abnormal fluctuations. The specific operation steps are as follows: S3.2.1: Preprocess the real-time health indicator data collected in S3.1 to remove noise and fill in missing values; S3.2.2: Compare the preprocessed data with the preset safety threshold to detect whether there are any risk events where the indicators exceed the threshold; S3.2.3: Calculate the fluctuation range of health indicators within a preset time window, and identify abnormal risks through abnormal fluctuation detection; S3.2.4: Integrate the detection results of threshold exceeding and abnormal fluctuations, and output the risk event type and trigger level.

9. A method for regulating personalized whole-body vibration health indicators based on artificial intelligence according to claim 8, characterized in that, S3.2.3 identifies abnormal risks through abnormal fluctuation detection, and the specific formula is as follows: ； in, Let be the health indicator value at time t. This represents the average of the indicators over the previous w time windows. The standard deviation of the index for the first w time windows. This is the fluctuation sensitivity coefficient.

10. The method for regulating personalized whole-body vibration health indicators based on artificial intelligence according to claim 1, characterized in that, The specific steps of S4 are as follows: S4.1: Based on the physical characteristics of the stiffness and motor torque of simple vibration equipment used in home, office and other scenarios, the energy value output by the AI ​​model is automatically converted into the gear parameters that the equipment can execute, and the vibration intervention parameters are automatically mapped and adjusted. S4.2: Based on users' daily routine data, intelligently plan personalized vibration training programs with short durations and high frequencies.

Images