Artificial intelligence health management system based on multi-source data fusion

Abstract

The invention discloses an artificial intelligence health management system based on multi-source data fusion, and relates to the field of intelligent health management, and the system comprises a multi-modal data collection and processing module which is used for obtaining related data; the target dynamic adjustment module is used for decomposing the health target of the user and adjusting exercise and diet schemes; the health file construction and storage module is used for generating a dynamic health trend graph and realizing data encryption storage and cross-device synchronization; the panoramic health analysis module is used for associating the multi-dimensional health data to generate risk tags; the personalized health intervention module is used for calling a food red-black list to generate customized courses and recipes; and the intelligent reminding and execution module is used for triggering reminding and executing a priority strategy. Risk assessment is realized through multi-modal data fusion and block chain encryption, and safe, reliable and full-scene coverage intelligent health management service is provided for a user depending on a personalized rule base and an offline early warning mechanism.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent health management, more particularly to an artificial intelligence health management system based on multi-source data fusion. BACKGROUND

[0002] The field of artificial intelligence health management is rapidly developing, aiming to integrate individual physiological, behavioral and environmental data through intelligent devices and algorithms to provide proactive health monitoring and intervention services for families and individuals. This technology is widely used in chronic disease prevention, sub-health improvement and personalized health promotion scenarios.

[0003] The existing technology has significant limitations in data integration depth. Traditional systems rely on a single or small number of data sources, lacking synchronous collection and deep fusion analysis of multi-modal data such as body composition, real-time heart rate, sleep quality, detailed diet, etc. This leads to one-sided health assessment, making it difficult to build accurate and dynamic health trend maps and identify early risks. Meanwhile, insufficient data cleaning and standardization further reduces the reliability of analysis, making it impossible to provide high-quality input for subsequent decision-making.

[0004] Existing systems are severely lacking in intervention intelligence. Their goal setting is rigid, lacking the ability to dynamically adjust plans based on machine learning algorithms comparing actual progress with expected values, and unable to automatically optimize exercise and diet plans based on real-time user data, resulting in low goal achievement rates. Intervention plans are mostly generic templates that fail to correlate multi-dimensional health data to generate risk labels and drive personalized strategies, and fail to generate customized courses and recipes based on personalized indicators such as body fat percentage and muscle mass, further reducing user compliance and management effectiveness.

[0005] The existing technology has gaps in scene adaptation, data security and regional value mining. Systems provide single-dimensional reminders, lack intelligent reminders based on time rules, physiological indicator thresholds and risk levels, and cannot guarantee offline scene usability. Most importantly, they fail to effectively match regional specialty food libraries with user geographic locations, limiting the value promotion and precise reach of high-quality health products. Meanwhile, data security is at risk, lacking reliable solutions for encrypted storage and cross-device synchronization through blockchain algorithms, hindering the establishment of multi-role collaboration trust and sensitive data flow.

[0006] Therefore, how to design an artificial intelligence health management system based on multi-source data fusion that can deeply integrate multi-source data, achieve dynamic closed-loop intervention, ensure the safety and availability of full-scene services, and effectively connect regional resources and cultural health needs is a problem that needs to be solved by those skilled in the art. SUMMARY

[0007] In view of this, the present application provides a multi-source data fusion-based artificial intelligence health management system, which dynamically generates and optimizes personalized health goals and intervention strategies by fusing and processing multi-modal health data through artificial intelligence technology, and provides precise, real-time, reliable and full-scene health monitoring, risk assessment and management services for users on the premise of ensuring user data privacy and security.

[0008] In order to achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0009] A multi-source data fusion-based artificial intelligence health management system comprises the following modules:

[0010] A multi-modal data acquisition and processing module is used to acquire user physiological data, motion data, diet records and physical examination reports, analyze emergency physiological indicators in real time on the device side and trigger local reminders, and perform data cleaning and standardization;

[0011] A target dynamic adjustment module is used to decompose user health goals and dynamically adjust exercise and diet programs based on machine learning algorithms compared with actual progress and expected values;

[0012] A health record construction and storage module is used to generate dynamic health trend maps, label abnormal risk points, and realize data encryption storage and cross-device synchronization through a blockchain algorithm;

[0013] A panoramic health analysis module is used to associate multi-dimensional health data to generate risk labels and drive personalized intervention strategies;

[0014] A personalized health intervention module is used to call food red and black lists to generate customized courses and recipes, match regional food material libraries according to user geographic locations, and load special rules for special groups;

[0015] An intelligent reminding and execution module is used to trigger reminders according to time rules and physiological index thresholds, execute priority strategies, and ensure offline scene availability.

[0016] Preferably, the multi-modal data acquisition and processing module comprises:

[0017] A wearable device connection unit is used to compatible multi-brand devices to collect physiological data, support local real-time analysis of heart rate abnormalities and trigger vibration reminders;

[0018] A multi-source input unit is used to receive manual basic data, voice and image diet records and physical examination report OCR identification;

[0019] A data cleaning unit is used to fill in missing values and perform normalization processing, and output structured data to the health record construction and storage module.

[0020] Preferably, the target dynamic adjustment module comprises:

[0021] Target decomposition unit: for disassembling the total target into stage targets;

[0022] Dynamic optimization unit: based on the sleep quality data and historical behavior data of the health record construction and storage module, generate exercise and diet adjustment strategies;

[0023] Progress feedback unit: output the adjustment scheme to the panoramic health analysis module.

[0024] Preferably, the health record construction and storage module comprises:

[0025] Index calculation unit: based on the body composition data of the multi-modal data acquisition and processing module, calculate BMI and body fat rate;

[0026] Trend map unit: for constructing dynamic health trend map;

[0027] Blockchain storage unit: receive the risk label data of the panoramic health analysis module, and realize encrypted storage through smart contract.

[0028] Preferably, the panoramic health analysis module comprises:

[0029] Data correlation unit: for spatiotemporal alignment of the exercise and diet scheme output by the target dynamic adjustment module, the BMI and body fat rate index of the health record construction and storage module, and the sleep data of the multi-modal data acquisition and processing module;

[0030] Dynamic risk assessment unit: for generating health risk level labels according to the aligned multi-source data;

[0031] Intervention driving unit: for outputting the risk label to the personalized health intervention module to generate a forbidden list, and synchronizing to the intelligent reminding and execution module to trigger an early warning.

[0032] Preferably, the personalized health intervention module comprises:

[0033] Recipe generation unit: based on the risk label and real-time index output by the panoramic health analysis module, call the food red and black list to generate a forbidden list;

[0034] Regional adaptation unit: for matching user positioning and local characteristic food material library;

[0035] Population customization unit: according to the pregnant woman and old person labels marked by the health record construction and storage module, load the exclusive rule library.

[0036] Preferably, the intelligent reminding and execution module comprises:

[0037] Rule engine unit: receives risk label data of panoramic health analysis module, matches time rules with physiological index threshold values;

[0038] Priority execution unit: used for sorting multiple reminder conflicts according to health risk levels;

[0039] Offline guarantee unit: links local analysis capabilities of multi-modal data acquisition and processing modules to guarantee triggering of a timing reminder when disconnected from a network.

[0040] According to the technical solution described above, compared with the prior art, the technical solution of the present application has the following advantages

[0041] Advantages:

[0042] 1. The system forms a real-time closed loop of "monitoring-analysis-intervention-feedback" through real-time multi-modal data acquisition and local analysis, target dynamic adjustment driven by machine learning, and intelligent reminder execution, significantly improving the response speed and adaptability of health management, and ensuring that the scheme is continuously optimized according to the user's state.

[0043] 2. Multi-dimensional data correlation analysis is used (dynamic risk labels are generated, breaking through the limitations of a single index; at the same time, through blockchain encryption storage and cross-device synchronization, medical-grade data can be reliably transferred under the premise of protecting user privacy, providing a reliable basis for precise intervention.

[0044] 3. Based on risk labels and user characteristics, a food red and black list is generated to generate a taboo list, and a local food library and a special rule library are matched, so that the health scheme has both scientificity and scene adaptability; offline reminder guarantee expands the applicable scenarios of the system, ensuring that high-risk users can obtain timely warnings in any environment. BRIEF DESCRIPTION OF DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description only illustrate the embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of the provided drawings.

[0046] Figure 1 A multi-source data fusion-based artificial intelligence health management system structure framework diagram is provided for the embodiments of the present application;

[0047] Figure 2 An interactive interface schematic diagram of a target decomposition unit is provided for the embodiments of the present application;

[0048] Figure 3 A nutritional classification architecture interface schematic diagram of a recipe generation unit is provided for the embodiments of the present application;

[0049] Figure 4 The card priority configuration interface of the intelligent reminding and executing module provided by the embodiment of the present application is shown. DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present application will be clearly and completely described with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0051] As shown in the figure, the embodiment provides an artificial intelligence health management system based on multi-source data fusion, which comprises the following modules: Figure 1

[0052] A multi-modal data acquisition and processing module is used to acquire physiological data, motion data, diet records and physical examination reports of a user, analyze emergency physiological indexes in real time at a device end and trigger local reminding, and perform data cleaning and standardization;

[0053] A target dynamic adjustment module is used to decompose a health target of a user and dynamically adjust a motion and diet scheme based on a machine learning algorithm compared with an actual progress and an expected value;

[0054] A health record construction and storage module is used to generate a dynamic health trend graph, mark abnormal risk points, and realize data encryption storage and cross-device synchronization through a blockchain algorithm;

[0055] A panoramic health analysis module is used to associate multi-dimensional health data to generate a risk label and drive a personalized intervention strategy;

[0056] A personalized health intervention module is used to call a food red and black list to generate a customized course and a recipe, match a regional food material library combined with a geographical position of a user, and load exclusive rules for special groups;

[0057] An intelligent reminding and executing module is used to trigger reminding according to a time rule and a physiological index threshold, execute a priority strategy, and guarantee offline scene availability.

[0058] The artificial intelligence health management system realizes efficient and accurate, dynamic closed-loop, safe and reliable and highly personalized health management by real-time acquisition, local analysis and cleaning of multi-source data, dynamic adjustment of a target scheme combined with machine learning, generation of accurate risk labels and driving of personalized intervention through multi-dimensional correlation analysis, guarantee of data security and cross-device synchronization relying on a blockchain, and ensuring of reminding availability in an offline scene.

[0059] ​The following further details each module in the above system:

[0060] In this embodiment, the multi-modal data acquisition and processing module is used to obtain user physiological data, motion data, diet records, and physical examination reports, analyze emergency physiological indicators in real time on the device side and trigger local reminders, and perform data cleaning and standardization. It includes:

[0061] Wearable device connection unit: used to compatible multi-brand devices to collect physiological data, support local real-time analysis of heart rate abnormalities and trigger vibration reminders; can dynamically adapt to mainstream brand devices such as Huawei, Apple, etc. through Bluetooth 5.0 protocol, call the original SDK of the device to obtain raw sensor data (such as heart rate, blood oxygen), and deploy a lightweight heart rate abnormality detection model on the device side. When the deviation of 5 consecutive sampling points from the user's resting heart rate baseline is ±20%, the wearable device vibration reminder is triggered directly, realizing real-time early warning in a network-free environment.

[0062] Multi-source input unit: used to receive manual basic data, voice and image diet records, and physical examination report OCR identification; support voice input diet records, image recognition food types, and OCR key indicator extraction (such as blood glucose, cholesterol) on physical examination report PDF, image, and structured storage to downstream modules.

[0063] Data cleaning unit: used to fill in missing values and normalize, output structured data to health record construction and storage module; uses the historical mean value filling method to handle missing values, and unifies multi-source data to the [0, 1] interval through Min-Max standardization, and outputs structured data table for health record module to call.

[0064] This module realizes high-precision acquisition and local real-time processing of multi-source heterogeneous health data, breaks through the limitations of traditional systems relying on a single data source, and provides a low-latency, high-reliability data foundation for dynamic health analysis.

[0065] In this embodiment, the target dynamic adjustment module is used to decompose user health targets and dynamically adjust exercise and diet plans based on machine learning algorithms comparing actual progress with expected values.

[0066] For example, Figure 2As shown, the target decomposition unit guides the user to set personalized health goals through an interactive interface. The user sets the target weight (e.g., 55 kg) by sliding a selector or manually inputting the exact value. From the calendar, the user selects the target achievement date, and the system automatically calculates the total period (48 weeks) and recommends a reasonable weight loss rate (0.3 kg / week) to avoid the health risks of aggressive programs. The system also provides an emotional option (e.g., "happy about better body shape") to associate health goals with psychological motivation and enhance long-term user compliance. This not only reduces the threshold for goal setting but also provides key input parameters for the subsequent dynamic optimization unit, achieving scientific goal decomposition and progress tracking.

[0067] Dynamic optimization unit: Based on the sleep quality data and historical behavior data of the health record construction and storage module, generate exercise and diet adjustment strategies; based on sleep quality data (deep sleep ratio <15% to reduce exercise intensity) and historical behavior deviation (continuous 3 days not meeting the standard), use PID algorithm to dynamically adjust exercise duration and diet calorie distribution to ensure target accessibility.

[0068] Progress feedback unit: output the adjustment scheme to the panoramic health analysis module.

[0069] This module solves the problem of traditional health management system target rigidity through a machine learning driven dynamic closed-loop optimization mechanism, significantly improving user compliance and target achievement rate.

[0070] In this embodiment, the health record construction and storage module is used to generate a dynamic health trend map, mark abnormal risk points, and realize data encryption storage and cross-device synchronization through a blockchain algorithm; including:

[0071] Index calculation unit: based on the body composition data of the multi-modal data acquisition and processing module, calculate BMI and body fat rate; based on body composition data, calculate BMI and body fat rate, and automatically update health baseline when continuous 7-day index fluctuation >5%.

[0072] Trend map unit: used to construct a dynamic health trend map; aggregate BMI, body fat rate and other data to generate a scalable timeline map, support click to view details; automatically mark metabolic risk red area when body fat rate is continuously above threshold value (male >25%, female >30%) for 3 days.

[0073] Blockchain storage unit: receives risk label data from the panoramic health analysis module, and realizes encryption storage through smart contract. Encrypt sensitive data (such as risk label) through asymmetric key encryption, store encrypted fingerprint using IPFS, and verify blockchain signature when synchronizing across devices to ensure data cannot be tampered with.

[0074] It solves the problems of privacy leakage and data island by constructing a dynamic visual personal health portrait and relying on blockchain technology to achieve secure storage and cross-device trusted synchronization of medical-grade data.

[0075] The panoramic health analysis module in this embodiment is used to associate multi-dimensional health data to generate risk labels and drive personalized intervention strategies; including:

[0076] The data association unit is used to dynamically adjust the movement and diet plan output by the target module, the BMI and body fat rate index of the health record construction and storage module, and the sleep data of the multi-modal data acquisition and processing module; to align the movement GPS track, diet timestamp, and sleep stage data on a 24-hour time axis (such as binding the morning run and breakfast record in space and time), and construct a multi-dimensional correlation matrix.

[0077] The dynamic risk assessment unit is used to generate health risk level labels based on the aligned multi-source data; after inputting the aligned data, it outputs 0-5 level risk labels according to the preset medical rule tree.

[0078] The intervention driving unit is used to output the risk labels to the personalized health intervention module to generate a contraindication list, and synchronously to the intelligent reminder and execution module to trigger an early warning.

[0079] This module generates precise dynamic risk labels through multi-dimensional spatio-temporal correlation analysis, breaking through the one-sidedness of traditional health assessment, and providing a scientific basis for personalized intervention.

[0080] The personalized health intervention module in this embodiment is used to call the food red and black list to generate customized courses and recipes, match regional food material libraries based on the user's geographic location, and load exclusive rules for special groups; including:

[0081] The recipe generation unit calls the food red and black list to generate a contraindication list based on the risk labels and real-time indicators output by the panoramic health analysis module;

[0082] As shown in Figure 3 The recipe generation unit realizes precise food recommendation through a multi-dimensional classification tree, including health goal orientation (first-level classification): hierarchical display according to user core needs, among which weight loss focuses on calorie control, muscle gain associates with protein supplementation, and diabetes emphasizes low GI value food materials; medical scene refinement (second-level classification): expanding comorbidity scenarios such as hypertension and hyperlipidemia under diabetes management, and setting up independent labels for special groups to ensure strong correlation between intervention plans and health risks; scientific food material screening (third-level classification): based on nutrition standards to build vertical categories, automatically filter contraindicated foods through the red and black list mechanism, and preferentially recommend local characteristic healthy food materials through the regional adaptation unit.

[0083] The design converts complex nutrition rules into visual interactive paths, providing implementation support for generating a taboo list for calling food red and black lists.

[0084] Regional adaptation unit: for matching user positioning with local characteristic food material library; associating user GPS coordinates with local agricultural product library (such as recommending selenium-rich millet in a certain place), and combining with festival to push cultural food therapy solutions (such as angelica and mutton soup on the winter solstice).

[0085] Population customization unit: according to the health record construction and storage module, mark pregnant women and the elderly, and load the exclusive rule library.

[0086] This module realizes the intervention scheme of regional culture adaptation and population precise customization, significantly improves the practicality of health management and the long-term compliance of users.

[0087] In this embodiment, the intelligent reminding and execution module is used to trigger reminders according to time rules and physiological index threshold values, execute priority strategies, and ensure offline scenario availability; including:

[0088] Rule engine unit: receiving risk label data from the panoramic health analysis module, matching time rules and physiological index threshold values; specific hierarchical definition of trigger conditions (L1: sitting > 1h→ pop-up; L3: abnormal heart rate + sitting→ vibration + voice), supporting doctors to customize physiological threshold values (such as postprandial blood glucose > 10mmol / L for diabetes).

[0089] Priority execution unit: for sorting multiple reminders in conflict according to health risk levels; sorting conflict reminders according to risk levels, and executing according to trigger time in the same level.

[0090] Offline guarantee unit: linking the local analysis capability of the multi-modal data acquisition and processing module to guarantee the triggering of timed reminders when disconnected from the network; caching the rule engine and sensor analysis model on the device side, triggering reminders based on local data when disconnected from the network (such as accelerometer to determine sitting), and automatically synchronizing logs after connecting to the network.

[0091] As shown in Figure 4 The intelligent reminding and execution module realizes scenario-based health management of users through modular card design, supports hierarchical configuration of functions and dynamic priority management, the core monitoring items default to display high-frequency indicators, the scene extension items are enabled as needed to meet the customization needs of special populations, the card can be flexibly pinned to deal with high-risk situations, the hidden card runs in the background to ensure data continuity, combined with offline adaptation design to ensure that key functions can still be recorded and executed in a network environment, and efficient cooperation of user flexible configuration and system intelligent scheduling is realized.

[0092] The working process of the system in this embodiment is further described in detail as follows.

[0093] 1) Daily health management closed loop;

[0094] When the user goes for a morning run, the multi-modal data acquisition module monitors heart rate in real time through the Huawei bracelet. When it detects that the resting heart rate has suddenly risen to 120 beats per minute, it immediately triggers a device-side vibration warning. After breakfast, the user takes a photo of the oatmeal and uploads it. The multi-source input unit identifies it as a low GI food and records 200kcal.

[0095] During the morning work, the target dynamic adjustment module detects that yesterday's exercise consumption did not meet the standard (only 80% completed), combined with sleep data, the dynamic optimization unit automatically reduces the exercise intensity by 10 minutes today, and pushes the adjustment suggestion to the phone pop-up window through the progress feedback unit. At the same time, the health record module updates the BMI trend graph, and finds that the body fat rate has been >25% for 3 consecutive days, marking the red metabolic risk area in the chart.

[0096] Before lunch, the panoramic health analysis module associates exercise deficiency, metabolic risk, and high-sodium diet records to generate an L3-level cardiovascular risk label, driving the personalized intervention module to generate a contraindication list: disable lunch pickles, and recommend selenium-rich millet porridge.

[0097] 2) High-risk warning and offline response;

[0098] During the user's lunch break, the intelligent reminder module detects that the user has been sitting for 2 hours, triggering an L1-level pop-up window to remind the user to stand up. Then the wearable device detects that the resting heart rate is continuously >100 beats per minute, combined with the pre-stored pregnant woman risk rules of the panoramic analysis module (L4-level heart rate abnormalities + sitting), the priority execution unit immediately triggers a vibration + voice alarm: "Detecting heart rate abnormalities, please rest immediately!"

[0099] At the same time, the personalized intervention module loads the pregnant woman's exclusive rule library, pushes the folic acid fortified recipe (spinach and liver soup), and disables caffeine drinks (the recipe generation unit calls the red and black list).

[0100] When the user enters a subway environment without network, the pre-synchronized encrypted health data in the blockchain storage unit still supports local calling; the sitting reminder completely relies on the device-side cached rule engine, and triggers a vibration reminder on time. All offline operation logs are automatically encrypted and returned to the cloud after network recovery.

[0101] The artificial intelligence health management system in this embodiment realizes the deep fusion of real-time collection of multi-modal data, dynamic closed-loop optimization and blockchain encryption storage, constructs a whole-process health management framework of "perception-analysis-intervention-feedback", not only realizes medical-grade risk assessment and precise adaptation of personalized health strategies, but also relies on the offline available localized intelligence and spatio-temporal correlation analysis capability, breaks through the fragmentation and lag of traditional health management, provides an end-to-end trusted solution for personal health portrait construction, risk early warning and scientific intervention, and effectively promotes the continuous evolution of health management services towards intelligentization, scene and security and reliability.

[0102] The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments. For the system disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and the relevant parts are described in the method part.

[0103] The above description of the disclosed embodiments enables a person skilled in the art to implement or use the present application. Various modifications to the embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An artificial intelligence health management system based on multi-source data fusion, characterized in that: Includes the following modules: Multimodal data acquisition and processing module: used to obtain user physiological data, exercise data, diet records and physical examination reports, analyze emergency physiological indicators in real time on the device side, trigger local reminders, and perform data cleaning and standardization; Dynamic goal adjustment module: used to break down user health goals and compare actual progress with expected values ​​based on machine learning algorithms, dynamically adjusting exercise and diet plans; Health record construction and storage module: used to generate dynamic health trend maps, mark abnormal risk points, and realize data encryption storage and cross-device synchronization through blockchain algorithms; Panoramic health analysis module: used to associate multi-dimensional health data to generate risk labels and drive personalized intervention strategies; Personalized health intervention module: used to generate customized courses and recipes based on the food red and black lists, match the regional food library based on the user's geographic location, and load exclusive rules for special groups; Intelligent reminder and execution module: used to trigger reminders based on time rules and physiological indicator thresholds, execute priority strategies, and ensure offline scenario availability.

2. The artificial intelligence health management system based on multi-source data fusion according to claim 1 is characterized in that: The multimodal data acquisition and processing module includes: Wearable device connection unit: used to collect physiological data from devices of multiple brands, support local real-time analysis of heart rate abnormalities and trigger vibration reminders; Multi-source input unit: used to receive manual basic data, voice and image diet records and physical examination report OCR recognition; Data cleaning unit: used to fill missing values ​​and normalize them, and output structured data to the health record construction and storage module.

3. The artificial intelligence health management system based on multi-source data fusion according to claim 1 is characterized in that: The target dynamic adjustment module includes: Target decomposition unit: used to break down the overall target into stage targets; Dynamic Optimization Unit: Generates exercise and diet adjustment strategies based on sleep quality data and historical behavior data from the health record construction and storage module; Progress feedback unit: outputs the adjustment plan to the panoramic health analysis module.

4. The artificial intelligence health management system based on multi-source data fusion according to claim 1 is characterized in that: The health record construction and storage module includes: Index calculation unit: Calculates BMI and body fat percentage based on body composition data from the multimodal data acquisition and processing module; Trend graph unit: used to construct dynamic health trend graphs; Blockchain storage unit: Receives risk tag data from the panoramic health analysis module and implements encrypted storage through smart contracts.

5. The artificial intelligence health management system based on multi-source data fusion according to claim 1 is characterized in that: The panoramic health analysis module includes: Data association unit: used for the exercise and diet plans output by the spatiotemporal alignment target dynamic adjustment module, the BMI and body fat percentage indicators of the health record construction and storage module, and the sleep data of the multimodal data acquisition and processing module; Dynamic risk assessment unit: used to generate health risk level labels based on aligned multi-source data; Intervention drive unit: used to output risk tags to the personalized health intervention module to generate a list of contraindications, and synchronize them to the intelligent reminder and execution module to trigger early warnings.

6. The artificial intelligence health management system based on multi-source data fusion according to claim 1 is characterized in that: The personalized health intervention module includes: Recipe generation unit: Based on the risk labels and real-time indicators output by the panoramic health analysis module, it calls the food red and black lists to generate a taboo list; Regional adaptation unit: used to match user location with local specialty food library; Population customization unit: Based on the health records, the module builds and stores labels for pregnant women and the elderly, and loads a dedicated rule library.

7. The artificial intelligence health management system based on multi-source data fusion according to claim 1 is characterized in that: The intelligent reminder and execution module includes: Rule engine unit: receives risk tag data from the panoramic health analysis module and matches time rules with physiological indicator thresholds; Priority execution unit: used to sort multiple reminder conflicts by health risk level; Offline support unit: Links the local analysis capabilities of the multimodal data acquisition and processing modules to ensure that scheduled reminders are triggered when the network is disconnected.

Images