A personalized health behavior quantification incentive system and method based on blood pressure feedback

Abstract

The application discloses a personalized health behavior incentive system and method based on blood pressure feedback, multi-dimensional data is collected through a PPG+ECG dual-mode sensor, an XGBoost model is combined to establish a behavior-blood pressure correlation mechanism, personalized incentive and model adaptive updating are realized, user health management compliance and blood pressure control effect are effectively improved, and an intelligent solution is provided for health management of a chronic disease population such as hypertension.

Description

Technical Field

[0001] This invention relates to the fields of health management and digital healthcare, specifically to an incentive mechanism based on physiological parameter feedback, and more particularly to a quantitative incentive system, method, and device for health behaviors based on blood pressure feedback. This technology integrates wearable device data with machine learning algorithms to establish a dynamic correlation model between health behaviors and blood pressure changes, providing personalized incentive feedback for patients with chronic diseases such as hypertension. It belongs to the interdisciplinary field of digital health and artificial intelligence. Background Technology

[0002] 1. Current Status of Technological Development With the integration of IoT and AI technologies, the health management field has transitioned from single-data monitoring to multi-dimensional behavioral intervention. The global wearable device market is projected to exceed $80 billion by 2025, with blood pressure monitoring devices accounting for 27%. However, existing technologies generally suffer from the "data silo" problem: 78% of users' health apps can only record basic indicators such as steps and heart rate, failing to quantify the specific impact of behavior on blood pressure; only 12% of chronic disease management solutions provide dynamic incentive feedback, and the incentive rules are mostly generic designs that do not consider individual physiological differences.

[0003] 2. Analysis of Existing Technical Solutions

[0004] - Patent CN202411449179.9: Achieving adaptive adjustment of LED backlight through the fusion analysis of environmental sensor data and user vital sign data. This technology uses an LSTM network to establish an environment-user state correlation model, but it is only applicable to lighting control scenarios and does not involve quantitative incentives for health behaviors.

[0005] - Patent CN202410584179.3: Utilizes multimodal biosensors (PPG, ECG) to collect data and optimizes blood pressure monitoring accuracy through support vector regression (SVR). Its core lies in sensor error compensation, but it does not establish a behavior-blood pressure causal relationship model.

[0006] - Patent CN202411038558.9: Constructing a time-series analysis model for physiotherapy data and employing the Dynamic Time Warping (DTW) algorithm to evaluate the physiotherapy effect. This technology is limited to the physiotherapy scenario, and the incentive feedback mechanism is only for adjusting system parameters, lacking user-end visual incentives.

[0007] 3. Deficiencies in existing technology

[0008] The existing technology has three major shortcomings: 1. Lack of correlation model: 83% of health management tools have not established a quantitative mapping relationship between behavior and blood pressure changes, and users cannot perceive the specific impact of "30 minutes of jogging" on systolic blood pressure.

[0009] 2. Lack of personalized incentives: Generalized reward rules (such as the daily 10,000 steps badge) are out of touch with users' physiological responses, causing 62% of users to abandon the service within 3 months.

[0010] 3. Weak model adaptability: Only 17% of the systems support online model updates, which cannot adapt to changes in individual physiological states (such as seasonal blood pressure fluctuations).

[0011] 4. The technical problem solved by the present invention

[0012] To address the aforementioned shortcomings, this invention proposes a health behavior quantitative incentive system based on blood pressure feedback, focusing on solving the following problems: - How to construct a personalized dynamic correlation model between "behavior and blood pressure" - How to design a dynamic incentive mechanism based on positive improvement of physiological parameters - How to achieve cross-user aggregation optimization and adaptive updating of the model

[0013] Summary of the Invention 1. Overview of Technical Solution This invention provides a quantitative incentive system for health behaviors based on blood pressure feedback. Figure 1 ),include: - Blood pressure monitoring module: Employs a PPG+ECG dual-mode sensor to achieve non-invasive blood pressure monitoring with an accuracy of ±2mmHg and a sampling frequency of 1Hz. - Behavioral data acquisition module: Integrates a 9-axis IMU, bioelectrical impedance module, and microphone array to simultaneously acquire 12-dimensional behavioral data, including motion (acceleration, angular velocity), diet (chewing audio analysis), and sleep (body movement, heart rate variability). - Data processing and incentive calculation engine: includes correlation analysis unit ( Figure 2 The system includes a quantitative calculation unit and a personalized mapping model constructed using the XGBoost algorithm. The input is a vector of behavioral factors (type, intensity, time), and the output is the predicted value of blood pressure change. - User incentive feedback module: Displays a 3D visualization interface via AR glasses or mobile app ( Figure 3 The score includes expected impact value (0-100 points), blood pressure trend curve, and behavioral target achievement rate. When the accumulated points reach a threshold, a health mall coupon reward is triggered. 2. Innovative Technological Features

[0014] Compared with existing technologies, this invention has three major innovations: 1. Multimodal behavior quantification technology: - Exercise intensity: MET (metabolic equivalent) value calculated from IMU data, with an error of <5%. - Dietary Calories: A CNN audio classification model is used to identify chewing sounds, and calorie intake is calculated by combining this with a food database. - Stress index: Calculated based on the LF / HF ratio of HRV (heart rate variability), with a sampling window of 300 seconds. 2. Dynamic Personalized Modeling Method: - Initial model training: Collect 7 days of basic user data to construct an XGBoost regression model of behavioral factor vectors and blood pressure changes. - Online update mechanism: The model is automatically fine-tuned at 3:00 AM every day, using the FTRL algorithm to adapt to changes in physiological state. - Cross-user optimization: Cloud servers aggregate data from 100,000 users and optimize general model parameters through federated learning. 3. Multi-dimensional incentive and feedback system: - Instant feedback: The expected impact value is displayed within 3 seconds of the action occurring. - Long-term incentives: Set up weekly / monthly points leaderboards, with the top 10% of users receiving physical medals. - Social Incentives: Support team challenges with friends; team points can be redeemed for health check-up packages. 3. Beneficial technical effects

[0015] This invention achieves three major technological advancements: 1. Improved quantification accuracy: The behavior-blood pressure association model achieved an R² of 0.87, a 42% improvement over traditional linear models. 2. Enhanced user adherence: The 3-month continuous usage rate increased to 81%, 3.3 times higher than the general incentive program. 3. Improved management effectiveness: Users' average systolic blood pressure decreased by 12.3 mmHg, and the compliance rate increased by 67%. Attached Figure Description

[0016] Figure 1 This is a block diagram of the health behavior quantification incentive system based on blood pressure feedback according to the present invention. Figure 2 A detailed schematic diagram of the data processing and stimulus computing engine. Figure 3 An AR visualization interface diagram for the user incentive feedback module. Figure 4 Flowchart of a method for quantifying incentives for health behaviors based on blood pressure feedback Figure 5 A schematic diagram of the hardware architecture of an electronic device. Detailed Implementation Example 1: Complete System Implementation Plan

[0017] 1. System Architecture Deployment - Hardware layer: The device uses the Huawei Watch GT 3 Pro as the terminal device, integrating the Maxim MAX30102 PPG sensor and the TI ADS1299 ECG chip. - Network Layer: Data is uploaded to the Huawei Cloud IoT platform via the BLE 5.2 protocol, with an end-to-end latency of <200ms. - Application Layer: Deployed on Huawei Cloud ECS servers, using a Docker containerized architecture, supporting 100,000 concurrent accesses.

[0018] - Data preprocessing: - Blood pressure data: Median filtering was used to remove motion artifacts, window size 5. - Behavioral data: Motion data was corrected using a Kalman filter, and dietary audio was extracted in the 1kHz-4kHz frequency band. - Model training process: Python

[0019] params = { 'objective': 'reg:squarederror', 'max_depth': 6, 'learning_rate': 0.05, 'subsample': 0.8, 'colsample_bytree': 0.8, 'n_estimators': 200 }

[0020] X = pd.get dummies(df[['behavior type', 'time_slot']]) X = pd.concat([X, df[['intensity', 'baseline_bp']]], axis=1)

[0021] model = xgb.train(params, dtrain, num boost round=100) ``` - Online update mechanism: - Daily new data is stored in Huawei Cloud OBS object storage. - Update model weights using the FTRL algorithm: \[ w {i,t+1} = \text{sign}(z{i,t}) \cdot \left( \frac{1}{\eta} + \sqrt{\sum {s=1}^tg {i,s}^2} \right)^{-1} \] Where \( z {i,t} = z {i,t-1} + g {i,t} - \sigma w {i,t} \), \( \sigma \) are the regularization coefficients.

[0022] - AR interface rendering: - Expected impact value: Utilizes a 3D bar chart with dynamic growth animation; height is proportional to score. - Blood pressure trend line: drawn using Unity's Line Renderer component, with color changing according to blood pressure value (green → yellow → red). - Behavior Completion Rate: Displayed as a circular progress bar; particle effects are triggered when the completion rate reaches 80%. - Reward distribution logic: SQL -- Points Accumulation Inquiry SELECT SUM(impact score) FROM user behavior WHERE user id = ? AND create time > DATE_SUB(NOW(), INTERVAL 7 DAY); -- Reward Triggering Conditions IF total_score >= 500 THEN INSERT INTO reward records VALUES(?, 'health coupon', 20); UPDATE user profile SET VIP level = vip level + 1 WHERE user id = ?; END IF; ```

[0023] In a clinical trial involving 300 people: - Model prediction error: Systolic blood pressure MAE = 3.2 mmHg, Diastolic blood pressure MAE = 2.1 mmHg - Changes in user behavior: Regular exercise rate increased from 28% to 67%, and low-salt diet adherence rate increased from 41% to 79%. - Blood pressure control effect: After 6 months, the average systolic blood pressure decreased by 14.2 mmHg (p<0.01). Example 2: Lightweight Equipment Implementation Plan

[0024] 1. Hardware optimization solutions - Employs Nordic nRF52840 main control chip, integrating STMicroelectronics LPS22HH pressure sensor - Remove the ECG module and extract HRV parameters using the PPG waveform: - Peak detection: Pan-Tompkins algorithm employed - RR interval calculation: Dynamic threshold adjustment to adapt to different motion states - Behavioral data collection: - Movement: Only accelerometer data is retained; walking / running is identified using a threshold method. - Diet: Simplified audio analysis, only identifying the number of chews (>15 times per minute is considered eating).

[0025] - The model size was compressed from 12MB to 1.8MB using TensorFlow Lite quantization. - Feature dimensionality reduction: - Original model: 23 dimensions (behavioral type × 3 + intensity × 5 + time × 3 + baseline blood pressure × 2) - Optimized: 9 dimensions (exercise duration + dietary calories + sleep efficiency + resting heart rate + baseline blood pressure) - Inference speed improvement: reduced from 82ms / inference to 23ms / inference (ARM Cortex-M4@64MHz)

[0026] - Display device: Switched to E-Ink electronic ink screen, reducing power consumption by 82%. - Feedback content: - Instant feedback: Only displays the expected impact value and simple emojis. - Long-term feedback: Weekly points report sent via SMS - Reward mechanism: Points can be redeemed for public transport card top-up services, catering to the needs of middle-aged and elderly users.

[0027] In a test involving 100 elderly people: - Device battery life: Extended from 3 days to 14 days (300mAh battery) - Operational complexity: System availability score (SUS) of 82.3 out of 100. - Blood pressure control effect: Systolic blood pressure decreased by an average of 11.7 mmHg (p<0.05) Example 3: Implementation Plan for Enterprise Health Management Scenarios

[0028] 1. System Expansion Design - Add an enterprise back-end management system: - It adopts a Spring Cloud microservice architecture, including modules for user management, data dashboards, and reward configuration. - Database sharding and partitioning: Sharding by enterprise ID hash, supporting simultaneous access by 1000 enterprises. - Customized incentive rules: - Department Ranking: Team points are tallied weekly / monthly, with the top 3 teams receiving a team-building fund. - Health Challenge: Featuring themed tasks such as "21-Day Low-Salt Diet," participants who complete the challenge receive an extra vacation.

[0029] - Transmission encryption: Employs the national standard SM4 algorithm with a key length of 256 bits. - Storage Encryption: Huawei Cloud KMS service manages the master key, and the data encryption key (DEK) is rotated periodically. - Access Control: Fine-grained permission management is implemented based on the RBAC model. The CEO role can view data across the entire company, while department managers are limited to their own departments.

[0030] - Provide a RESTful API for third-party systems to call: http POST / api / v1 / behavior / record HTTP / 1.1 Host: health-incentive.huaweicloud.com Content-Type: application / json Authorization: Bearer {access_token} { "user_id": "10001", "behavior_type": "exercise", "intensity": 4.5, "duration": 30, "timestamp": 1672531200 } ``` - Response example: json { "code": 200, "message": "success", "data": { "impact_score": 28, "total_score": 452, "next_reward": "50 points to bronze badge" } } ```

[0031] In a pilot program at a company with 500 employees: - Employee participation rate: increased from 12% to 76%. - Average medical expenses per person: decreased by 23%, with an estimated annual cost saving of 870,000 yuan. - Productivity: Absences due to illness decreased by 41%, and project delivery cycles were shortened by 15%. The technical solution of this invention effectively solves the industry problem of low adherence to health management among patients with chronic diseases through three core technologies: multimodal behavior quantification, dynamic personalized modeling, and multi-dimensional incentive feedback. Actual test data shows that the system can improve the blood pressure control target achievement rate by 67%, demonstrating significant clinical value and social benefits.

Claims

1. A quantitative incentive system for health behaviors based on blood pressure feedback, characterized in that, include: The blood pressure monitoring module is used to periodically collect the user's blood pressure data; The behavioral data collection module is used to collect multi-dimensional behavioral data related to user health. The multi-dimensional behavioral data includes at least two of the following: exercise type and duration, food type and calorie intake, resting heart rate, sleep duration and quality, and stress index. The data processing and incentive calculation engine is connected to the blood pressure monitoring module and the behavior data acquisition module respectively. It is used to establish a personalized correlation between specific behavior types, behavior intensity and subsequent blood pressure changes based on historical blood pressure data and current multi-dimensional behavior data, through a personalized machine learning model trained based on the user's personal historical data. Based on the personalized correlation, it calculates the expected impact value of the current behavior on blood pressure. The expected impact value is a quantitative score of having a positive impact on blood pressure. The user incentive feedback module is connected to the data processing and incentive calculation engine. It is used to present users with visual incentive feedback information that includes the expected impact value, historical blood pressure trend and behavior achievement status. When the expected impact value accumulates to a preset threshold, virtual or physical rewards are triggered.

2. The system according to claim 1, characterized in that, The personalized machine learning model is a mapping model between behavioral factor vectors and blood pressure changes. The behavioral factor vectors include behavioral type, behavioral intensity, and behavioral occurrence time, while the blood pressure changes are the average blood pressure changes within a preset time period after the behavior occurs.

3. The system according to claim 1, characterized in that, The user incentive feedback module displays visual incentive feedback information in a graphical interface, including bar charts showing historical blood pressure trends, pie charts showing the percentage of behaviors that meet the standards, and numbers showing the current expected impact value.

4. The system according to claim 1, characterized in that, The virtual or physical rewards include at least one of the following: points, badges, level upgrades, coupons, and redemption rights for health products / services, and different reward types correspond to different expected cumulative threshold values.

5. The system according to claim 1, characterized in that, Also includes: The data communication module is used to synchronize the blood pressure data, multi-dimensional behavioral data, and incentive feedback information to the cloud server; The cloud server is used to perform cross-user data aggregation analysis to optimize the parameters of the personalized machine learning model and feed the optimized model parameters back to each user's system.

6. The system according to claim 5, characterized in that, When performing cross-user data aggregation analysis, the cloud server uses a clustering algorithm to divide users into different groups and optimizes the parameters of the personalized machine learning model for each group.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the functions of the system as described in any one of claims 1-6, including collecting blood pressure and behavioral data, establishing personalized associations, calculating expected impact values, and generating incentive feedback information.

8. A method for quantifying and incentivizing health behaviors based on blood pressure feedback, characterized in that, The system applied to any one of claims 1-6 comprises: Periodically acquire users' blood pressure monitoring data; Collect users' multi-dimensional health behavior data in real time or periodically; Based on a personalized machine learning model trained on users' personal historical data, the expected impact of current multi-dimensional health behavior data on users' blood pressure is analyzed. The personalized machine learning model establishes a personalized correlation between specific behavior types, behavior intensity and subsequent blood pressure changes. The expected impact value is calculated, and the expected impact value is a quantitative score for having a positive impact on blood pressure; Based on the expected impact value, generate and output incentive feedback to the user that includes visual incentive feedback information and triggers virtual or physical rewards when the expected impact value accumulates to a preset threshold.

9. The method according to claim 8, characterized in that, "Analyzing the expected impact of current multi-dimensional health behavior data on users' blood pressure based on personalized machine learning models" specifically includes: Extract behavioral factor vectors from current multidimensional health behavior data. The behavioral factor vectors include behavior type, behavior intensity, and behavior occurrence time. The behavioral factor vector is input into the personalized machine learning model, and the personalized machine learning model outputs the predicted positive trend value of blood pressure improvement as the expected impact value.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in claim 8 or 9, including the steps of collecting data, establishing correlations, calculating expected impact values, and generating incentive feedback.

Images