Dynamic safety grading home health monitoring intelligent single system and use method thereof
By using 3D-TOF body posture recognition equipment and a deep learning framework that integrates multimodal data fusion, along with a dynamic safety grading system, the problems of high false alarm rate, long response time, and inefficient resource allocation in traditional home health monitoring systems have been solved, achieving efficient, accurate home health monitoring and rapid response.
Patent Information
- Application Number
- CN202510830095.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-20
- Publication Date
- 2025-11-18
AI Technical Summary
Traditional home health monitoring systems suffer from problems such as reliance on manual judgment for work order triggering, high false alarm rates, data silos leading to inaccurate early warnings, fragmented response processes, lack of dynamic grading mechanisms, conflict between privacy and functionality, inability to identify sudden risks in real time, long response times, and inefficient resource allocation.
The system employs 3D-TOF body posture recognition equipment, health monitoring equipment, and environmental sensors for real-time monitoring. It combines a multimodal fusion deep learning framework for data analysis, a dynamic safety grading system to assess risk levels, and automatically dispatches work orders, including three levels: emergency, early warning, and routine, and links with the community service system.
Significantly reduce false alarm rate, shorten emergency response time, optimize resource allocation, improve user experience, improve work order response efficiency and resource utilization, and provide a closed-loop feedback mechanism.
Smart Images
Figure CN120975422A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application belongs to the technical field of smart home and health service, and relates to a dynamic safety grading home health monitoring intelligent work order system and a use method thereof. BACKGROUND
[0002] The traditional home health monitoring system mainly adopts the following technical solutions: 1. Single device monitoring: relying on independent devices such as smart bracelets, emergency call buttons or cameras, for example: 1) Wearable devices (such as smart bracelets): detecting falls through acceleration sensors, or monitoring basic physiological data such as heart rate and step count; 2) Camera monitoring: using visual recognition technology to detect falls or abnormal behavior; 3) Emergency button: the user actively triggers the alarm, lacking automatic early warning capability; 2. Manual inspection and response: community services rely on regular manual inspection or user-initiated reporting, and work order dispatching requires manual review of priority; 3. Decentralized data management: health data (heart rate, blood sugar), environmental data (temperature and humidity, air quality) and behavior data (falls, body movement) are stored in a decentralized manner, and a unified analysis platform has not been formed.
[0003] Therefore, the traditional health monitoring system has the following limitations: 1. Work order triggering relies on manual judgment: 1) Abnormal events (such as night falls and respiratory pauses) need to be reported by the user or family members, and sudden risks cannot be identified in real time; 2) High false positive rate of wearable devices (such as misjudging vigorous exercise as a fall), leading to invalid work order dispatching; 2. Data silos lead to inaccurate early warning: 1) Health data (heart rate, blood sugar), behavior data (falls, body movement) and environmental data (PM2.5) are not analyzed together, making it difficult to distinguish between real risks and false positives; 2) Events are not handled in association, for example: air pollution and an elderly person falling at the same time, leading to confusion in work order priority; 3. Response process is fragmented and inefficient: 1) The community service system is not connected with the health data platform, and emergency events (such as ambulance calls after a fall) rely on manual dispatch, with an average response time of more than 3 minutes; 4. Privacy and functionality conflict: 1) Camera monitoring infringes on privacy, while alternative solutions (such as wearable devices) cannot cover high-risk scenarios (such as bathroom falls) due to technical limitations; 5. Lack of dynamic grading mechanism: 1) All abnormal events are handled with fixed priority, without dynamically adjusting response strategy according to severity (e.g. apnea and night rise are not distinguished in level).
[0004] Therefore, a dynamic security grading home health monitoring intelligent work order system and its use method are designed to overcome the above problems. SUMMARY
[0005] The purpose of the present application is to overcome the shortcomings of the prior art, and to provide a dynamic security grading home health monitoring intelligent work order system and its use method, which has simple and reasonable structure, improves work order response efficiency, reduces false positive rate, optimizes resource allocation and enhances user experience.
[0006] The present application is realized by the following technical solutions: a dynamic security grading home health monitoring intelligent work order system, which comprises a collection system, a data processing and analysis system, a dynamic security grading system and a work order execution system, which work in association with each other. The collection system is connected with 3D-TOF body posture recognition equipment, health monitoring equipment and environmental sensors, respectively, for real-time monitoring of human body posture, health data and environmental conditions. The data processing and analysis system comprises a health data platform, which analyzes and operates the platform after converting the detected data. The dynamic security grading system performs security evaluation on the analyzed and operated data, and performs grade scoring after evaluation. The work execution system processes the scored data by corresponding task dispatching.
[0007] As a preferred embodiment, the body posture recognition equipment is installed in the bedroom or bathroom, respectively, for real-time monitoring of falling or night rise action retardation behavior. The health monitoring equipment comprises a smart wristwatch, a smart mattress, an Internet of Things blood pressure meter or a blood glucose meter. The environmental sensors comprise a smoke alarm, a water immersion alarm and a door magnetic alarm.
[0008] As a preferred embodiment, the health data platform comprises AI model construction and training based on a multi-modal fusion deep learning framework, pre-training is performed through public medical data sets and self-built 3D-TOF body posture action library, and migration learning is performed in combination with historical health data to optimize small sample scenarios.
[0009] As a preferred embodiment, the grade scoring is divided into three levels, namely, first-level emergency work order, second-level warning work order and third-level daily work order. The first-level emergency work order directly outputs first-aid instructions from the health data platform, calls an ambulance and other first-aid units. The second-level warning work order is generated and sent to family members and property management for subsequent processing. The third-level daily work order generates a daily report and transmits it to a community hospital for follow-up.
[0010] A use method of a dynamic security grading home health monitoring intelligent work order system, the method comprising the following steps: 1) Build an intelligent device environment: collect through intelligent devices, including 3D-TOF body recognition devices, health monitoring devices, and environmental sensor devices, respectively collect the fall or night rise action retardation behavior in the bedroom or bathroom, real-time heart rate and respiration, apnea, blood pressure, and blood glucose meter conditions; 2) Build a health data platform: through the construction and training of AI models, based on a multi-modal fusion deep learning framework: LSTM+CNN hybrid model, pre-trained through public medical data sets and self-built 3D-TOF body action library, combined with historical health data transfer learning, optimize small sample scenarios; 3) Add the intelligent device built in step 1) to the AI model in step 2) for actual scene operation, input data, and analyze output data results; 4) Dynamic safety classification work order through analyzed output data, and task assignment, including: First-level emergency work order: direct call to emergency system; Second-level warning work order: generate environment disposal suggestions and push to family members and property; Third-level daily work order: generate health report summary and associate community doctor follow-up; 5) After the work order is executed, it is returned to the health data platform, from first-level to second-level, and second-level to third-level for circulating work order assignment.
[0011] As preferred: the actual scene operation logic in the AI model in step 3) is: 1) Time series analysis: identify apnea: ≥10 seconds and frequency >5 times / hour, heart rate drop: 1 minute drop >30% with body motion disappearance, both are time series anomalies; 2) Spatial behavior pattern: combined with 3D-TOF data, data trunk inclination angle >45°+touching ground static time >10 seconds, heart rate and respiration in physiological data to determine whether the anomaly is a real fall; 3) Environmental correlation risk: PM2.5 exceeds standard: >75 μg / m³, superimposed on respiratory rate: >22 times / minute, both trigger compound warning.
[0012] Compared with the prior art, the present application has the following beneficial effects: The present application adopts: 1. Work order response efficiency is improved: the response time of emergency events (such as falls) is shortened from 3-5 minutes in traditional manual processing to 5 seconds, and the average ambulance arrival time is reduced by 40%; 2. False positive rate is significantly reduced: 3D-TOF body recognition technology makes the fall detection accuracy rate reach 99.2% (traditional camera scheme is 85%), and the apnea false positive rate is <1 time / week; 3. Resource optimization: Dynamic classification mechanism improves community service resource utilization by 60% (such as prioritizing first-level work orders and avoiding low-risk events from occupying emergency channels); 4. Enhanced user experience: Closed-loop feedback mechanism allows family members to view work order status (such as "ambulance has departed") in real time through mobile app, reducing anxiety complaints by 75%. BRIEF DESCRIPTION OF DRAWINGS
[0013] Figure 1 is the overall flowchart of the present application. DETAILED DESCRIPTION
[0014] In order for those skilled in the art to more clearly understand the purpose, technical solution and advantages of the present application, the present application will be further described below in conjunction with the drawings and examples.
[0015] In the description of the present application, it should be understood that the orientation or positional relationship indicated by the terms "up", "down", "left", "right", "in", "out", "horizontal", "vertical" and the like is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present application, and does not indicate or imply that the device or element referred to must have a particular orientation, therefore it cannot be understood as a limitation on the present application.
[0016] The present application will be described in detail below in conjunction with the drawings: as shown in Figure 1 a dynamic security classification home health monitoring intelligent work order system, the system includes a collection system, a data processing and analysis system, a dynamic security classification system, a work order execution system, which work with each other, the collection system is connected with 3D-TOF body state recognition device, health monitoring device, environmental sensor respectively, real-time monitoring of human body state, health data, environmental conditions, the data processing and analysis system includes a health data platform, which converts the detected data for platform analysis and operation, the dynamic security classification system performs security evaluation on the analyzed and operated data, and performs grade scoring after evaluation, and the work execution system processes the corresponding tasks according to the scored data.
[0017] The body state recognition device is installed in the bedroom or bathroom, respectively, to monitor the behavior of falling or getting up slowly at night, the health monitoring device includes a smart wristwatch, a smart mattress, an Internet of Things blood pressure meter or a blood glucose meter, and the environmental sensor includes a smoke alarm, a water immersion alarm and a door magnetic alarm.
[0018] The health data platform includes AI model construction and training, based on a multi-modal fusion deep learning framework, pre-training is performed through public medical data sets and self-built 3D-TOF body state action library, and migration learning is performed in combination with historical health data to optimize small sample scenarios.
[0019] The grade score is divided into three levels, namely, first-level emergency work order, second-level early warning work order, and third-level daily work order. The first-level emergency work order directly outputs first-aid instructions from the health data platform, calls an ambulance, and the like. The second-level early warning work order is sent to the family and property for subsequent processing. The third-level daily work order is generated into a daily report and transmitted to a community hospital for follow-up.
[0020] A use method of a dynamic security grading intelligent work order system for home health monitoring, the method comprising the following steps: 1) Building an intelligent device environment: collecting through intelligent devices, wherein the devices include 3D-TOF body posture recognition devices, health monitoring devices, and environmental sensor devices, respectively collecting fall or night-rise action retardation behaviors in a bedroom or bathroom, real-time heart rate and respiration, apnea, and blood pressure and blood glucose meter conditions; 2) Building a health data platform: constructing and training an AI model based on a multi-modal fusion deep learning framework: an LSTM+CNN hybrid model, pre-training through public medical data sets and self-built 3D-TOF body action libraries, combining historical health data transfer learning, and optimizing small sample scenarios; 3) Adding the intelligent devices built in step 1) to the AI model in step 2) for actual scene operation, inputting data, and analyzing output data results; 4) Dynamically grading work orders through analyzed output data, and assigning tasks, wherein the grading includes First-level emergency work order: directly calling a first-aid system; Second-level early warning work order: generating environment disposal suggestions and pushing to the family and property; Third-level daily work order: generating a health report summary and associating community doctors for follow-up; 5) After the work order is executed, it is returned to the health data platform, and the first level is converted to the second level, and the second level is converted to the third level for circulating work order assignment.
[0021] The actual scene operation logic in the AI model in step 3) is as follows: 1) Time series analysis: identifying apnea: ≥10 seconds and frequency >5 times / hour, heart rate drop: 1-minute drop >30% with body movement disappearance, both of which are time series anomalies; 2) Spatial behavior pattern: combining 3D-TOF data, data trunk inclination angle >45°+touching the ground for >10 seconds, and combining heart rate and respiration in physiological data to determine whether the anomaly is a real fall; 3) Environmental correlation risk: PM2.5 exceeds the standard: >75 μg / m³, superimposed on accelerated breathing frequency: >22 times / minute, which triggers a composite early warning.
[0022] The core steps and design features of the present application are as follows: I. Data collection 1. 3D-TOF body recognition device: deployed in bedroom, bathroom and other areas, non-inductive monitoring of falling, night action delay and other behaviors (avoiding privacy leakage); 2. Health monitoring device: smart wristwatch (real-time heart rate / respiration), smart mattress (apnea detection), Internet of Things blood pressure meter / glucometer (data automatic upload); 3. Environmental sensor: smoke, water immersion, door magnetic alarm device, linkage safety warning.
[0023] II. Data processing and analysis 1. Health data platform: 1) AI model construction and training, based on multi-modal fusion deep learning framework (such as LSTM+CNN hybrid model), pre-trained through public medical data set (such as MIMIC-III vital sign data) and self-built 3D-TOF body action library (containing 3D point cloud data of 10 types of actions such as falling and squatting), combined with historical health data (100,000+ records of abnormal events of homebound elderly) for transfer learning, to optimize small sample scenarios (such as apnea events); 2) Input data: 3D-TOF body data: three-dimensional coordinates of human body key points (such as head, torso, and limb spatial position changes), sampling frequency 20Hz; Health time series data: smart wristwatch (heart rate variability, respiration rate), smart mattress (apnea interval, body movement frequency), Internet of Things blood pressure / glucometer (blood pressure / glucometer fluctuation curve); Environmental data: PM2.5 concentration, smoke concentration, water immersion sensor state, door magnetic switch frequency; 3) Output results: Risk score (0-100 points): dynamically calculated according to event severity (such as falling = 95 points, PM2.5 exceeding = 60 points) Abnormal event classification: life risk events (respiratory arrest, severe fall), environmental risk events (fire, water leakage), health warning events (sleep apnea, blood pressure surge); False alarm filtering: through multi-sensor cross verification (such as falling with normal heart rate and continuous respiration is determined as "active lying down").
[0024] 2. Dynamic safety grading module: 1) AI analysis logic: Time series analysis: identify time series anomalies such as apnea (≥10 seconds and frequency >5 times / hour), heart rate drop (1 minute drop >30% with body movement disappearance); Spatial behavior pattern: Combine 3D-TOF data (trunk inclination angle > 45° + touch ground static time > 10 seconds) with physiological data (abnormal heart rate / respiration) to determine real fall; Environment-related risk: PM2.5 exceeds standard (> 75 μg / m³) combined with accelerated breathing rate (> 22 times / minute) triggers compound warning; 2) Classification strategy: First-level emergency work order: AI output first aid instructions (including GPS positioning and health record coding), directly occupy system resources (such as automatically calling an ambulance); Second-level warning work order: Generate environmental disposal suggestions (such as starting air purification to 80% power), and push to family members and property management; Third-level daily work order: Generate health report summary (such as "past 7 days night body movement times increased by 20%"), associate community doctors for follow-up.
[0025] Three, work order execution 1. Intelligent work order dispatch engine: match service resources according to safety level (such as first-level work order directly connected to first aid center, second-level work order dispatched to property maintenance); 2. Feedback loop mechanism: service execution results (such as ambulance arrival time, purification equipment status) are fed back to the platform to optimize subsequent classification strategies.
[0026] Four, core technology 1. Multi-modal data fusion 1) 3D-TOF body state data and physiological data (heart rate, respiration) are spatio-temporally aligned, and key features (such as heart rate changes at the moment of falling) are extracted through attention mechanism; 2) Environmental data (PM2.5, smoke) and health data are jointly modeled to identify compound risks (such as air pollution-induced respiratory abnormalities); 2. Dynamic safety classification algorithm 1) Dynamically adjust work order level based on risk score threshold (such as automatically upgrading to second-level work order for two consecutive apnea events); 2) Work order priority adaptive matching: first-level work order occupies system resources (such as automatically dialing emergency number and sending location), second-level work order is processed in parallel (such as starting purification equipment and pushing family member APP at the same time).
[0027] Five, implementation rules 1. Equipment deployment: install 3D-TOF equipment and intelligent mattress in the bedroom, configure anti-slip alarm pad in the bathroom, and deploy environmental sensors in the living room; 2. Service interface: health data platform connects with community medical system, property management system, and first aid center platform through API, supporting automatic dispatch and status tracking of work orders.
[0028] The present application has the following characteristics: 1. Multimodal data fusion trigger work order: integrate 3D body posture recognition (falling), physiological monitoring (heart rate / respiration), environmental sensor (PM2.5) data, and construct a dynamic early warning model; 2. Dynamic classification of security level: based on AI algorithm, risk quantification is performed on abnormal events (such as apnea frequency, falling posture confidence), and automatic division of work order priority (first level emergency to third level reminder); 3. Closed-loop work order dispatching and feedback: through the linkage of Internet of Things devices and community service systems, realize the whole process automation of "monitoring-analysis-dispatching-execution-feedback", shorten the response time.
[0029] The specific embodiments described herein are merely illustrative of the principles of the present application and its efficacy, and are not intended to limit the present application. Any person skilled in the art can modify or change the above embodiments without departing from the spirit and scope of the present application. Therefore, all equivalent modifications or changes made by those skilled in the art without departing from the spirit and technical idea disclosed by the present application shall be covered by the claims of the present application.
Claims
1. A dynamic safety grading intelligent work order system for home health monitoring, comprising a data acquisition system, a data processing and analysis system, a dynamic safety grading system, and a work order execution system, which work in conjunction with each other, characterized in that: The acquisition system is connected to a 3D-TOF body posture recognition device, a health monitoring device, and an environmental sensor to monitor human body posture, health data, and environmental conditions in real time. The data processing and analysis system includes a health data platform, which converts the detected data and performs platform analysis and calculations. The dynamic safety grading system performs a safety assessment on the analyzed and calculated data and assigns a grade score. The work execution system dispatches corresponding tasks to process the graded data.
2. The intelligent work order system for dynamic safety grading home health monitoring according to claim 1, characterized in that: The body posture recognition devices are installed in the bedroom or bathroom to monitor falls or slow movements when getting up at night in real time. The health monitoring devices include smartwatches, smart mattresses, IoT blood pressure monitors or blood glucose meters, and the environmental sensors include smoke detectors, water leak detectors, and door magnetic alarms.
3. The intelligent work order system for dynamic safety grading home health monitoring according to claim 1, characterized in that: The health data platform includes AI model building and training, based on a multimodal fusion deep learning framework. It is pre-trained using publicly available medical datasets and a self-built 3D-TOF body movement library, and transfer learning is performed using historical health data to optimize small sample scenarios.
4. The intelligent work order system for dynamic safety grading home health monitoring according to claim 1, characterized in that: The rating system is divided into three levels: Level 1 Emergency Work Order, Level 2 Early Warning Work Order, and Level 3 Daily Work Order. Level 1 Emergency Work Orders are generated by the health data platform, which directly outputs emergency instructions to call ambulances and other emergency units. Level 2 Early Warning Work Orders are generated and sent to family members and property management for follow-up. Level 3 Daily Work Orders generate daily reports and send them to the community hospital for follow-up.
5. The method of using the intelligent work order system for dynamic safety grading home health monitoring according to any one of claims 1-4, characterized in that: The method includes the following steps: 1) Establish an intelligent device environment: Data is collected through intelligent devices, including 3D-TOF body posture recognition devices, health monitoring devices, and environmental sensor devices. These devices collect data on falls or slow movements at night in the bedroom or bathroom, as well as real-time heart rate, respiration, sleep apnea, blood pressure, and blood glucose levels. 2) Building a health data platform: Through the construction and training of AI models, based on the multimodal fusion deep learning framework: LSTM+CNN hybrid model, pre-training is performed using publicly available medical datasets and a self-built 3D-TOF body movement library, combined with historical health data transfer learning, to optimize small sample scenarios; 3) Add the intelligent devices built in step 1) to the AI model in step 2) for real-world scenario operation, input data, and analyze and output data results; 4) Dynamically classify safety work orders based on the analyzed output data and assign tasks, where the classification includes... Level 1 Emergency Work Order: Directly calls the emergency medical services system; Level 2 early warning work order: Generates environmental handling suggestions and pushes them to family members and property management in parallel; Level 3 routine work orders: Generate health report summaries and link them to community doctor follow-ups; 5) After the work order is completed, it is sent back to the health data platform, where it is converted from Level 1 to Level 2, and then from Level 2 to Level 3 for cyclical work order dispatch.
6. The method of using the intelligent work order system for dynamic safety grading home health monitoring according to claim 5, characterized in that: The logic for running the AI model in the actual scenario in step 3) is as follows: 1) Time series analysis: Identify apnea: ≥10 seconds and frequency >5 times / hour, and sudden drop in heart rate: >30% decrease in 1 minute accompanied by loss of body movement, both of which are time series abnormalities; 2) Spatial behavior pattern: Combining 3D-TOF data, the data shows that the trunk tilt angle is >45° and the ground contact time is >10 seconds. The physiological data, such as heart rate and respiration, are combined to determine whether the abnormality is a real fall. 3) Environmental risks: PM2.5 exceeding the standard: >75μg / m³, coupled with increased breathing rate: >22 breaths / minute, triggers a combined warning.