Intelligent home maintenance system

Abstract

The invention relates to the technical field of intelligent health management, and discloses an intelligent home care system which comprises monitoring equipment located at a user side, the monitoring equipment is used for monitoring the activity state and health data of a user and environment data of a user activity area in real time, and the monitoring equipment is in wireless communication connection with a cloud server. The cloud server comprises a data receiving and processing module, an AI algorithm module, a user data storage module, a risk triggering module and a health management module, and the data receiving and processing module is used for receiving and preprocessing data from the user side monitoring equipment; the AI algorithm module analyzes the data processed by the data receiving and processing module based on a multi-modal data fusion method and generates an early warning signal; and when the risk triggering module detects a risk event, the health management module or the AI emergency platform is automatically triggered. According to the system, the physiological data of the user is collected in real time, a multi-modal data fusion framework is provided, and the health event recognition accuracy is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent health management systems, in particular to an intelligent home care system. BACKGROUND

[0002] With the continuous development of smart home and health management technology, intelligent home care systems have gradually become an important tool for improving people's quality of life. However, the existing technology still has certain limitations in realizing intelligence and efficient operation, mainly in the following aspects: traditional home care systems usually rely on multiple sensors to collect real-time data. However, the acquisition time and sampling frequency of different types of sensor data are different, resulting in a spatio-temporal asynchronous problem between data. This problem makes it difficult for the system to obtain accurate and consistent results when processing and analyzing data, thereby affecting the decision-making and response speed of the system. How to effectively solve the synchronization problem of different sensor data has become one of the key challenges to improve the performance of home care systems. At present, most home care systems separate the first-aid system and the health management system and run independently, and the interoperability between the two is poor, forming a data island. The first-aid system mainly deals with emergencies, and the health management system is responsible for long-term health monitoring and intervention. Although both are important parts of home care, due to the lack of effective data sharing and fusion mechanism, when a health crisis occurs, the first-aid system cannot obtain relevant historical data in the health management system in time, affecting the rescue decision and effect. Therefore, how to break the data island problem between the two and make the system work more intelligently is an important problem to be solved in the current home care system. SUMMARY

[0003] In view of the above technical problems in the related art, the present application provides an intelligent home care system, which can solve the above problems.

[0004] To achieve the above technical purposes, the technical scheme of the present application is as follows: The intelligent home maintenance system comprises a monitoring device at the user end, which is used to monitor the activity state, health data and environmental data of the user activity area in real time. The monitoring device is wirelessly connected to a cloud server. The cloud server comprises a data receiving and processing module, an AI algorithm module, a user data storage module, a risk triggering module and a health management module. The data receiving and processing module receives data from the user end monitoring device and performs preprocessing. The AI algorithm module analyzes the data processed by the data receiving and processing module based on a multi-modal data fusion method and generates an early warning signal. When the risk triggering module detects a risk event, it automatically triggers an AI first aid platform. The user data storage module stores the health data and behavior data of the user for long-term analysis and health management. The health management module is used for health management of the user.

[0005] Further, the monitoring device comprises a fall monitoring radar, a vital sign detection radar, a temperature and humidity sensor and a gas sensor. The activity state comprises falling and getting out of bed. The health data comprises respiratory data and heart rate data. The environmental data comprises temperature and humidity data and gas data.

[0006] Further, the monitoring device sends data to the cloud server through Wi-Fi or 4G / 5G network.

[0007] Further, the data receiving and processing module performs data cleaning and validity filtering on the received data.

[0008] Further, the multi-modal data fusion method comprises the following steps: S100, inputting multi-modal data; S200, time alignment processing of multi-modal data to ensure synchronization of data of different modalities in time; S300, using a lightweight convolutional neural network to extract features of data of each modality, projecting features extracted by different modalities to the same feature space and unifying dimensions; S400, using dot product method to calculate attention scores between different modal features, performing Softmax normalization processing on the attention scores to obtain an attention weight matrix; S500, combining modality confidence and reliability factor to assign real-time weight to data of each modality, performing normalization processing on the real-time weight to ensure that the sum of the weights is 1; S600, using the normalized weight to weight and sum the features to obtain a decision score, and dividing the risk level according to the decision score.

[0009] Further, the decision score is compared with a risk level threshold, so as to divide the risk level, which includes a first-level blue warning, a second-level orange warning and a third-level red warning.

[0010] Further, when the warning signal is the first-level blue warning and the second-level orange warning, a health management module is triggered to form a health management cycle; when the warning signal is the third-level red warning, an AI first-aid platform is triggered to form a first-aid response cycle.

[0011] Further, the health management module comprises a data analysis unit, a user health baseline unit, a health suggestion generation unit, a multi-terminal pushing unit, an effect tracking unit and a dynamic optimization unit, wherein the user health baseline unit contains a set of user health data thresholds and health labels; the data analysis unit performs deep analysis based on the historical health data and behavior data of the user and the data contained in the user health baseline unit; the health suggestion generation unit provides personalized health management suggestions for the user according to the analysis results and generates a health report of the user; the multi-terminal pushing unit is used to notify the health management suggestions to the mobile phone or television of the user; the effect tracking unit is used to track the effect of executing the health management suggestions; and the dynamic optimization unit adjusts various threshold data of the user health baseline according to the service tracking data of the AI first-aid platform.

[0012] Further, the AI first-aid platform comprises a warning receiving unit, a service scheduling unit and a service tracking unit, wherein the warning receiving unit is used to receive the risk warning signal from the cloud server; the service scheduling unit quickly schedules the nearest service personnel to go to the home of the user according to the warning information; and the service tracking unit tracks the response situation and service progress of the service personnel in real time to ensure that the service is completed in time.

[0013] Further, the medical care to home service network is further included, which is connected with medical care units in a nationwide range, and after receiving the scheduling instruction of the AI first-aid platform, the medical care to home service network quickly responds and makes service arrangement, and feeds back the service result to the first-aid platform and the cloud server.

[0014] The system of the present application has the following advantages: the system collects the physiological data of the user in real time, provides a multi-modal data fusion architecture, improves the health event recognition accuracy, and through three-level warning decision, a data feedback mechanism of first-aid and health double cycle is constructed to establish a service closed loop architecture, the architecture can enter the first-aid cycle or the health cycle respectively, and a data feedback channel is established between the two cycles, so as to form a double cycle system closed loop service composed of the first-aid response cycle and the health management cycle, and the blank of realizing complete service closed loop technology in the field of home monitoring in China is filled. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to make the technical solutions in the embodiments of the present application or the prior art clearer, the accompanying drawings needed in the embodiments will be briefly introduced below. Obviously, the accompanying drawings in the following description only represent some embodiments of the present application, and other drawings can be obtained by those of ordinary skill in the art without any creative effort based on these drawings.

[0016] The present application will be further described in detail below according to the accompanying drawings.

[0017] Fig. 1 is a schematic diagram of an intelligent home care system according to an embodiment of the present application; Fig. 2 is a flowchart of a multi-modal data fusion method according to an embodiment of the present application; Fig. 3 is a double-cycle schematic diagram composed of an emergency response cycle and a health management cycle according to an embodiment of the present application. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments only represent some embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art belong to the scope of protection of the present application.

[0019] As shown in Figs. 1-3 , an intelligent home care system is disclosed according to the present application, which comprises a monitoring device at a user end, the monitoring device being used to monitor the activity state of a user, health data and environmental data of a user activity area in real time, the monitoring device being wirelessly connected to a cloud server, the cloud server comprising a data receiving and processing module, an AI algorithm module, a user data storage module, a risk triggering module and a health management module, wherein the data receiving and processing module is used to receive data from the monitoring device at the user end and perform preprocessing; the AI algorithm module analyzes the data processed by the data receiving and processing module based on a multi-modal data fusion method and generates an early warning signal; the risk triggering module automatically triggers an AI emergency platform when a risk event is detected; the user data storage module stores health data and behavior data of a user for long-term analysis and health management; and the health management module is used for health management of a user.

[0020] Embodiment I: The monitoring device includes fall monitoring radar, vital sign detection radar, temperature and humidity sensor, gas sensor and other monitoring elements, which are deployed in the user's home, such as the bedroom, bathroom and other key areas, for real-time monitoring of the user's activity state (such as falling, getting out of bed, etc.), health data (such as heart rate, breathing), and environmental data such as temperature and humidity, gas, etc.

[0021] Embodiment two: The data collected by the monitoring device is sent to the cloud server through Wi-Fi or 4G / 5G network, received by the data receiving and processing module, and preprocessed, such as filtering based on device parameter range (such as normal heart rate range 60-100bpm, extreme value outside the range is judged as abnormal, marked as invalid); based on time sequence continuity verification (such as environmental temperature sudden rise / drop in a short time (such as 1 minute from 25℃→50℃), and no other device data to support, judged as sensor failure data, rejected); short-term data loss caused by network interruption, if the missing time is short (such as 1-2 sampling periods), it can be supplemented by interpolation method (such as linear interpolation, cubic interpolation and other processing logic); if the missing time is long (such as more than 10 sampling periods) or the key data (such as the skeleton point sequence of the fall radar) is missing, it is marked as "data invalid section" to avoid affecting the subsequent AI algorithm judgment.

[0022] Embodiment three: The multi-modal data fusion method includes the following specific steps: (1) Input of multi-modal data The multi-modal data includes the following data: Fall monitoring radar data (K): skeleton point sequence, sampling rate is 25fps, dimension is 128.

[0023] Vital sign detection radar data (Q): respiratory rate, heart rate, blood oxygen, tremor, sampling rate is 1Hz, dimension is 64.

[0024] Environmental sensor data (V): temperature, humidity, gas concentration, sampling rate is 0.2Hz, dimension is 32.

[0025] (2) Time alignment processing (ensure that the data of different modalities are synchronized in time, which is convenient for subsequent processing) Take the fall monitoring radar data (K) 25fps as the reference, the vital sign detection radar data (Q) has 1 valid point every 25 frames, and the environmental sensor data (V) has 1 valid point every 125 frames. Add-1e9 mask to the unsampled time points.

[0026] (3) Feature extraction and projection (convert the data of different modalities into comparable form, which is convenient for subsequent calculation of attention weight) A lightweight convolutional neural network (CNN) is used to extract features from the data for each modality. The features extracted from different modalities are projected onto the same feature space with a uniform dimension of 128.

[0027] (4) Cross-modal attention calculation (capturing the inherent correlation between data from different modalities to provide a basis for subsequent dynamic weight allocation) Attention scores between different modal features are calculated using a dot product method, resulting in a 128-dimensional score matrix. Invalid time points are masked to ensure that attention scores are only calculated for valid time points. The attention scores are then Softmax normalized to obtain the attention weight matrix.

[0028] For standardized multimodal feature vectors (such as feature vector K_vec of K, feature vector Q_vec of Q, and feature vector V_vec of V), the attention score is calculated as follows: For any two feature vectors of different modalities (such as K_vec and Q_vec, K_vec and V_vec, Q_vec and V_vec), a vector dot product operation is performed: the corresponding elements of the two feature vectors are multiplied one by one, and then all the product results are summed to obtain the individual attention score value of the modal feature. By traversing all modal combinations according to the above rules, the final generated attention scores will form an attention score matrix with a dimension of 128 (consistent with the unified dimension of the features).

[0029] (5) Dynamic weight allocation (dynamically allocate weights based on the importance and reliability of different modal data to improve decision-making accuracy.) By combining modality confidence and reliability factors, real-time weights are assigned to the data for each modality. These real-time weights are then normalized to ensure that the sum of the weights is 1. Dynamic weight allocation can be achieved using the following formula: in, The modal confidence level is the probability value output by the model (e.g., a fall detection confidence level of 0.92), which takes the value [0, 1]. The reliability factor is dynamically calculated, representing long-term sensor stability, and takes a value of [0, 5]. The calculation formula is as follows: ; (Instant accuracy) is the accuracy of the most recent 100 tests (sliding window). (Historical stability) is the sensor's failure rate over the past 30 days, calculated using the following formula: ; This is the attenuation factor (default 0.7, updated hourly).

[0030] (6) Output decision (Initiate the corresponding service process according to the risk level to ensure that users receive timely and effective assistance.) The features are weighted and summed using the normalized weights to obtain a decision score, and the risk level is divided according to the decision score.

[0031] The risk level is divided according to the decision score: LEVEL3 (red): Score ≥ 0.9, trigger emergency cycle.

[0032] LEVEL2 (orange): 0.7 ≤ decision score < 0.9, trigger intensive intervention of health cycle.

[0033] LEVEL1 (blue): When 0.5 ≤ decision score < 0.7, trigger regular intervention of health cycle.

[0034] Further, the multi-modal data fusion method of the present application can generate an early warning model integrated in an intelligent home care system.

[0035] Embodiment Four: When the AI emergency platform completes service tracking, the smart contract is automatically triggered, the treatment results (such as vital sign data, prescription information) are written into the blockchain, and the user data storage module of the cloud server is updated synchronously. The core function of the blockchain smart contract is to store emergency service records, health intervention data, and other key information, ensuring that the data cannot be tampered with, and providing a secure and trusted basis for double-cycle data interaction. When the user health baseline unit needs to call the historical emergency data in the blockchain, it can access through the smart contract authorization to ensure data privacy and security.

[0036] Embodiment Five: The dynamic optimization unit can update the weight parameters of the early warning model every day at 2 am based on the double-cycle data (emergency records, health suggestion execution data) of the past 24 hours using the gradient descent method. At the same time, a dynamic optimization rule table is developed to clearly define the optimization logic in different scenarios, as follows: | Optimization trigger scenario | Health baseline adjustment rule | Early warning model parameter adjustment | | ---- | ---- | ---- | | Level3 emergency event | Heart rate baseline = (historical baseline + average heart rate during emergency) × 0.6 (weight) | Level3 threshold reduced by 3% (if recurrence rate > 15%) | | Level2 early warning for 3 consecutive times | Blood pressure threshold increased by 5% (if intervention effectiveness < 60%) | Life sign modality weight increased by 0.1 in attention calculation | | Health intervention effectiveness > 80% | Blood glucose baseline remains unchanged | Level2 threshold reduced by 5% |

[0037] Embodiment Six: The fall detection is illustrated by the following specific example. The elderly Li Laoma (75 years old) slips on the ground oil stains while cooking in the kitchen. The system detects the fall event in real time through multi-modal data fusion, triggering a red alarm.

[0038] (1) Sensor data as follows (2) Time alignment processing python # Dynamic interpolation alignment (target 25fps) aligned_K = cubic_spline_interpolation( orig_data=[72, 68, 118], orig_times=[0,1000,2000], target_times=[0,40,80,120] # Align Q's time axis ) # Output: K = [72.0, 71.2, 72.8, 118.0] aligned_V = linear_interpolation( orig_data=[0.35], orig_times=[0], target_times=[0,40,80,120] ) # Output: V = [0.35, 0.35, 0.35, 0.35]。

[0039] The alignment results are as follows: (3) Feature extraction and projection python # Fall feature extraction (Q) Q_feat = self.fall_feature_extractor(Q) # Output: [0.21, 0.19, 0.92, 0.96] # Fall probability features # Vital sign feature extraction (K) K_feat = self.vital_feature_extractor(aligned_K) # Output: [0.05, 0.04, 0.06, 0.87] # Heart rate anomaly features # Environment feature extraction (V) V_feat = self.env_feature_extractor(aligned_V) # Output: [0.33, 0.33, 0.33, 0.33] # Low friction risk (4) Cross-modal attention computation python # 1. Compute attention scores (d_k=128) attn_scores = (Q_feat @ K_feat.T) / np.sqrt(128) # Score matrix: [[ 0.002, 0.002, 0.002, 0.018], [ 0.002, 0.002, 0.002, 0.017], [ 0.009, 0.009, 0.010, 0.080], # <- Key frame [ 0.010, 0.010, 0.011, 0.083]] # 2. Apply time mask (Vital signs radar 1Hz -> mark invalid points) mask = [[1,1,1,0], [1,1,1,0], [1,1,1,0], [1,1,1,1]] # Last time point valid attn_scores = np.where(mask, attn_scores, -1e9) # 3. Softmax normalization attn_weights = softmax(attn_scores, axis=-1) # Weight matrix: [[0.25, 0.25, 0.25, 0.25], [0.25, 0.25, 0.25, 0.25], [0.24, 0.24, 0.24, 0.28], # Start focusing on heart rate from 3rd frame [0.00, 0.00, 0.00, 1.00]] # Last frame fully focused on heart rate anomaly # 4. Feature fusion fused_feat = attn_weights @ V_feat # Output: [0.33, 0.33, 0.34, 0.33] # fused feature vector (5) Dynamic weight allocation python # 1. Calculate modality confidence c_Q = [0.21, 0.19, 0.92, 0.96] # Fall feature confidence c_K = [0.05, 0.04, 0.06, 0.87] # Vital signs confidence c_V = [0.33, 0.33, 0.33, 0.33] # Environment feature confidence # 2. Calculate reliability factors (device historical data) R_Q = 4.2 # Fall radar (new device, high reliability) R_K = 3.1 # Vital signs radar (battery aging) R_V = 1.8 # Environment sensor (kitchen oil contamination) # 3. Real-time weight optimization w_Q = exp(c_Q * R_Q) = [2.36, 2.18, 42.7, 46.2] w_K = exp(c_K * R_K) = [1.16, 1.13, 1.20, 13.5] w_V = exp(c_V * R_V) = [1.80, 1.80, 1.80, 1.80] # 4. Normalize weights sum = w_Q + w_K + w_V = [5.32, 5.11, 45.7, 61.5] w_Q_norm = [0.44, 0.43, 0.93, 0.75] # Fall radar weight w_K_norm = [0.22, 0.22, 0.03, 0.22] # Vital signs weight w_V_norm = [0.34, 0.35, 0.04, 0.03] # Environment weight (6) Output decision decision_score = w_Q_norm * Q_feat + w_K_norm * K_feat + w_V_norm * V_feat # Output: [0.15, 0.14, 0.89, 0.94] # Final risk score.

[0040] 0.94 > 0.9 red alert threshold, thus triggering highest level alarm.

[0041] Example Seven: The tiered response is illustrated by the following specific examples.

[0042] (1) LEVEL3 response: Start emergency dispatch For example: Vital sign radar detects apnea > 30 seconds.

[0043] (a) Warning trigger: AI algorithm module generates LEVEL3 warning; (b) Resource dispatch: Warning receiving unit of AI emergency platform activates service dispatch unit, notifies home medical service network; (c) Medical execution: Home medical service network dispatches ambulance for treatment according to patient location and electronic medical record information, and synchronizes user's real-time vital signs to cloud server at the same time; (d) Treatment feedback: After treatment, home medical service network feeds back treatment information to emergency platform and cloud server.

[0044] Emergency date: 2025-XX-XX XX-XX", Treatment result: success, Vital parameters: heart rate: 88, blood pressure: "130 / 85", Treatment prescription: nifedipine 5mg, oxygen inhalation.

[0045] (2) LEVEL2 response: Health management cycle generates intensive intervention plan For example: Nighttime out of bed > 2 hours, abnormal blood pressure trend, etc., trigger LEVEL2.

[0046] (a) Update user health baseline: Blood pressure > 160 / 100 mmHg for 3 consecutive times → mark "hypertension deterioration"; (b) Health management suggestion generation: low-sodium diet, exercise program; (c) Multi-terminal push: sent to user's mobile phone or TV terminal; (d) Effect tracking as shown in the following table; .

[0047] (3) LEVEL1 response: generate daily health report For example: reduce deep sleep duration, trigger regular intervention of health cycle, generate daily health report.

[0048] For example: Health report in XXXX, X (Li **) Core indicators | **Parameter** | **This month value** | **Trend** | **Normal range** | |----------------|------------|----------|--------------| | Night heart rate | 72bpm | ↑5% | 60-100bpm | | Deep sleep duration | 1.2h | ↓15% | >1.5h | (a) Intervention suggestions Sleep improvement: avoid strong light exposure after 22 o'clock (current bedroom light: 50 lux → recommended <30 lux).

[0049] Exercise plan: increase morning walking (activity this week: 3800 steps / day → target 5000 steps).

[0050] (b) Doctor's comments "Blood pressure fluctuations are related to decreased sleep quality, it is recommended to prioritize the above program, and reevaluate after two weeks."

[0051] Example eight: The double-cycle data feedback is illustrated by the following specific examples.

[0052] (1) Emergency data feedback to health cycle After LEVEL3 treatment is completed, key vital sign data (such as heart rate and blood pressure during emergency) are updated through smart contract to update the dynamic baseline (such as threshold) in the user's health baseline unit.

[0053] (a) Emergency stage The patient suddenly has chest pain, radar detects heart rate >150bpm (LEVEL3); The service scheduling unit assigns the order to the home medical service network; The emergency team arrives with a defibrillator in 8 minutes and successfully rescues.

[0054] (b) Health feedback User health baseline update: heart rate from 78 to 82.

[0055] (2) Health data feedback to emergency cycle Health intervention effects (e.g. chronic disease control rate) feedback to the early warning model.

[0056] For example: if the intervention efficiency is greater than 80%, reduce the LEVEL2 threshold by 5%; if the emergency event recurrence rate is greater than 15%, increase the LEVEL3 threshold by 3%.

[0057] For example: rehabilitation of myocardial infarction patients.

[0058] Health management: (a) Generate rehabilitation plan: Daily heart rate monitoring (warning value < 100bpm), Low-fat diet.

[0059] (b) Effect feedback: After two weeks, the resting heart rate is stable at 80-85bpm, The system automatically reduces the heart rate LEVEL3 threshold (150→145bpm).

[0060] (3) Cross-system data security exchange protocol of blockchain, blockchain smart contract.

[0061] (4) Regularly update the AI algorithm module according to the feedback data, and adjust the health suggestion generation dynamically.

[0062] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. An intelligent home maintenance system, characterized in that, The system includes a monitoring device located at the user end, which monitors the user's activity status, health data, and environmental data of the user's activity area in real time. The monitoring device wirelessly connects to a cloud server, which includes a data receiving and processing module, an AI algorithm module, a user data storage module, a risk triggering module, and a health management module. The data receiving and processing module receives data from the user-end monitoring device and preprocesses it. The AI ​​algorithm module analyzes the data processed by the data receiving and processing module based on a multimodal data fusion method and generates early warning signals. When the risk triggering module detects a risk event, it automatically triggers the health management module or an AI emergency rescue platform. The user data storage module stores the user's health and behavioral data for long-term analysis and health management. The health management module is used for the user's health management.

2. The intelligent home maintenance system according to claim 1, characterized in that, The monitoring equipment includes a fall detection radar, a vital signs detection radar, a temperature and humidity sensor, and a gas sensor. The activity status includes falling and getting out of bed. The health data includes respiratory data and heart rate data. The environmental data includes temperature and humidity data and gas data.

3. The intelligent home maintenance system according to claim 1, characterized in that, The monitoring device sends data to a cloud server via Wi-Fi or 4G / 5G networks.

4. The intelligent home maintenance system according to claim 1, characterized in that, The data receiving and processing module performs data cleaning and validity filtering on the received data.

5. The intelligent home maintenance system according to claim 1, characterized in that, The multimodal data fusion method includes the following steps: S100, Input multimodal data; S200. Perform time alignment processing on multimodal data to ensure that data from different modalities are synchronized in time; S300: Use a lightweight convolutional neural network to extract features from data of each modality, project the features extracted from different modalities onto the same feature space, and unify the dimensions; S400. Use point integral to calculate the attention score between different modal features, and perform Softmax normalization on the attention score to obtain the attention weight matrix. S500 combines modal confidence and reliability factors to assign real-time weights to the data of each modality, and normalizes the real-time weights to ensure that the sum of the weights is 1. S600. Use the normalized weights to perform a weighted summation of the features to obtain the decision score, and classify the risk level according to the decision score.

6. The intelligent home maintenance system according to claim 5, characterized in that, The decision score is compared with the risk level threshold to classify the risk level, which includes Level 1 Blue Alert, Level 2 Orange Alert, and Level 3 Red Alert.

7. The intelligent home maintenance system according to claim 6, characterized in that, When the warning signal is a Level 1 blue warning or a Level 2 orange warning, the health management module is triggered to form a health management cycle; when the warning signal is a Level 3 red warning, the AI ​​emergency rescue platform is triggered to form an emergency response cycle.

8. The intelligent home maintenance system according to claim 1, characterized in that, The health management module includes a data analysis unit, a user health baseline unit, a health suggestion generation unit, a multi-terminal push unit, an effect tracking unit, and a dynamic optimization unit. The user health baseline unit contains a set of various health data thresholds and health tags for the user. The data analysis unit performs in-depth analysis based on the user's historical health data and behavioral data, as well as the data contained in the user health baseline unit. The health advice generation unit provides personalized health management advice to users based on the analysis results and generates a health report for the user; the multi-terminal push unit is used to notify users of health management advice on their mobile phones or TVs; the effect tracking unit is used to track the effect of implementing health management advice; and the dynamic optimization unit adjusts various threshold data of the user's health baseline based on the service tracking data of the AI ​​emergency rescue platform.

9. The intelligent home maintenance system according to claim 1, characterized in that, The AI ​​emergency rescue platform includes an early warning receiving unit, a service dispatching unit, and a service tracking unit. The early warning receiving unit is used to receive risk warning signals from a cloud server. The service dispatching unit quickly dispatches the nearest service personnel to the user's home based on the early warning information. The service tracking unit tracks the response status and service progress of the service personnel in real time to ensure that the service is completed in a timely manner.

10. The intelligent home maintenance system according to claim 1, characterized in that, It also includes a home-based medical care service network, which connects medical units nationwide. After receiving the dispatch instructions from the AI ​​emergency rescue platform, the home-based medical care service network responds quickly, makes service arrangements, and feeds back the service results to the emergency rescue platform and cloud server.

Images