Multi-mode home-based care health monitoring method and system based on voice interaction suitable for aging

Abstract

The invention discloses a multi-mode home-based care health monitoring method and system based on voice interaction suitable for aging, and belongs to the technical field of smart care and health monitoring. According to the system, physiological data of an intelligent ring (M1), behavior data of a home environment sensor (M2) and equipment state monitoring information are fused, wherein the behavior data comprise a water / electricity / gas meter and an emergency pager which are wirelessly transmitted by zigbee, a button type noctilucent coating is larger than or equal to 0.5 mm, and buzzing is 60-70 dB after triggering. And a health management closed loop without APP operation is realized through aging-suitable voice interaction (supporting more than eight dialects and wake-up word custom). The core method comprises multi-source data acquisition, six-dimensional health assessment, multi-dimensional risk grading and grading service response, the problems of single data, insufficient aging adaptability, service disjunction and the like of existing equipment are solved, the high-risk identification accuracy rate reaches 93%, the emergency call response time is shorter than 90 seconds, and the method is suitable for home-based care scenes of old people living alone or both, and has a wide application prospect. And the monitoring safety and convenience are obviously improved.

Description

Technical Field

[0003] This invention belongs to the field of smart elderly care and health monitoring technology. Specifically, it relates to a method and system that integrates data from wearable devices (smart ring (M1)) and home environment sensors (M2), and enables health status assessment and risk warning for the elderly without the need for an app through age-friendly voice interaction. It also links with a grid worker app (M7) to achieve a service loop. This method and system are suitable for 24 / 7 home-based elderly care scenarios for elderly people living alone or in pairs. Background Technology

[0004] Existing home-based elderly care monitoring technologies have the following core shortcomings:

[0005] 1. Limited Data Dimensions: Most systems rely on a single sensor (such as heart rate monitoring in smart bracelets), failing to comprehensively assess health status. For example, the Huawei Smart Selection Elderly Care Bracelet only monitors heart rate via PPG, without combining it with ECG analysis of cardiac electrical activity, making it difficult to identify early myocardial infarction risk;

[0006] 2. Disconnect between environmental and health data: Human body sensors (such as the Xiaomi Human Body Sensor (M2-1)) only record activity trajectories, and smart water meters (M2-2) / electricity meters (M2-3) only monitor usage, without linking to health data. For example, a sudden decrease in water consumption by an elderly person may indicate mobility issues, but the existing system cannot link it to abnormal heart rate monitored by PPG.

[0007] 3. Disconnect between early warning and service: The monitoring system can only issue alarms (such as fall alarms), but lacks a closed-loop response mechanism with community service stations (M5). For example, although the Quick Angel elderly care system can locate the location of the elderly, it does not achieve automatic matching of "risk level (R1 / R2 / R3) → service priority";

[0008] 4. Static evaluation model: The model uses a fixed threshold (e.g., a heart rate > 100 beats / min triggers an alert) and does not consider the individualized baseline of the elderly (e.g., the "normal heart rate" of an elderly person with long-term hypotension may be less than 60 beats / min), resulting in a false alarm rate > 30%.

[0009] 5. Lack of age-friendly interaction: Most systems require the elderly to use an app to complete tasks such as marking meals and providing feedback on needs. Due to the elderly's insufficient digital skills, the actual usage rate is less than 30%, making it difficult to effectively realize the monitoring value.

[0010] 6. Insufficient device status monitoring: The lack of real-time monitoring of the battery level and data upload status of wearable devices (such as smart rings) makes it impossible to intervene in a timely manner when the device goes offline;

[0011] 7. Delayed emergency call response: Most existing emergency call devices are fixed wired installations, which have poor deployment flexibility and complicated triggering methods (such as requiring a long press for more than 5 seconds), which are not suitable for the emergency needs of the elderly. Summary of the Invention

[0012] (I) Purpose of the Invention

[0013] To address the shortcomings of existing technologies, this paper provides a multimodal home-based elderly care health monitoring method and system with age-friendly voice interaction. Through the fusion analysis of physiological data from a smart ring (M1) and behavioral data from the home environment (M2), combined with a dynamic baseline model and age-friendly voice interaction, it achieves a closed-loop management system encompassing "health assessment - risk warning - proactive services," specifically including:

[0014] • Senior users: No need to operate the app; users can complete interactions such as marking meals, making requests, and confirming abnormalities through voice dialogue or an emergency call button.

[0015] • Device side: The smart ring (M1) has dedicated buttons (for ECG testing only) and provides voice prompts via a 4G voice gateway (M6); a new emergency call button with Zigbee wireless transmission has been added to enhance emergency response; device status monitoring covers battery level and data upload, triggering intervention in case of abnormalities;

[0016] • Server-side: Grid workers receive orders and provide service feedback through the APP (M7), forming a complete closed loop of "monitoring-early warning-service".

[0017] (II) Technical Solution

[0018] 1. Overview of Core Technology Solutions

[0019] This invention achieves multimodal monitoring through **three-layer data fusion + age-friendly interaction + device status monitoring + enhanced emergency call functionality**.

[0020] Physiological data layer: The smart ring (M1) collects data such as PPG, acceleration, and temperature (ECG requires the user to actively press a button to collect data) to construct a 6-dimensional health age assessment system (H1-H6);

[0021] • Environmental Behavior Layer: Home sensors (M2) (human body sensor (M2-1), water / electricity / gas meter (M2-2 / M2-3 / M2-4, Zigbee wireless transmission), emergency call button (M2-5, Zigbee wireless transmission)) record daily routines and emergency needs, and identify abnormal behavior patterns;

[0022] • Device Status Layer: Real-time monitoring of the smart ring (M1) battery level and data upload status; in case of abnormality, voice prompts and service triggers are provided. • Service Response Layer: Based on the results of fusion analysis, a tiered response (passive door-to-door service / active intervention / emergency rescue) is achieved through voice interaction, emergency call button and grid member APP (M7).

[0023] 2. Monitoring methods (see attached flowchart)Figure 1 )

[0024] Step S1: Multi-source data acquisition and preprocessing

[0025] • Data acquisition from the S1.1 smart ring (M1):

[0026] o Timed data collection: PPG (red light + infrared, 100Hz), skin temperature (±0.1℃), and triaxial acceleration (±8g) are collected every 5 minutes; only PPG data are collected at 3:00 AM every day (ECG is not collected to avoid disturbing sleep);

[0027] ο Triggering data collection:

[0028] • Post-meal PPG: When an elderly person says "I have eaten breakfast", "I have eaten lunch", or "I have eaten dinner" to the 4G voice gateway (M6), the gateway (M6) confirms the message via voice and triggers high-frequency PPG collection 2 hours after the meal (once every 5 minutes for 1 hour).

[0029] • ECG detection: Only when PPG detects a heart rate fluctuation >20 beats / min, the 4G voice gateway (M6) will prompt "Please press the ring button to collect ECG". The user will actively press the ring (M1) button (M1-4) to collect ECG (single lead, 10 seconds). Button (M1-4) is only used for this function.

[0030] Preprocessing: Motion artifacts are filtered using da262 acceleration data (Formula 1: Corrected PPG waveform = Original waveform - 0.3 × 10 ...

[0031] Acceleration amplitude); Nighttime EDA data is temperature compensated (Formula 2: Compensated resistance = Original resistance × (37℃ / Real-time temperature)).

[0032] • S1.2 Home Environment (M2) Data Collection:

[0033] ο Human body sensor (M2-1) (1 per room): Records activity frequency (e.g., number of daily round trips between bedroom and living room) and dwell time (staying in a room for more than 2 hours);

[0034] ο Smart water meter (M2-2): Pulse type (accuracy ±1%), uses Zigbee wireless transmission, records usage every hour, and calculates daily average fluctuations (e.g., water consumption decreases by more than 50% compared to the previous day);

[0035] Smart meter (M2-3): Uses Zigbee wireless transmission to record usage every hour and calculate daily average fluctuations;

[0036] Smart gas meter (M2-4): Internet of Things meter (leakage alarm threshold > 0.1% LEL), using Zigbee wireless transmission, records usage and leakage status every hour;

[0037] Emergency call device (M2-5):

[0038] • Button type: Deployed at the bedside or in the living room, using Zigbee wireless transmission, a 1-second press triggers an emergency help signal, the button has a luminous coating thickness (≥0.5mm), and a buzzer sound (60-70dB) after triggering;

[0039] • Pull-cord type: Deployed in the toilet, using Zigbee wireless transmission, pulling for 1 second triggers an emergency help signal, followed by a buzzer (60-70dB);

[0040] o Data alignment: Physiological data, environmental data and emergency call signals are linked by timestamps (error < 1 second) (e.g., "emergency call to the toilet + sudden increase in heart rate").

[0041] • S1.3 Equipment Status Monitoring:

[0042] The cloud platform (M4) monitors the smart ring's (M1) battery level in real time (below 20% is considered insufficient) and data upload status (exceeding the limit).

[0043] (If no new data is collected after 6 hours, the session is considered interrupted).

[0044] If the above conditions are met, the 4G voice gateway (M6) will prompt "Your ring is low on power / has not uploaded data, please charge it" once a day. If the prompt is not confirmed after 3 consecutive prompts (the user does not respond by pressing the gateway button), it will be judged as a device malfunction and included in the environmental malfunction indicators.

[0045] • S1.4 Grid Member APP (M7) Data Interaction: The cloud platform (M4) synchronizes elderly health, environmental data, equipment status, voice interaction records and emergency call signals to the grid member APP (M7) to provide a basis for risk assessment and service dispatch.

[0046] Step S2: Multimodal Feature Extraction and Health Assessment

[0047] • S2.16-dimensional health age feature extraction (with appendix) Figure 2 (Evaluation system diagram)

[0048] Heart age (H1): Based on ECG data actively collected by the user, it is compared with the historical baseline (resting ECG data collected by the user on a regular basis as prompted) to calculate the premature beat frequency (number of abnormalities / total number of collections);

[0049] ο Vascular Age (H2): PWV is calculated by the time difference (ΔT) between the R wave of ECG (user actively acquired) and the peak value of PPG (Formula 3: PWV = 0.2m / ΔT, where 0.2m is the distance from finger to heart), after filtering for motion artifacts;

[0050] Metabolic age (H3): PPG fluctuation range 2 hours after meal (Formula 4: Fluctuation range = (postprandial peak - preprandial baseline) / preprandial baseline × 100%);

[0051] οNeural Age (H4): Gateway (M6) timed (e.g., 9:00, 15:00, 19:00) voice test response time for “Please say health” (average of 3 consecutive times) + nighttime EDA fluctuation (excluding temperature influence);

[0052] οImmune age (H5): The time it takes for HRV to return to baseline when body temperature is >37.3℃ (health threshold <48 hours);

[0053] ο Musculoskeletal age (H6): Gait symmetry of triaxial acceleration (left-right amplitude difference <20%) + ring (M1) rotation angle (0-360°) to assess joint range of motion.

[0054] • S2.2 Health Assessment Model: Each dimension uses a dynamic baseline score (Formula 5):

[0055] \[

[0056] Dimensional score = 100 - |current value |- 21-day baseline mean | baseline standard deviation 40

[0057] \]

[0058] (Score range 0-100 points, <60 points is the risk warning threshold).

[0059] •S2.3 Assessment Results Synchronization: The cloud platform (M4) synchronizes the 6-dimensional health age score and dynamic baseline score to the grid worker APP (M7) to help predict the health risks of the elderly.

[0060] Step S3: Multi-dimensional Risk Integration and Classification

[0061] • S3.1 Health Risk Identification:

[0062] a single dimension score <60 points (e.g., cardiac age (H1) 55 points);

[0063] ο Abnormal associated environmental data (e.g., abnormal heart age (H1) + bedroom stay > 2 hours + abnormal equipment status);

[0064] The emergency call button (M2-5) trigger signal is directly identified as a high-risk associated item.

[0065] • S3.2 Voice Confirmation Process: When an abnormal activity pattern is detected (such as delayed wake-up), the gateway (M6) will sequentially say "Hello! Are you awake? Please answer" (5 minutes apart, 3 times in total). If there is no response, the environmental abnormality indicator will be increased by 2. If there is a response but the abnormality is confirmed, the indicator will be increased by 1.

[0066] • S3.3 Risk Level (R) Classification:

[0067] Low risk (R1): 60-70 points in one dimension, with no abnormal environmental data (including normal equipment status);

[0068] Medium risk (R2): 1 dimension score <60 points + 1 environmental anomaly (including minor anomalies in voice confirmation and equipment status anomalies),

[0069] Or ≥2 dimensions with scores <60 points + 1 environmental anomaly;

[0070] High risk (R3): ≥2 dimensions score <60 points + ≥2 environmental anomalies (including 3 unresponsive calls, equipment malfunction lasting 3 days), **or emergency call button (M2-5) trigger signal**.

[0071] •S3.4 Risk Level Result Synchronization: The cloud platform (M4) synchronizes the risk level (R1 / R2 / R3) and the judgment criteria to the grid worker's APP (M7) to facilitate early intervention.

[0072] Step S4: Service Response Closed Loop

[0073] Passive service:

[0074] When an elderly person presses the large button (M6-2) on the 4G voice gateway (M6), the gateway (M6) will prompt with a voice message, "Hello! What do you need? You can say 'door-to-door service' or 'health consultation'." After recognizing the need, the message will be sent to the grid worker's APP (M7), and the community service station (M5) will arrive at the door within 15 minutes.

[0075] Once the emergency call button (M2-5) is triggered, the cloud platform (M4) directly pushes a high-risk alert to the grid worker's APP (M7), and the community service station (M5) initiates the rescue chain within 90 seconds.

[0076] • Proactive service:

[0077] Low risk (R1): The 4G voice gateway (M6) broadcasts health advice (such as "insufficient walking today, it is recommended to exercise for 10 minutes"), which is synchronized to the grid worker APP (M7);

[0078] Medium risk (R2): Grid workers receive orders via APP (M7) and follow up by phone or visit the site within 1 hour (if the equipment is faulty, bring the charging equipment);

[0079] High risk (R3): The service station (M5) initiates the rescue chain within 90 seconds (contacting the emergency contact person + grid member arriving within 15 minutes + coordinating with the nearest community hospital), and the progress is updated in real time to the APP (M7).

[0080] Step S5: Dynamic Updates and Model Optimization

[0081] • A 21-day dynamic baseline is updated daily at 2:00 AM and synchronized to the grid worker's APP (M7);

[0082] • For every 100 high-risk event data points collected (including voice confirmation, abnormal equipment status, emergency call records, and grid worker service records), the risk classification model is retrained using the random forest algorithm (input features: 6-dimensional score + environmental anomaly indicators + equipment status + voice interaction results + emergency call signals), and after optimization, it is synchronized to the cloud (M4) and APP (M7).

[0083] 3. Monitoring System (See attached architecture diagram) Figure 3 )

[0084] • Smart ring terminal (M1):

[0085] o Sensor module (M1-1): PPG (660nm / 940nm dual wavelength), ECG electrode (Ag / AgCl), triaxial accelerometer, temperature sensor, EDA sensor, power sensor;

[0086] Control module (M1-2): ARM Cortex-M4 core, supporting edge computing (motion artifact filtering, preliminary feature extraction);

[0087] ο Communication module (M1-3): BLE 5.0, communicates with the 4G voice gateway (M6) (encryption method AES-128), and periodically uploads data and power information;

[0088] ο button (M1-4): Only used to trigger ECG acquisition, no other functions.

[0089] • 4G voice gateway (M6):

[0090] ο Voice interaction module (M6-1): Supports Chinese dialect recognition (accuracy ≥92%, dialect library coverage includes ≥8 languages ​​such as Cantonese and Sichuanese), TTS broadcast (medium-speed clear female voice), used to prompt ECG acquisition, device charging and abnormal confirmation, and has a customizable voice wake-up word function;

[0091] ο Large physical button (M6-2): Diameter ≥2cm, short press triggers voice dialogue, long press for 5 seconds triggers emergency call, supports user response to device status prompts;

[0092] o Data aggregation unit (M6-3): Receives data from the ring (M1), environmental sensor (M2) and emergency call device (M2-5), and aligns them by timestamp;

[0093] Local warning unit (M6-4): Triggers audible and visual alarm (85dB indoors) when high risk (R3).

[0094] • Home environment sensor network (M2):

[0095] ο Human body sensor (M2-1) (infrared + microwave dual detection);

[0096] Smart water meter (M2-2): Pulse type (accuracy ±1%), Zigbee wireless transmission;

[0097] Smart meter (M2-3): Zigbee wireless transmission;

[0098] Smart gas meter (M2-4): Internet of Things meter (leakage alarm threshold > 0.1% LEL), Zigbee wireless transmission;

[0099] ο Emergency call button (M2-5): Includes bedside / living room button type and toilet pull cord type, Zigbee wireless transmission, press and hold for 1 second to trigger emergency call, button type has a luminous coating thickness (≥0.5mm), and all have a buzzer prompt (60-70dB) after triggering.

[0100] • Cloud platform (M4):

[0101] ο Health Assessment Engine (M4-1): Calculates 6-dimensional scores and dynamic baselines, and synchronizes them to the grid worker's APP (M7);

[0102] Risk Decision Engine (M4-2): Integrates multimodal data (including equipment status and emergency call signals) to output risk levels (R1 / R2 / R3), and the model is iterated regularly;

[0103] Service dispatch engine (M4-3): Pushes dispatch information (including location, risk details, equipment status, voice recording, and emergency call signal) to the grid worker APP (M7);

[0104] o Device Status Monitoring Engine (M4-4): Real-time monitoring of ring (M1) power consumption and data upload, triggering voice prompts from the gateway (M6).

[0105] • Grid worker APP (M7):

[0106] ο Dispatch Receiving Module (M7-1): Real-time acquisition of dispatch information, device status, emergency call records, and historical data;

[0107] o Service Feedback Module (M7-2): Upload on-site arrival time and service content (such as equipment charging, health check status, and emergency rescue results);

[0108] ο Health Record Module (M7-3): Stores the elderly’s historical assessment results, service records and emergency call records.

[0109] Community service terminal (M5):

[0110] o Service station large screen (M5-1): Displays the real-time risk level (R1 / R2 / R3) of the elderly in the jurisdiction, equipment status (normal / abnormal), emergency call alarm, and grid worker service progress;

[0111] It links with the grid worker's APP (M7) to assist in overall planning and scheduling. Attached Figure Description

[0112] This invention includes four key figures to fully present the logical flow, evaluation system, architectural composition, and risk decision-making mechanism of a multimodal home-based elderly care health monitoring method and system with age-friendly voice interaction. The descriptions of each figure are as follows:

[0113] Figure 1 Flowchart of Age-Friendly Voice Interaction and Grid Worker Collaboration

[0114] Function Description:

[0115] It demonstrates the entire process logic from multi-parameter data acquisition (P1) to service response closed loop (F).

[0116] • Data acquisition layer (S1-1~S1-5): Covers 5 scenarios including smart ring automatic data acquisition (limited PPG acquisition in the early morning), ECG triggered data acquisition (PPG fluctuation linked to voice prompts), meal voice markers (linked to post-meal health monitoring), device status monitoring (battery / data interruption warning), and emergency call (one-click triggering of Zigbee transmission, including a 60-70dB buzzer prompt).

[0117] • Data Processing Layer (B): Enables multi-source data fusion preprocessing through timestamp alignment and noise filtering;

[0118] • Health assessment layer (S2): Calculates dynamic baseline scores based on 6-dimensional health age characteristics (H1~H6);

[0119] • Risk grading layers (R1~R3): Based on dimensional scores (60-point threshold) and the number of outliers, low / medium / high risk are divided. Emergency calls directly trigger high risk (R3).

[0120] • Service Response Layer (S4): Links with the grid worker's APP to achieve "risk level → service action" matching (voice suggestions, door-to-door intervention, rescue chain activation), and completes dynamic baseline updates through feedback.

[0121] Cross-graph association: with Figure 3 (System Architecture) Data Flow (M1→M6→M4→M7) Figure 4 The risk logic's tiered conditions (I5 emergency call trigger R3) are fully coordinated, unifying the references for multiple stages.

[0122] Figure 2 6-Dimensional Health Age Assessment System Diagram

[0123] Function Description:

[0124] We break down the core dimensions and calculation logic of health status assessment to provide a quantitative basis for risk classification.

[0125] • 6-dimensional feature layers (H1~H6):

[0126] οH1 (Heart Age): Premature beat frequency is calculated by actively collecting ECG data and comparing it with historical baseline.

[0127] οH2 (vascular age): The time difference between the ECG R wave and PPG peak values, combined with a motion artifact filtering algorithm to calculate pulse wave velocity (PWV);

[0128] οH3 (metabolic age): Analysis of postprandial PPG fluctuations to assess metabolic levels;

[0129] οH4 (Neural Age): A dual-dimensional assessment of gateway voice response time (supporting more than 8 dialects) and nighttime EDA fluctuations;

[0130] οH5 (Immune Age): Analysis of Heart Rate Variability (HRV) Recovery Time in Abnormal Body Temperature Scenarios;

[0131] οH6 (musculoskeletal age): A fusion assessment of gait symmetry and joint range of motion (monitored by ring rotation angle);

[0132] • Dynamic baseline layer (H7): Generates personalized health thresholds (different from fixed standards) through a 6-dimensional data fusion algorithm;

[0133] • Application Layer (M7): Evaluation results are synchronized with the grid worker's app, providing support Figure 1 Risk classification, Figure 4 Decision-making logic.

[0134] Cross-graph association: H1~H6 as Figure 1 (S2) Figure 4 (I1) is the core input, enabling a quantitative connection between "feature extraction and risk assessment".

[0135] Figure 3 Monitoring System Architecture Diagram

[0136] Function Description:

[0137] It presents the hierarchical relationship and data interaction path between hardware devices, cloud platforms, and service terminals, and clarifies the functional boundaries of each module.

[0138] Terminal layer (M1~M2):

[0139] οM1 (Smart Ring): Integrates multiple sensors (M1-1), edge computing (M1-2), BLE communication (M1-3), and a dedicated ECG button (M1-4);

[0140] οM2 (Home Sensor): Covers human body sensors (M2-1), water, electricity and gas meters (M2-2~M2-4, Zigbee wireless transmission), and emergency call devices (M2-5, button-type luminous coating ≥0.5mm, buzzer 60-70dB after triggering).

[0141] Gateway layer (M6): Enables age-friendly operation through voice interaction (M6-1, supporting ≥8 dialects + custom wake words) and physical buttons (M6-2), aggregates multi-source data (M6-3) and triggers local alerts (M6-4);

[0142] • Cloud Layer (M4): Deploys four major engines: health assessment (M4-1), risk decision-making (M4-2), service scheduling (M4-3), and device status monitoring (M4-4), providing algorithmic support for hierarchical decision-making;

[0143] • Service layer (M5~M7): The grid worker APP (M7) receives dispatch orders and provides service feedback, while the community terminal (M5) displays risks on a large screen, forming a "monitoring-disposal" closed loop.

[0144] Cross-graph association: with Figure 1 Data flow (M1→M6→M4→M7) Figure 4 The input items (I2~I5 correspond to M2~M6 data) are completely matched, unifying the device and module references.

[0145] Figure 4 Risk classification and equipment status logic diagram

[0146] Function Description:

[0147] Risk assessment rules are presented in the form of a decision tree, clearly defining the mapping relationship between "input conditions → risk level → response action".

[0148] ·**Input layer (I1~I5)**: Corresponding to Figure 1 Data acquisition (I1= Figure 2 H1~H6, I5= Figure 3 M2-5 includes a buzzer alert) Figure 3 Device status (I3= Figure 3 M1 battery level / upload);

[0149] ·**Judgment Layer (Q1~Q4)**: Based on three levels of criteria—emergency call (Q1), dimensional score (Q2), and number of anomalies (Q3~Q4)—low (R1), medium (R2_light / R2_heavy) and high (R3) risks are accurately classified.

[0150] • Response layer (R1~R3): Association Figure 1 Service Response (S4) enables differentiated handling of "low-risk voice suggestions (supporting dialects), medium-risk tiered intervention, and high-risk activation of the rescue chain".

[0151] Cross-graph association: with Figure 1 Risk classification (R1 to R3 determination criteria) Figure 3 The equipment references (M1 to M2 data inputs) are fully coordinated to ensure a closed-loop decision-making logic.

[0152] Synergistic Value of the Attached Figures:

[0153] 4 attached diagrams from the process ( Figure 1 ),Evaluate( Figure 2 ), architecture ( Figure 3 ),decision making( Figure 4 It covers the entire chain of "data collection → health assessment → risk classification → service response" in four dimensions. By unifying device / module labels, data flow, and judgment conditions, it achieves cross-graph logic self-consistency and clearly presents the core innovations of this invention: "multimodal fusion, age-friendly interaction (including dialect and wake word customization), and emergency call enhancement (including night light and buzzer)". Detailed Implementation

[0154] Example 1: Smart Ring Low Battery Intervention Procedure

[0155] 1. Equipment status monitoring:

[0156] The cloud platform (M4) detected that the smart ring (M1) had 15% battery (<20%) and no new data had been uploaded for 6 hours, triggering the 4G voice gateway (M6) to issue a voice prompt: "Your ring is low on battery, please charge it" (supports Sichuan dialect broadcast).

[0157] 2. Confirmation Process:

[0158] If the user does not respond by pressing the gateway (M6) button after being prompted once a day for three consecutive days, it is determined to be a device malfunction and included in the environmental malfunction indicators.

[0159] 3. Risk Classification:

[0160] Based on the elderly person's metabolic age (H3) score of 62 (60-70 points), the risk level was initially assessed as low (R1) → but due to equipment malfunction, the risk level was upgraded to medium (R2).

[0161] 4. Service Response:

[0162] The grid worker receives the order (including the "device malfunction" label) through the APP (M7), arrives at the door within 1 hour, charges the ring and checks the device's function, and uploads the service record to the APP (M7).

[0163] Example 2: Early warning process for the risk of myocardial infarction in elderly people living alone

[0164] 1. Data Acquisition:

[0165] • The smart ring (M1) collects PPG every 5 minutes. When it detects a sudden increase in heart rate from 65 beats / min to 120 beats / min (lasting for 10 seconds), the 4G voice gateway (M6) will prompt "Please press the ring button to collect ECG" (Cantonese broadcast). The user presses the button (M1-4) to collect ECG.

[0166] • The human body sensor (M2-1) showed that the elderly person had not left the bedroom for two consecutive hours, and the smart water meter (M2-2) showed that the water consumption for the day was 60% lower than the previous day;

[0167] • The 4G voice gateway (M6) will call out "Hello! Are you awake? Please answer" three times (5 minutes apart). If there is no response, the environmental anomaly indicator will increase by 2.

[0168] 2. Feature extraction and evaluation:

[0169] • Heart age (H1): ST segment elevation was detected on ECG (0.2mV). Compared with the baseline of 3 months, the premature beat frequency was 5 times / 10 minutes, and the score was 45 points (<60 points).

[0170] • Vascular age (H2): PWV = 12 m / s (baseline 8 m / s), score 52 (<60).

[0171] 3. Risk Classification:

[0172] If the criteria of "≥2 dimensions <60 points + 2 environmental anomalies (including no response)" are met, the risk level is determined to be high (R3).

[0173] 4. Service Response:

[0174] • The cloud platform (M4) pushes an alert to the grid worker's APP (M7), including location (bedroom), risk details (suspected myocardial infarction), and voice recording;

[0175] • After receiving the dispatch order, the grid worker will arrive at the door within 12 minutes and simultaneously call the community hospital ambulance. The service record will be fed back through the APP (M7).

[0176] Example 3: Metabolic Function Assessment of Elderly Couples Living Together

[0177] 1. Data Acquisition:

[0178] • When an elderly person says “I have breakfast” to the 4G voice gateway (M6) (triggered by the custom wake-up word “Xiao Hu Reminder”), the gateway (M6) confirms with voice “Breakfast time has been recorded”, triggering PPG high-frequency collection 2 hours after the meal, and calculating the PI fluctuation amplitude of 18% (>15%) 1 hour after the meal;

[0179] • The smart meter (M2-3) shows stable daily activity levels (6-8 AM, 4-6 PM) with no environmental anomalies;

[0180] The smart ring (M1) has 80% battery and data upload is normal (no abnormalities in device status).

[0181] 2. Feature extraction and evaluation:

[0182] • Metabolic age (H3) score: 58 (<60), all other dimensions (H1 / H2 / H4 / H5 / H6) score: >70.

[0183] 3. Risk Classification:

[0184] One dimension anomaly + 0 environmental / equipment anomalies are judged as low risk (R1).

[0185] 4. Service Response:

[0186] • The 4G voice gateway (M6) broadcasts: "Your post-meal metabolism fluctuates greatly. It is recommended to reduce refined carbohydrates and increase protein intake" (broadcast in Hokkien).

[0187] • The grid worker's APP (M7) simultaneously receives health advice. One week later, the system automatically retests the results. The post-meal PI fluctuation drops to 12%, and the metabolic age (H3) score increases to 65 points. The results are synchronized to the APP (M7).

[0188] Example 4: Emergency Assistance Procedure for Elderly People Living Alone

[0189] 1. Data Acquisition:

[0190] An elderly person suddenly felt unwell in the toilet and pulled the pull-cord emergency call button (M2-5). Pulling it for 1 second triggered a signal, and the device emitted a 65dB buzzer to alert the user. The signal was then transmitted via Zigbee to the 4G voice gateway (M6).

[0191] 2. Risk Classification:

[0192] The cloud platform (M4) receives an emergency call signal and immediately identifies it as high-risk (R3);

[0193] 3. Service Response:

[0194] When a grid worker receives an emergency dispatch order containing location information (toilet) via the grid worker's APP (M7), the community service station simultaneously contacts the emergency contact person, and the grid worker arrives at the scene to provide assistance within 8 minutes. Service records are uploaded to the APP in real time.

Claims

1. A multi-modal home health monitoring system for age-friendly voice interaction, characterized in that, include: • Smart Ring Terminal (M1): A sensor module (M1-1) integrating PPG (660nm / 940nm dual wavelength), ECG electrodes (Ag / AgCl), a triaxial accelerometer, a temperature sensor, and a power sensor; a control module (M1-2) using an ARM Cortex-M4 core to achieve edge computing; a communication module (M1-3) that communicates with a 4G voice gateway (M6) via BLE 5.0 with AES-128 encryption and periodically uploads data and power information; and a dedicated ECG button (M1-4) used only to trigger ECG acquisition. When the PPG detects a heart rate fluctuation >20 beats / min, the 4G voice gateway (M6) will provide a voice prompt, and the elderly person will press the dedicated ECG button (M1-4) to collect 10 seconds of single-lead ECG data (250Hz). • 4G Voice Gateway (M6): Supports Chinese dialect recognition (accuracy ≥92%, dialect library covers ≥8 languages ​​including Cantonese and Sichuanese), TTS broadcast (medium-speed clear female voice), and has a customizable voice wake-up word function. It is used for voice interaction module (M6-1) for ECG collection prompts, device charging reminders, and abnormal confirmation; a large physical button (M6-2) with a diameter ≥2cm, which responds to voice prompts with a short press and triggers an emergency call with a long press for 5 seconds; a data aggregation unit (M6-3) that receives data from the smart ring (M1) and home environment sensor (M2) and aligns the data through timestamps (error <1 second); and a local early warning unit (M6-4) that triggers an indoor 85dB audible and visual alarm when there is a high risk (R3). • Home Environment Sensor Network (M2): One human body sensor (M2-1) is installed in each room, using infrared + microwave dual detection technology to record the frequency of elderly people's activities and the time they stay; pulse-type (accuracy ±1%), Zigbee wireless transmission smart water meter (M2-2); Zigbee wireless transmission smart electricity meter (M2-3); IoT meter (leakage alarm threshold > 0.1% LEL), Zigbee wireless transmission smart gas meter (M2-4); deployed at the bedside (button type), living room (button type), and toilet (pull-cord type), using Zigbee wireless transmission, a 1-second long press triggers an emergency help signal, the button type button has a luminous coating thickness ≥ 0.5mm, and all emit a 60-70dB buzzer prompt after triggering (M2-5). • Cloud Platform (M4): Based on data from the smart ring (M1), a dynamic baseline scoring model (dimension score = 100 - |current value - 21-day baseline mean| / baseline standard deviation × 40) is used to calculate a 6-dimensional health age score, which is then synchronized to the health assessment engine (M4-1) of the grid worker APP (M7). A risk decision engine (M4-2) integrates 6-dimensional health age scores, abnormal environmental data, device status, voice interaction records, and emergency call signals to classify risk levels (R1 / R2 / R3) and periodically trains models using the random forest algorithm; a service dispatch engine (M4-3) pushes dispatch information to the grid worker APP (M7) based on the risk level, including the elderly's location and risk details; and a device status monitoring engine (M4-4) monitors the smart ring's (M1) battery level (below 20% is considered insufficient) and data upload status (no new data for more than 6 hours is considered interrupted), triggering voice prompts from the 4G voice gateway (M6) when abnormalities occur. • Grid worker APP (M7): Real-time receiving of dispatch information pushed by the cloud platform (M4), including the elderly's risk level, device status, emergency call records, etc. (M7-1); Service feedback module (M7-2) after the grid worker provides door-to-door service, uploading service content, door-to-door service time, and health check results. A health record module (M7-3) that stores the elderly’s historical health assessment results, device status records, emergency call records, and service feedback information. • Community Service Terminal (M5): A service station screen (M5-1) that displays the real-time risk level (R1 / R2 / R3) of elderly people in the jurisdiction, device status (normal / abnormal), emergency call alarm, and grid worker service progress.

2. The system of claim 1, wherein, The control module (M1-2) of the smart ring terminal (M1) uses acceleration data to filter motion artifacts. The formula is: Corrected PPG waveform = Original waveform - 0.3 × Acceleration amplitude.

3. The system according to claim 1, characterized in that, The specific logic of the risk decision engine (M4-2) of the cloud platform (M4) in classifying risk levels is as follows: • Low risk (R1): A health age score of 60-70 points in one dimension, with no environmental abnormalities (normal activity of human body sensor, normal water, electricity and gas meter usage), and normal equipment status (battery level ≥20%, normal data upload). Health advice is given by voice through the 4G voice gateway (M6), and the grid member APP (M7) provides simultaneous monitoring. • Medium risk (R2): If a candidate's health age score is less than 60 points in any of the 1 dimensions, and there are no environmental abnormalities or equipment in normal condition, the grid worker will dispatch an order via the APP (M7) and follow up by phone. If a person's health age score is less than 60 points in any of the three dimensions, and there is one environmental anomaly (such as a human body sensor detecting that the person has stayed in a room for more than 2 hours) or an equipment malfunction (battery level less than 20% or data upload interruption and unconfirmed), a grid worker will provide door-to-door service within 1 hour. If the equipment malfunction is due to an abnormality, the worker must bring a charging device. • High risk (R3): Health age score <60 points in ≥2 dimensions, and ≥2 environmental abnormalities (such as human body sensor detecting 2 hours of stay in the bedroom, water consumption decrease of 50%) or equipment status abnormalities (battery power <20% and data upload interruption for 3 days); or receive an emergency call signal triggered by the emergency call button (M2-5), start the rescue chain, contact the elderly's relatives, the grid member arrives at the door within 15 minutes, and coordinate with the nearest community hospital. When the emergency call is triggered, the rescue chain start time is shortened to 90 seconds, and the grid member APP (M7) updates the rescue progress in real time.

4. The system according to claim 1, characterized in that, When the voice interaction module (M6-1) of the 4G voice gateway (M6) detects that the smart ring (M1) has insufficient power or interrupted data upload, it will give a voice prompt once a day (supporting dialect broadcast). If there is no confirmation after 3 consecutive prompts (the elderly do not respond by pressing the large physical button), it will be judged as a device malfunction and included in the risk judgment criteria of the risk decision engine (M4-2).

Images