Home-based care privacy protection type activity detection method based on infrared and microwave radar cooperation

Abstract

The invention belongs to the field of home-based care intelligent monitoring, aims to solve the problems of privacy disclosure, low collaboration efficiency, single monitoring and the like in the prior art, and provides a wearable and non-imaging activity detection method. A passive infrared sensor and a 24GHz non-imaging microwave radar are adopted to construct a cooperative system, an edge calculation and privacy enhancement module is matched, and through three-level hierarchical scheduling of infrared pre-triggering and microwave radar verification and scenarized dynamic weight fusion, the anti-interference capability and low-power-consumption balance are improved; triple privacy protection of non-imaging acquisition of a signal layer, local calculation of a processing layer and encryption and desensitization of a transmission layer is constructed, and privacy leakage is completely eradicated; six types of activity states are identified through multi-feature joint, a rule baseline is constructed in combination with 7-day data, and'real-time + rule deviation 'two-dimensional alarm is realized. The method does not need the cooperation of the elderly, reduces the false alarm rate by more than 40%, adapts to various home scenes and the requirements of the elderly, and is high in practicability and easy to popularize.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring technology for home-based elderly care, and in particular to a privacy-protecting activity detection method for home-based elderly care based on the synergy of infrared and microwave radar. It is suitable for accurate monitoring of the daily activity status of the elderly in non-wearable and non-imaging scenarios, while achieving zero privacy leakage protection. Background Technology

[0002] As society ages, home-based elder care has become the mainstream model, and monitoring the daily activities of the elderly is a core requirement for ensuring their safety. Existing home-based elder care monitoring technologies are mainly divided into two categories: wearable and non-wearable. Wearable devices require the elderly to wear them voluntarily, leading to problems such as poor compliance, easy loss of contact, and unsuitability for disabled elderly. In non-wearable monitoring, camera-based monitoring poses a serious risk of privacy breaches and is difficult for the elderly to accept.

[0003] To balance monitoring needs with privacy protection, existing technologies often employ a combination of millimeter-wave radar and infrared thermal imaging, using complementary data to monitor falls, heart rate, and other information. However, this approach still has significant drawbacks: infrared thermal imaging is inherently a weak imaging technology, which may still reveal private information such as the elderly person's posture and behavioral details; sensor collaboration methods mostly involve parallel data overlay, lacking scenario-based adaptive scheduling, and are easily affected by home environment factors (such as furniture obstruction and temperature fluctuations), leading to a high false alarm rate; furthermore, most solutions only focus on detecting abnormal events (falls, abnormal heart rate), failing to dynamically capture and predict trends in the elderly person's daily activities, thus making it difficult to meet the refined care needs of home-based elderly care.

[0004] Furthermore, some existing technologies employ a single infrared or microwave radar sensor. A single infrared sensor has weak identification capabilities for static elderly individuals, easily leading to missed detections; a single microwave radar sensor is susceptible to electrical interference in complex home environments and lacks sufficient accuracy in recognizing small movements (such as getting up or changing sitting posture). Therefore, there is an urgent need for a home-based elderly care detection method that is completely non-imaging, highly efficient in sensor collaboration, and combines privacy protection with multi-dimensional activity monitoring capabilities. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing home-based elderly care monitoring technologies, such as high risk of privacy leakage, low efficiency of sensor collaboration, single monitoring dimensions, and weak anti-interference capabilities. This invention proposes a privacy-protecting activity detection method and system for home-based elderly care based on the collaboration of infrared and microwave radar. Under the premise of completely avoiding privacy leakage, this invention enables accurate monitoring and prediction of the daily activity status, behavioral patterns, and abnormal events of the elderly.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a home-based elderly care privacy-protecting activity detection method based on the synergy of infrared and microwave radar, comprising the following steps:

[0007] S1: System deployment and initialization, building the hardware foundation and environmental benchmark feature library for non-imaging collaborative monitoring;

[0008] S2: Layered collaborative triggering monitoring achieves a balance between low power consumption and high-precision monitoring through a layered scheduling mechanism of "infrared pre-triggering + microwave radar precise verification";

[0009] S3: Activity state recognition and pattern modeling, which classifies activity states based on dual-source multi-dimensional feature fusion and constructs a personalized activity pattern baseline;

[0010] S4: Anomaly detection and privacy enhancement processing. It identifies risk events through a two-dimensional anomaly detection mechanism and protects privacy and security through a full-link protection strategy. It combines real-time status and regular deviations to determine anomalies and achieves privacy protection from the signal layer, processing layer and transmission layer.

[0011] S5: Alarms and adaptive optimization, triggering hierarchical alarms and continuously optimizing system parameters and baseline patterns.

[0012] As a further aspect of the present invention, step S1 specifically comprises:

[0013] S11: Home Terminal Deployment. Monitoring terminals are deployed in key home activity areas such as bedrooms, bathrooms, and living rooms. Each terminal integrates one passive infrared (PIR) sensor, one 24GHz non-imaging microwave radar sensor, and a local edge computing unit. The sensors all adopt a non-imaging design. The microwave radar shields the radio frequency imaging function and only extracts the target's motion and presence characteristics. The infrared sensor only collects the human body's thermal radiation change signal and does not generate any image-related privacy data.

[0014] S12: Sensor parameter calibration. The dual sensors are precisely calibrated through the edge computing unit: For the infrared sensor, the thermal radiation threshold range is set (adapting to the 8-14μm band thermal radiation intensity corresponding to the human body temperature of 36-37.5℃), and the interference from environmental heat sources such as heaters and water heaters is shielded through a multi-frame differential algorithm; For the microwave radar, the detection distance range (0.5-5m) and signal sampling frequency (10Hz) are set, and the electromagnetic interference from household appliances such as microwave ovens and routers is eliminated through a wavelet transform-based spatial frequency domain filtering algorithm, thereby improving the signal-to-noise ratio.

[0015] S13: Environmental feature library construction. Input environmental information such as furniture location, fixed heat source distribution, and wall material in the home scene to establish an environmental benchmark feature library as a background reference for subsequent signal recognition and reduce the probability of environmental false triggers.

[0016] As a further aspect of the present invention, step S2 specifically comprises:

[0017] S21: Sleep level monitoring. When there is no effective infrared signal trigger, the microwave radar is in a low-power sleep state (power consumption ≤ 0.5W), and only the infrared sensor monitors in real time with a period of 2 seconds, which greatly reduces the system energy consumption.

[0018] S22: Trigger-level wake-up. When the infrared sensor detects a change in thermal radiation signal exceeding a set threshold (e.g., thermal radiation intensity fluctuation ≥ 0.2 W / m²) and the signal duration is ≥ 0.5 s (transient environmental interference is eliminated through a sliding window algorithm), the microwave radar wake-up command is immediately triggered, and the system enters the collaborative monitoring state.

[0019] S23: Precise monitoring layer fusion. After the microwave radar is woken up, it collects data synchronously with the infrared sensor. The edge computing unit performs dynamic weight fusion of the dual-source data: In dynamic scenarios (such as walking or getting up), the infrared weight is set to 0.4 and the microwave radar weight is set to 0.6, which improves the dynamic recognition accuracy by relying on the motion trajectory characteristics of the radar; In static scenarios (such as sitting or lying down), the infrared weight is set to 0.2 and the microwave radar weight is set to 0.8, which improves the reliability of steady-state monitoring by relying on the static existence characteristics of the radar.

[0020] As a further aspect of the present invention, step S3 specifically comprises:

[0021] S31: Multi-dimensional feature extraction: Extract thermal radiation change rate, signal duration, and heat source movement trajectory features from infrared sensors; extract target distance change rate, motion amplitude, static existence duration, and Doppler frequency shift features from microwave radar, and construct a fusion feature vector containing 12 dimensions.

[0022] S32: Lightweight Neural Network Classification. The MobileNetV3 lightweight convolutional neural network model is used to classify and identify fused feature vectors, accurately distinguishing six activity states: sitting, standing, walking, getting up / sitting down, lying down, and abnormal postures (falling, curling up). Abnormal postures are determined by a combination of microwave radar distance change threshold (≥0.8m / s) and infrared signal stability (no change after change for ≥3s), excluding interference from normal squatting, sitting, lying down and other postures of the elderly.

[0023] S33: Time series analysis to build a baseline. Collect 7 days of activity data of the elderly and use the ARIMA time series analysis algorithm to build a personalized activity pattern baseline, including the frequency of activities in each time period, distribution of residence areas, state switching cycle, and duration of bed rest. For example, record the elderly’s routine pattern of getting up at 6:00-7:00 every day and going to bed from 20:00 to 6:00 the next day to generate a dynamic baseline threshold.

[0024] As a further aspect of the present invention, step S4 specifically comprises:

[0025] S41: Dual-dimensional anomaly judgment adopts a dual judgment mechanism of "real-time state anomaly + regular deviation anomaly": real-time state anomaly is detected when dangerous postures such as falling or curling up are detected, and a level one alarm is triggered immediately; regular deviation anomaly is when the current activity state deviates from the baseline by more than a set threshold (such as the length of time lying in bed is ±2 hours from the baseline, or the time spent not getting up during the normal period is more than 1 hour), and a level two alarm is triggered. Before the level two alarm, a secondary verification is performed by microwave radar (such as confirming the static existence time of the target) to rule out sensor false triggering.

[0026] S42: End-to-end privacy protection ensures zero privacy leakage through a triple protection strategy: ① Signal layer: Sensors only collect raw non-imaging signals and do not generate any images, body posture, or other privacy data; ② Processing layer: All data is processed in the local edge computing unit, without transmitting raw signals, only uploading status results (such as "normal bed rest" or "fall alarm"); ③ Transmission layer: Alarm information and status data are transmitted using the TLS 1.3 encrypted transmission protocol, with differential privacy technology for data anonymization, removing any characteristic information that could be associated with the elderly individual.

[0027] As a further aspect of the present invention, step S5 specifically comprises:

[0028] S51: Tiered alarm triggering. Level 1 alarms are triggered via local audio-visual prompts (volume and brightness can be adjusted adaptively) + push notifications to the guardian's mobile APP + SMS synchronization with emergency contacts, while uploading the time and location of the anomaly. Level 2 alarms are only pushed to the guardian's APP, prompting them to pay attention to the elderly person's status.

[0029] S52: Adaptive parameter optimization. Every 30 days, the system automatically optimizes the baseline and sensor parameters (such as adjusting the infrared thermal radiation threshold and microwave radar dynamic weights) based on newly collected activity data, using a gradient descent algorithm to adapt to changes in the elderly's activity level (such as slower movement and reduced activity range), thereby improving long-term monitoring accuracy. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the main steps of the present invention;

[0031] Figure 2 This is a detailed flowchart of step S1 of the present invention;

[0032] Figure 3 This is a detailed flowchart of step S2 of the present invention;

[0033] Figure 4 This is a detailed flowchart of step S3 of the present invention;

[0034] Figure 5 This is a detailed flowchart of step S4 of the present invention. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. The components of the embodiments of this disclosure described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the detailed description of the embodiments of this disclosure provided below is not intended to limit the scope of the claimed disclosure, but merely represents selected embodiments of this disclosure. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without inventive effort are within the scope of protection of this disclosure.

[0036] Example 1: Bedroom Scene Monitoring

[0037] This embodiment corresponds to a home-based elderly care bedroom scenario, and fully reproduces the entire process of the method of the present invention:

[0038] S1: System deployment and initialization, building the hardware foundation and environmental benchmark feature library for non-imaging collaborative monitoring;

[0039] The specific steps of S1 are as follows:

[0040] S11: A monitoring terminal is deployed 1.8m above the head of the bed in the bedroom. The infrared sensor faces the bed surface and the area around the bed, and the microwave radar detection range covers the entire bedroom space. The terminal integrates a passive infrared sensor, a 24GHz non-imaging microwave radar and a local edge computing unit. All sensors adopt a non-imaging design.

[0041] S12: During the initialization phase, environmental information such as the bed position and air conditioner position is entered to build a bedroom environment feature library; the infrared sensor thermal radiation threshold is calibrated to 8-12μW / cm² (adapted to the thermal radiation intensity corresponding to a human body temperature of 36-37.5℃), and environmental heat source interference such as the air conditioner outdoor unit is shielded through a multi-frame differential algorithm; the microwave radar detection distance is calibrated to 0.5-3m and the sampling frequency is 10Hz, and electromagnetic interference from the bedroom router is eliminated through a spatial frequency domain filtering algorithm.

[0042] S2: Layered collaborative triggering monitoring achieves a balance between low power consumption and high-precision monitoring through a layered scheduling mechanism of "infrared pre-triggering + microwave radar precise verification";

[0043] The specific steps of S2 are as follows:

[0044] S21: In the nighttime sleep mode, the microwave radar is in a low-power sleep state (power consumption ≤ 0.5W), and only the infrared sensor continuously monitors with a period of 2 seconds.

[0045] S22: When the elderly person gets up in the early morning, the infrared sensor detects a change in thermal radiation of 15μW / cm², and the signal duration is ≥0.6s (meeting the trigger threshold), immediately triggering the microwave radar wake-up command.

[0046] S23: In the precise monitoring phase, since getting up in the bedroom is a dynamic scenario, the dynamic weight is set to 0.3 for infrared and 0.7 for microwave radar; the edge computing unit integrates dual-source data to identify the "getting up" state.

[0047] S3: Activity state recognition and pattern modeling, which classifies activity states based on dual-source multi-dimensional feature fusion and constructs a personalized activity pattern baseline;

[0048] The specific steps for S3 are as follows:

[0049] S31: Extract the thermal radiation change rate and signal duration features from the infrared sensor, and the distance change rate and motion amplitude features from the microwave radar, construct a fused feature vector, and accurately identify the "getting up" state using the MobileNetV3 model. S32: The system records the time of this getting up and updates the activity trajectory. Combining this with historical data from 7 consecutive days, a personalized baseline is constructed using the ARIMA algorithm. For example, it determines that the elderly person's usual time to get up is 6:00-7:00, and their time in bed is 20:00-6:00 the next day.

[0050] S4: Anomaly detection and privacy enhancement processing: risk events are identified through a two-dimensional anomaly detection mechanism, while privacy and security are protected through a full-link protection strategy.

[0051] The specific steps of S4 are as follows:

[0052] S41: If the elderly person does not lie down during the normal time period after getting up and deviates from the baseline for more than 1.5 hours, the abnormal deviation judgment is triggered and a level 2 alarm is activated; if the elderly person's distance suddenly changes by 0.9 m / s and the infrared signal remains unchanged for 4 seconds after the sudden change, it is determined to be an abnormal real-time status (fall) and a level 1 alarm is triggered.

[0053] S42: All data is processed in the local edge computing unit, and only the status results of "bedroom fall alarm" or "deviation of activity pattern" are pushed to the guardian. The original signal and privacy data are not transmitted throughout the process, and alarm information is transmitted using TLS 1.3 encryption.

[0054] S5: Alarm and adaptive optimization, triggering hierarchical alarms and continuously optimizing system parameters and baseline patterns;

[0055] The specific steps of S5 are as follows:

[0056] S51: Level 1 alarm triggers local audio and visual alerts (volume and brightness adjustable) + push notification to the guardian's mobile app + SMS notification to emergency contacts; Level 2 alarm only pushes notifications to the guardian's app, prompting them to pay attention to the elderly person's status.

[0057] S52: Every 30 days, the system automatically optimizes the infrared thermal radiation threshold and dynamic weight based on newly collected bedroom activity data to adapt to changes in the elderly's activity level (such as reducing the infrared trigger threshold when movement slows down).

[0058] Example 2: Bathroom Scene Monitoring

[0059] This embodiment takes into account the characteristics of a humid and frequently changing bathroom environment and adjusts the method parameters accordingly:

[0060] S1: System deployment and initialization, building the hardware foundation and environmental benchmark feature library for non-imaging collaborative monitoring;

[0061] The specific steps of S1 are as follows:

[0062] S11: A monitoring terminal with a moisture-proof and sealed design is deployed in the center of the bathroom ceiling. The infrared sensor faces the shower area and the sink area, and the microwave radar detection range covers the entire bathroom.

[0063] S12: During the initialization phase, environmental information such as the location of the water heater and bathroom heater is recorded, and interference from fixed heat sources is shielded through a multi-frame differential algorithm; the microwave radar is activated in a strong anti-interference mode, and electromagnetic interference from the bathroom heater and hair dryer is filtered through a spatial frequency domain filtering algorithm, with the detection distance calibrated to 0.5-4m and the sampling frequency to 10Hz.

[0064] S2: Layered collaborative triggering monitoring achieves a balance between low power consumption and high-precision monitoring through a layered scheduling mechanism of "infrared pre-triggering + microwave radar precise verification";

[0065] The specific steps of S2 are as follows:

[0066] S21: When there is no effective infrared signal, the microwave radar is in a low-power sleep state, and only the infrared sensor monitors with a period of 2 seconds.

[0067] S22: Since the bathroom is a dynamic high-frequency scene, the trigger threshold is adjusted so that the microwave radar can be woken up when the infrared signal duration is ≥0.3s; when the elderly enter the bathroom, the infrared sensor detects the change in thermal radiation and continues for 0.4s, triggering the microwave radar to wake up.

[0068] S23: Precision monitoring stage. Since the bathroom is mainly a dynamic activity area, the dynamic weight is set to infrared 0.4 and microwave radar 0.6. The fusion of dual-source data improves the accuracy of dynamic status recognition.

[0069] S3: Activity state recognition and pattern modeling, which classifies activity states based on dual-source multi-dimensional feature fusion and constructs a personalized activity pattern baseline;

[0070] The specific steps for S3 are as follows:

[0071] S31: Extract the thermal radiation change rate and signal duration characteristics of the infrared sensor, as well as the distance change rate and motion amplitude characteristics of the microwave radar, and identify the "standing" and "walking" states using the MobileNetV3 model.

[0072] S32: Combine bathroom activity data from 7 consecutive days to build a personalized baseline, for example, record the elderly person's daily routine washing time as 7:00-8:00.

[0073] S4: Anomaly detection and privacy enhancement processing: risk events are identified through a two-dimensional anomaly detection mechanism, while privacy and security are protected through a full-link protection strategy.

[0074] The specific steps of S4 are as follows:

[0075] S41: If a sudden change in distance of the elderly is detected by 1.0 m / s and the infrared signal remains stable for 3.5 s, it is determined to be an abnormal real-time status (fall) and a level one alarm is triggered.

[0076] S42: All data is processed locally, and the status result of the "bathroom fall alarm" is only pushed to the guardian. There is no original signal or image transmission, and differential privacy technology is used to de-identify the alarm information.

[0077] S5: Alarm and adaptive optimization, triggering hierarchical alarms and continuously optimizing system parameters and baseline patterns;

[0078] The specific steps of S5 are as follows:

[0079] S51: Level 1 alarm triggers local audio and visual alerts + guardian's APP push notification + emergency contact SMS notification, simultaneously uploading the time and location of the anomaly.

[0080] S52: Every 30 days, the system automatically optimizes the infrared trigger threshold and microwave radar weight based on bathroom activity data to adapt to changes in the elderly's activity level (such as adjusting the radar detection distance when the range of motion shrinks).

Claims

1. A privacy-preserving activity detection method for home-based elderly care based on the synergy of infrared and microwave radar, characterized in that, Includes the following steps: S1: System deployment and initialization, deploying monitoring terminals integrating passive infrared sensors, non-imaging microwave radar sensors and local edge computing units in key areas of the home, calibrating sensor parameters and building a home environment feature database; S2: Layered collaborative triggering monitoring adopts a layered scheduling mechanism of "infrared pre-triggering + microwave radar precise verification" to achieve a balance between low power consumption and high-precision monitoring; S3: Activity state recognition and pattern modeling, extracting multi-dimensional features from dual sensors, identifying activity states through lightweight neural networks, and constructing personalized activity pattern baselines based on time series analysis; S4: Anomaly detection and privacy enhancement processing. It identifies risk events through a two-dimensional anomaly detection mechanism and protects privacy and security through a full-link protection strategy. It combines real-time status and regular deviations to determine anomalies and achieves privacy protection from the signal layer, processing layer and transmission layer. S5: Alarms and adaptive optimization, triggering tiered alarms and continuously optimizing system parameters and baseline patterns based on newly collected data.

2. The home-based elderly care privacy-protecting activity detection method based on the synergy of infrared and microwave radar according to claim 1, characterized in that, S1 includes the following sub-steps: S11: Home terminal deployment, deploy monitoring terminals in key home activity areas such as bedrooms, bathrooms, and living rooms. The passive infrared sensor only collects human thermal radiation change signals, and the non-imaging microwave radar sensor shields radio frequency imaging function and only extracts target motion features and presence features. S12: Sensor parameter calibration, the thermal radiation threshold range of the passive infrared sensor is calibrated to 8-12μW / cm² by the local edge computing unit, the environmental heat source interference is shielded by the multi-frame differential algorithm, the detection range of the non-imaging microwave radar sensor is calibrated to 0.5-5m and the sampling frequency is 10Hz by the non-imaging microwave radar sensor, and electromagnetic interference is eliminated by the spatial frequency domain filtering algorithm. S13: Environmental feature database construction, inputting environmental information such as furniture location and fixed heat source distribution in home scenarios, and establishing an environmental benchmark feature database.

3. The home-based elderly care privacy-protecting activity detection method based on the synergy of infrared and microwave radar as described in claim 1, characterized in that, S2 includes the following sub-steps: S21: Sleep level monitoring. When there is no effective infrared signal trigger, the non-imaging microwave radar sensor is in a low-power sleep state, and only the passive infrared sensor monitors in real time with a period of 2 seconds. S22: Trigger-level wake-up: When the passive infrared sensor detects a change in thermal radiation signal exceeding a set threshold and the duration is ≥0.5s, the non-imaging microwave radar sensor is woken up. S23: Precise monitoring level fusion. After the non-imaging microwave radar sensor is woken up, it synchronously collects data with the passive infrared sensor. The local edge computing unit performs dynamic weight fusion on the dual-source data. In the dynamic scene, the infrared weight is 0.4 and the microwave radar weight is 0.

6. In the static scene, the infrared weight is 0.2 and the microwave radar weight is 0.

8.

4. The home-based elderly care privacy-protecting activity detection method based on the synergy of infrared and microwave radar as described in claim 1, characterized in that, S3 includes the following sub-steps: S31: Multi-dimensional feature extraction, extracting thermal radiation change rate and signal duration features from the passive infrared sensor, and extracting target distance change rate, motion amplitude, and static existence duration features from the non-imaging microwave radar sensor; S32: Lightweight neural network classification, using the MobileNetV3 lightweight convolutional neural network model to identify 6 activity states: sitting, standing, walking, getting up / sitting down, lying down, and abnormal posture; S33: Time series analysis to build a baseline. Collect activity data for 7 consecutive days and build a personalized activity pattern baseline using the ARIMA time series analysis algorithm.

5. The home-based elderly care privacy-protecting activity detection method based on the synergy of infrared and microwave radar according to claim 1, characterized in that, S4 includes the following sub-steps: S41: Dual-dimensional anomaly determination. Real-time state anomaly is triggered when a dangerous posture such as falling or curling up is detected, triggering a level one alarm. Regular deviation anomaly is triggered when the deviation between the current activity state and the baseline exceeds a set threshold, triggering a level two alarm. Before the level two alarm, secondary verification is performed by the non-imaging microwave radar sensor. S42: End-to-end privacy protection ensures zero privacy leakage through a triple protection strategy: ① Signal layer: Sensors only collect raw non-imaging signals and do not generate any images, body posture, or other privacy data; ② Processing layer: All data is processed in the local edge computing unit, without transmitting raw signals, only uploading status results (such as "normal bed rest" or "fall alarm"); ③ Transmission layer: Alarm information and status data are transmitted using the TLS 1.3 encrypted transmission protocol, with differential privacy technology for data anonymization, removing any characteristic information that could be associated with the elderly individual.

6. The home-based elderly care privacy-protecting activity detection method based on the synergy of infrared and microwave radar according to claim 1, characterized in that, S5 includes the following sub-steps: S51: Tiered alarm triggering. Level 1 alarms are triggered simultaneously via local audio and visual alerts, push notifications to the guardian's mobile app, and SMS messages to emergency contacts. Level 2 alarms are only triggered via the guardian's mobile app. S52: Adaptive parameter optimization. The system automatically optimizes the baseline and sensor parameters every 30 days based on newly acquired data.

Images