Intelligent safety monitoring method integrating multi-parameter health monitoring and emergency alarm positioning

Abstract

The application relates to an intelligent safety monitoring method and system integrating multi-parameter health monitoring and emergency alarm positioning, a computer device and a storage medium. The method comprises the following steps: synchronously collecting PPG optical signals, piezoelectric pressure signals, motion posture signals and body surface temperature signals; dynamically calibrating a mapping relationship between pulse wave transmission time and blood pressure based on historical resting data, and calculating physiological parameters; extracting physiological and motion characteristics, inputting a lightweight AI model for real-time inference, and identifying the probability of a dangerous state; when the dangerous probability exceeds a threshold value and continuously exceeds a set time or an active trigger signal is received, starting local sound-light vibration alarm, satellite positioning and sending alarm information; if a false alarm cancellation operation is received, stopping the alarm and inputting the current parameters into the model for false alarm learning, otherwise repeating the alarm. The application realizes millisecond-level local danger identification and full-automatic emergency alarm, is not dependent on the cloud, and has high reliability.

Description

Technical Field

[0001] This application relates to the field of wearable health monitoring device technology, and in particular to an intelligent safety monitoring method, watch, computer device and storage medium that integrates multi-parameter health monitoring and emergency alarm positioning. Background Technology

[0002] With an aging population and increasing demand for outdoor safety, wearable health monitoring devices are becoming increasingly popular. Existing smartwatches typically process collected physiological data via cloud servers and send alerts to the linked device upon detecting anomalies. However, this cloud-based processing method suffers from significant response delays when the network is unstable, and emergency alarms often require the user to manually press a physical button to trigger. When the wearer loses consciousness or is unable to operate the device due to fainting, falls, or other reasons, the alarm may not be triggered automatically, potentially leading to missed opportunities for timely rescue. Summary of the Invention

[0003] Based on this, it is necessary to provide an intelligent safety monitoring method, watch, computer device and storage medium that integrates multi-parameter health monitoring and emergency alarm positioning, which can realize local real-time danger identification and fully automatic emergency alarm, in order to address the above-mentioned technical problems.

[0004] Firstly, this application provides an intelligent safety monitoring method integrating multi-parameter health monitoring and emergency alarm location. The method includes: The wearer's multi-parameter signals are collected simultaneously, including PPG optical signals, piezoelectric force signals, motion posture signals, and body surface temperature signals. The mapping relationship between pulse wave conduction time and blood pressure is dynamically calibrated based on the wearer's historical resting data, and physiological parameters are calculated using the mapping relationship based on the currently collected multi-parameter signals. Physiological and motion features are extracted from the multi-parameter signals, and the physiological and motion features are input into a lightweight AI model for local real-time inference. A hazard index threshold algorithm is used to identify the probability of a dangerous state. The dangerous state includes at least one of the following: fall, collapse, fainting, hypoxia, abnormal blood pressure, hypothermia, fever, or poisoning by harmful gases. When the probability exceeds a preset threshold and the duration of the dangerous state exceeds a set time, or when an active triggering signal is detected, an emergency alarm process is triggered; the emergency alarm process is to activate local sound, light and vibration alarms, activate satellite positioning, and send alarm information to the bound terminal, the alarm information including positioning data and physiological data summary; If the emergency alarm process is completed and a preset false alarm cancellation operation is received, the local audible and visual alarm and the alarm information are stopped, and the current physiological parameters are input into the lightweight AI model for false alarm learning; if the emergency alarm process is completed, the dangerous state continues, and the false alarm cancellation operation is not received, the local audible and visual alarm and the alarm information are repeated at set intervals.

[0005] In one embodiment, the identification of the fall and the drop includes: Based on the motion attitude signal, the synthetic acceleration and attitude angle change are calculated, and the peak value of the impact acceleration is extracted from the synthetic acceleration. A fall is determined when the peak impact acceleration exceeds the impact acceleration fall threshold, the change in attitude angle exceeds the attitude angle change threshold, and the duration of attitude stabilization after the fall exceeds the attitude stabilization duration threshold. When the peak value of the impact acceleration exceeds the impact acceleration fall threshold and the duration of the excess exceeds the impact acceleration duration threshold, it is determined to be a fall.

[0006] In one embodiment, the identification of syncope includes: Heart rate variability is calculated based on the PPG optical signal; Syncope is defined as follows: when the heart rate is below the severe bradycardia heart rate threshold or above the severe tachycardia heart rate threshold and the heart rate variability is below the heart rate variability threshold; or when the blood oxygen saturation is below the severe hypoxemia threshold and the posture angle obtained based on the movement posture signal is below the supine posture angle threshold for a continuous period exceeding the supine duration threshold.

[0007] In one embodiment, the identification of hypoxia includes: The rate of change in blood oxygen saturation and the magnitude of heart rate increase were calculated based on the PPG optical signal. When the blood oxygen saturation is below the low blood oxygen threshold for a sustained period exceeding the low blood oxygen duration threshold, or when the blood oxygen saturation is between the lower limit and the upper limit of the low blood oxygen range, the decrease in blood oxygen saturation exceeds the blood oxygen decrease threshold, and the increase in heart rate exceeds the heart rate compensation increase threshold, it is determined to be hypoxia.

[0008] In one embodiment, the identification of the hypothermia and the fever includes: The skin temperature change rate is calculated based on the surface temperature signal, and the difference between the skin temperature and the ambient temperature is calculated. When the skin temperature is between the lower limit and the upper limit of the mild hypothermia range and the rate of decrease in skin temperature exceeds the hypothermia warning change rate threshold, it is determined to be a mild hypothermia warning. When the skin temperature is between the lower limit and the upper limit of the moderate hypothermia range, a moderate hypothermia alarm is triggered. When the skin temperature is below the severe hypothermia threshold and the difference between the skin temperature and the ambient temperature is below the severe hypothermia temperature difference threshold, a severe hypothermia alarm is triggered. When the skin temperature exceeds the fever threshold, it is determined to be a fever.

[0009] In one embodiment, the identification of the blood pressure abnormality includes: The severity of low blood pressure and high blood pressure is classified according to the different ranges of systolic and diastolic blood pressure.

[0010] In one embodiment, the identification of the harmful gas poisoning includes: The respiratory rate is acquired based on the piezoelectric pressure signal, and the heart rate and blood oxygen saturation are acquired based on the PPG optical signal. The ratio of heart rate to respiratory rate is calculated. When the respiratory rate is below the bradycardia threshold or above the tachycardia threshold, and the heart rate is above the poisoning tachycardia threshold or below the poisoning bradycardia threshold, and the ratio of heart rate to respiratory rate is below the low heart rate-to-respiratory ratio threshold or above the high heart rate-to-respiratory ratio threshold, and the blood oxygen saturation is below the poisoning blood oxygen threshold and continues to decrease beyond the poisoning blood oxygen decrease rate threshold, it is determined to be poisoning by the harmful gas.

[0011] Secondly, this application also provides a smart safety monitoring watch that integrates multi-parameter health monitoring and emergency alarm positioning. The watch includes: The watch case has an arc-shaped bottom surface that matches the direction of the radial artery; The watch strap is mounted on the watch case; The main control module, located inside the watch case, is used to execute the aforementioned intelligent safety monitoring method that integrates multi-parameter health monitoring and emergency alarm location. The sensor module includes a PPG optical sensor, a piezoelectric pressure sensor array, an accelerometer and a gyroscope disposed on the watch case, and a flexible temperature sensor embedded inside the watch strap. The communication and positioning module includes a satellite positioning submodule and a wireless communication submodule disposed within the casing; The emergency alarm module includes a physical SOS button, a speaker, and a vibration motor located on the case. The power management module is located inside the watch case.

[0012] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the aforementioned intelligent safety monitoring method integrating multi-parameter health monitoring and emergency alarm location.

[0013] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the aforementioned intelligent safety monitoring method integrating multi-parameter health monitoring and emergency alarm location.

[0014] The aforementioned intelligent safety monitoring method, watch, computer device, and storage medium, which integrates multi-parameter health monitoring and emergency alarm positioning, synchronously collects multimodal physiological and motion signals and performs real-time inference using a locally running lightweight AI model. This eliminates the need for cloud networks to identify dangerous states, shortening the response time from the occurrence of an anomaly to alarm triggering. By using a dangerous state probability exceeding a preset threshold and a duration meeting certain conditions as the automatic triggering condition, and being compatible with active triggering signals, it achieves fully automatic and manual dual-mode alarms, ensuring timely response for the wearer in any situation. Dynamic calibration of the mapping relationship between pulse wave transit time and blood pressure improves the measurement accuracy of multi-parameter physiological indicators. By introducing a false alarm cancellation operation and feeding current physiological parameters back to the lightweight AI model for learning, the false alarm rate can be gradually reduced. When a dangerous state persists and no cancellation operation is received, the alarm and information transmission are repeated, ensuring reliable transmission of alarm information. In summary, this application improves the real-time performance, accuracy, and reliability of health monitoring for high-risk groups while maintaining low power consumption and miniaturization. Attached Figure Description

[0015] Figure 1 This is an application environment diagram of an intelligent safety monitoring method that integrates multi-parameter health monitoring and emergency alarm location in one embodiment; Figure 2 This is a flowchart illustrating an intelligent safety monitoring method that integrates multi-parameter health monitoring and emergency alarm location in one embodiment. Figure 3 This is a schematic diagram of a smart safety monitoring watch that integrates multi-parameter health monitoring and emergency alarm positioning in one embodiment. Figure 4 This is an internal structural diagram of a computer device in one embodiment.

[0016] Attached label: 102-Smart safety monitoring watch; 104-Backend server; 1-Watch case; 2-Watch strap; 3-OLED touch screen; 4-Physical SOS button; 5-Speaker. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. First, to facilitate understanding of the technical solutions provided by the embodiments of this application, the background technology involved in the embodiments of this application will be described below.

[0018] With the accelerating aging of the population and the increasing prevalence of outdoor adventures and independent travel, the demand for real-time health monitoring and safety alerts for high-risk groups such as children, the elderly, patients with chronic diseases, and field workers has risen significantly. Currently, smartwatches have begun to integrate health monitoring functions such as heart rate, blood oxygen, and fall detection. However, the physiological parameter measurements of these devices (especially blood pressure) largely rely on a single PPG (photoplethysmography) sensor, and the measurement accuracy is severely affected by motion artifacts and ambient light interference, making it difficult to meet clinical reference standards. More importantly, the identification of dangerous events (such as fainting, cardiac arrest, and severe falls) usually requires transmitting the collected physiological data to a cloud server for model inference. When network signals fluctuate or are delayed, the total response time from the occurrence of the event to the triggering of the alarm generally exceeds 2 seconds, easily missing the golden rescue opportunity. Furthermore, the alarm mechanisms of existing devices are mostly triggered by the user actively pressing a physical button. If the wearer loses consciousness or has limited limb movement due to sudden fainting, a fall, poisoning, or other reasons, the alarm will not be activated. Therefore, how to achieve accurate monitoring of multi-parameter physiological signals and millisecond-level local identification of dangerous states without relying on cloud networks, and how to automatically issue an emergency alarm with precise location when the wearer has no active operation capability, is a technical problem that urgently needs to be solved in this field.

[0019] Therefore, embodiments of this application provide an intelligent safety monitoring method, watch, computer device, and storage medium that integrates multi-parameter health monitoring and emergency alarm positioning.

[0020] The intelligent safety monitoring method integrating multi-parameter health monitoring and emergency alarm location provided in this application embodiment can be applied to, for example... Figure 1In the application environment shown, the smart safety monitoring watch 102 acts as the executor, synchronously collecting the wearer's physiological and motion signals through its built-in multimodal sensors. It utilizes a local lightweight AI model for real-time inference and danger state identification, achieving millisecond-level response independently without relying on a cloud network. When a dangerous state is detected or an active trigger signal is detected, the watch automatically activates local audible, visual, and vibration alarms, initiates satellite positioning, and sends alarm information containing location data and a summary of physiological data directly to the bound terminal (such as the guardian's mobile phone, community service platform, or backend server 104) via its wireless communication module. The bound terminal can be a smartphone, tablet, personal computer, or IoT monitoring device. The backend server 104 can be a standalone physical server, a server cluster consisting of multiple servers, or a cloud platform. It receives alarm information, stores historical data, and supports data statistical analysis. However, the real-time identification of dangerous states and the triggering of emergency alarms are all completed locally by the smart safety monitoring watch 102, without relying on a server for real-time processing.

[0021] It should be noted that the collection, storage, processing, and transmission of personal data such as physiological parameters and location information involved in the embodiments of this application have all been performed with the explicit informed consent of the user or their legal guardian. Users can view, manage, or delete their personal data at any time through the watch's accompanying application software or device settings. All data processing is solely for health monitoring, hazard identification, and emergency alarm purposes, and will not be used for any other unauthorized purposes. The implementation of the embodiments of this application strictly complies with the requirements of relevant privacy regulations to ensure that user data security and privacy rights are not infringed.

[0022] Firstly, in the field of wearable health monitoring, existing devices generally suffer from problems such as high response latency due to reliance on cloud networks for hazard identification, and the inability to send distress signals in a coma or disabled state due to the need for active triggering of alarms. Therefore, in one embodiment, such as... Figure 2 As shown, an intelligent safety monitoring method integrating multi-parameter health monitoring and emergency alarm location is provided, which is then applied to... Figure 1 Taking the Smart Security Monitoring Watch 102 as an example, the following steps are included: Step S1: Synchronously acquire multi-parameter signals from the wearer, including PPG optical signals, piezoelectric force signals, motion posture signals, and body surface temperature signals.

[0023] Specifically, the Smart Safety Monitoring Watch 102 incorporates a PPG optical sensor, a piezoelectric pressure sensor array, a three-axis accelerometer and gyroscope, and a flexible temperature sensor. These sensors operate synchronously at set sampling frequencies: the PPG sensor at 10Hz (one data point every 100ms), the accelerometer and gyroscope at 20Hz (one data point every 50ms), the piezoelectric pressure sensor array at 5Hz (one data point every 200ms), and the flexible temperature sensor at 1Hz (one data point every 1s). Timestamp alignment is achieved through hardware or software triggering to ensure strict temporal correspondence among the signals. The PPG optical signal is used to extract heart rate, blood oxygen saturation, and pulse wave conduction time; the piezoelectric pressure signal is used to capture pulse wave pressure waveforms and minute pressure changes caused by respiration; motion posture signals include three-axis acceleration components (ax, ay, az), pitch angle θ, and roll angle φ; and the body surface temperature signal directly reflects skin temperature.

[0024] Step S2: Dynamically calibrate the mapping relationship between pulse wave conduction time and blood pressure based on the wearer's historical resting data, and calculate physiological parameters using the mapping relationship based on the currently collected multi-parameter signals.

[0025] Specifically, after initial use or reset, perform the calibration procedure: The wearer should wear the watch while at rest (sitting for 5 minutes, keeping the wrist level with the heart) and simultaneously measure three systolic blood pressure (SBP) and diastolic blood pressure (DBP) measurements using a standard upper arm electronic blood pressure monitor. Record the pulse wave transit time (PWTT) at each measurement. PWTT is defined as the time interval from the start of the rising limb of the PPG signal to the point of maximum slope of the rising limb, in milliseconds (ms). Using the three calibration data, establish a linear regression model: , The coefficients a1, b1, a2, and b2 are obtained through least squares fitting. For example, for SBP, the solution makes... The smallest a1, b1, Let be the systolic blood pressure measured in the i-th time. Let a1 be the pulse wave conduction time of the i-th measurement; a2 and b2 are calculated similarly. During routine monitoring, the current PWTT is recalculated every 5 minutes and substituted into the above formula to obtain the real-time blood pressure value. Simultaneously, the blood pressure is dynamically fine-tuned using changes in pulse wave amplitude in the piezoelectric pressure signal: the ratio of the real-time pulse wave amplitude to the amplitude at the most recent calibration is calculated. If the ratio deviates from 1 by more than ±10%, automatic recalibration is triggered: 30 seconds of continuous data collection is used to refit the mapping coefficients. Furthermore, the user is prompted to perform a resting calibration every 7 days. The physiological parameters calculated include: heart rate (obtained by averaging the peak intervals of the PPG signal), blood pressure (systolic / diastolic pressure), blood oxygen saturation (obtained by looking up the ratio R of the dual-wavelength PPG signal), and skin temperature (read directly from the temperature sensor).

[0026] Step S3: Extract physiological and motion features from the multi-parameter signal, and input the physiological and motion features into the lightweight AI model for local real-time inference. Use the danger index threshold algorithm to identify the probability of dangerous states. Dangerous states include at least one of the following: fall, descent, fainting, hypoxia, abnormal blood pressure, hypothermia, fever, or poisoning by harmful gases.

[0027] Specifically, feature extraction employs a sliding window approach (e.g., a window length of 3 seconds and a step size of 1 second) to extract feature vectors characterizing the wearer's physiological and motor states from the collected signals. The extracted physiological features may include heart rate variability (HRV), pulse wave time (PWTT), respiratory rate (RR), and rate of change in blood oxygen saturation; the extracted motor features may include synthetic acceleration, change in posture angle, and peak impact acceleration. The extracted feature vectors are then input into a pre-trained lightweight AI model. In this embodiment, the lightweight AI model uses a convolutional neural network (CNN) architecture, with the following specific structure: Input layer: Accepts a one-dimensional feature vector with dimension N (N depends on the types of features extracted; in this embodiment, N=16, including heart rate variability, pulse wave transit time, respiratory rate, peak composite acceleration, attitude angle change, blood oxygen saturation rate, temperature change rate, etc.). First convolutional layer: Depthwise separable convolution, kernel size 3, stride 1, input channel 1, output channel 16. This layer contains depthwise convolution and pointwise convolution, with a total of 3×1×16+1×1×1×16=64 parameters; The second convolutional layer is a depthwise separable convolution with a kernel size of 3, a stride of 1, 16 input channels, and 32 output channels. The number of parameters is 3×3×16 + 1×1×16×32 = 144 + 512 = 656. Global average pooling layer: The feature map (length L×32) output by the second convolutional layer is globally averaged to obtain a vector of length 32; Fully connected layer: 32 inputs, 8 outputs (corresponding to 8 dangerous states), with a parameter count of 32 × 8 = 256; Output layer: Using the Sigmoid activation function (any one of the prior art), output the probability value (0 to 1) for each dangerous state.

[0028] The total number of model parameters is approximately 64 + 656 + 256 = 976, far less than 100KB (approximately 3.9KB based on float32, and about 1KB after 8-bit integer quantization). The model uses 8-bit integer (INT8) quantization and can run on the watch's built-in AI processing unit (such as the NPU in Rockchip RV1103B, with a computing power of 0.5 TOPS). Actual testing shows that a single inference time is ≤50ms and inference power consumption is <50mW.

[0029] The model is trained using supervised learning. The training dataset contains a large amount of real and simulated scenario data: synchronous multi-parameter signals were collected from over 200 subjects (covering children, adults, and the elderly) under various activity states (sitting, walking, running, fall simulation, fainting simulation, different ambient temperatures, etc.), and professional personnel labeled them with danger state tags (fall, collapse, fainting, hypoxia, abnormal blood pressure, hypothermia, fever, poisoning by harmful gases). The specific training process is as follows: Data preprocessing: Normalize the feature vector of each sample to the interval [0,1]. The normalization formula is as follows: ,in , The minimum and maximum values ​​of this feature in the training set; Loss function: Binary cross-entropy loss is used. The loss is calculated and summed for each of the 8 output nodes. The loss for a single node is... ,in For true labels (0 or 1). Predict probabilities for the model; Optimizer: Adam, initial learning rate 0.001, momentum parameters β1=0.9, β2=0.999; Batch size: 64; Training epochs: 50 epochs; Quantization: After training, the model is converted to TensorFlow Lite format and post-training quantization is performed using INT8. The calibration dataset is a subset of the training set (approximately 1000 samples). The quantized model is approximately 1KB in size, resulting in about a 4x speedup for inference on the NPU.

[0030] The model outputs eight probability values. Simultaneously, a threshold-based expert rule is set as a backup judgment (e.g., a syncope alarm is directly triggered when the heart rate is below 40 beats / minute and the heart rate variability is below 20ms). Dangerous states include falls, syncope, hypoxia, abnormal blood pressure, hypothermia, fever, and poisoning by harmful gases. In this step, as long as any of the above dangerous states is identified (i.e., the probability exceeds the threshold or the expert rule is met), the condition for triggering an alarm is considered met.

[0031] Step S4: When the probability of a dangerous state exceeds a preset threshold and the duration of the dangerous state exceeds a set time, or when an active trigger signal is detected, an emergency alarm procedure is triggered. The emergency alarm procedure involves activating local audible, visual, and vibration alarms, initiating satellite positioning, and sending alarm information to the bound terminal. The alarm information includes location data and a summary of physiological data.

[0032] Specifically, the preset thresholds are set according to different dangerous situations, for example, the fall probability threshold is 0.7 and the fainting probability threshold is 0.8; the set time is 3 seconds. The active trigger signal comes from the physical SOS button (pressed for more than 1 second). After triggering, the following actions are performed: (1) Activate local sound, light and vibration alarm (high frequency sound above 85dB, 200Hz vibration); (2) Activate satellite positioning (BeiDou / GPS dual mode, 1Hz refresh rate, accuracy ≤5 meters); (3) Send alarm information to the bound terminal through cellular communication (4G Cat1), including device identification, timestamp, location data, and a summary of physiological data for the last 10 seconds. The total time is ≤1 second.

[0033] Step S5: If the emergency alarm process is completed and a preset false alarm cancellation operation is received, stop sending local audible and visual alarms and alarm information, and input the current physiological parameters into the lightweight AI model for false alarm learning; if the emergency alarm process is completed, the dangerous state continues and no false alarm cancellation operation is received, repeat the sending of local audible and visual alarms and alarm information at set intervals.

[0034] Specifically, the preset false alarm cancellation procedure is: press the SOS button three times consecutively within 5 seconds. Upon detection, the alarm and information transmission are immediately terminated, and physiological parameters from 30 seconds before and after the alarm are packaged as negative samples for online model learning. Online learning employs the Elastic Weight Consolidation (EWC) method, adding a regularization term to the loss function to protect important parameters, and performing minor iterative fine-tuning. For example, all physiological parameters (heart rate, blood pressure, blood oxygen, temperature, and movement posture) and feature data from 30 seconds before and after the alarm are packaged as negative samples. (All labels are 0). Online fine-tuning is performed using the Elastic Weight Consolidation (EWC) method. The EWC loss function is: in For cross-entropy loss, For the i-th model parameter, These are the parameters of the original model. Let λ be the diagonal element of the Fisher information matrix (calculated from the training set, representing the average of the squared logarithmic gradients of the model with respect to the output probability at the i-th parameter, used to measure the importance of the parameter), λ = 10. Fine-tuning is performed for 10 epochs using a learning rate of 0.0001.

[0035] If the dangerous situation persists and no cancellation action is received after 10 seconds, the local alarm and information transmission (location data refresh) will be repeated every 30 seconds until the dangerous situation disappears or a cancellation action is received.

[0036] Based on the above, this embodiment achieves millisecond-level hazard identification through local multimodal signal acquisition and lightweight AI inference, eliminating reliance on the cloud; automatic and manual dual triggering ensures alarms can be triggered in any state; false alarm learning gradually reduces the false alarm rate; and repeated alarms ensure reliable information transmission. This solves the problems of high response latency and unreliable alarms.

[0037] Existing fall detection devices often rely on a single acceleration peak, which is prone to misjudgment and cannot distinguish between a fall and a descent. Therefore, in one embodiment, the identification of falls and descents includes: The synthetic acceleration and attitude angle change are calculated based on the motion attitude signal, and the peak value of the impact acceleration is extracted from the synthetic acceleration. A fall is defined as a situation where the peak impact acceleration exceeds the impact acceleration fall threshold, the change in attitude angle exceeds the attitude angle change threshold, and the duration of attitude stabilization after a fall exceeds the attitude stabilization duration threshold. A fall is defined as an event when the peak impact acceleration exceeds the impact acceleration fall threshold and the duration of the excess exceeds the impact acceleration duration threshold.

[0038] Specifically, for the extraction of the peak impact acceleration, let the triaxial acceleration value at the current sampling time t (sampling interval 50ms) be... (Unit: m / s) 2 ). Composite acceleration The calculation formula is: Attitude angle change: Pitch angle obtained by gyroscope integration and roll angle (Unit: °). Calculate the sampled values ​​of two adjacent frames (100ms interval, i.e., 2 sampling points, t−2): and Change in: Peak impact acceleration Amax: The maximum value of the composite acceleration is taken within a sliding window (1 second in length, i.e., 20 sampling points).

[0039] Secondly, the impact acceleration drop threshold (denoted as) Set to 8m / s 2 For child users, the speed can be reduced to 6m / s in the app. 2 For field workers, the speed can be increased to 10 m / s. 2 These parameter values ​​were obtained based on statistical analysis of data from simulated fall experiments involving 50 volunteers: the peak impact accelerations measured during simulated forward, backward, and lateral falls on a soft mat ranged from 7.2 to 12.5 m / s². 2 Between, take the median of 8 m / s 2 It can capture most real falls while avoiding the occurrence of walking (maximum speed of about 3m / s). 2 ) or running (maximum approximately 6m / s 2 (Misjudgment). This embodiment uses the default value of 8m / s. 2 Attitude angle change threshold (denoted as...) or The angle is set to 45°, based on human anatomy: when falling, the body changes from an upright posture (pitch angle approximately 90°) to a lying posture (approximately 0°), a change of approximately 90°. Half of this value, 45°, is taken as the minimum significant change threshold. The postural stability duration threshold (denoted as...) The duration is set to 3 seconds, based on clinical research: if a person is unable to stand up on their own after a fall, there is usually no significant active movement of the body within 3 seconds (the composite acceleration remains between 0.8 and 1.2 m / s²). 2 (within the range), therefore a stable duration of 3 seconds indicates that the wearer may have lost the ability to move. Judgment logic: When m / s 2 and( ° or When the temperature reaches 0.8°C, the system starts timing and checks whether the composite acceleration remains at 0.8 m / s² for the next 3 seconds. 2 Up to 1.2 m / s 2 The acceleration is considered to be within a certain range (approaching a stationary state, as the accelerometer output is approximately 0g~1g when the human body is horizontally stationary, and a range of 0.8~1.2g is reasonable). If the combined acceleration remains within this range for 3 seconds, it is considered a fall event; if there is a significant fluctuation (such as the wearer attempting to get up, causing the combined acceleration to exceed this range), the judgment is terminated. This combination of conditions effectively excludes situations such as rapid bending over (posture change but small impact acceleration), clapping (large impact acceleration but unchanged posture), and quickly regaining balance after a jump (short postural stability time).

[0040] In addition, the impact acceleration fall threshold is set at 9 m / s². 2 The duration threshold is set to 0.5 seconds, based on the fact that the impact acceleration of a body falling freely from a height of 1 meter and hitting the ground typically exceeds 10 m / s². 2 Taking into account the human body's buffering capacity, we take 9 m / s. 2 This serves as a threshold for distinguishing severe falls. A typical fall impact waveform has a half-width at half-maximum (FWHM) of approximately 0.3–0.6 seconds; using 0.5 seconds effectively excludes spikes caused by brief impacts (such as clapping or wrist flicking). Judgment logic: If… m / s², and If the state of m / s² continues for ≥0.5 seconds, it is considered a fall.

[0041] Based on the above, by introducing attitude angle change and attitude stabilization duration as auxiliary conditions for fall detection, the false alarm rate is reduced. Separating the judgment logic for falls from falls allows for the adoption of corresponding alarm strategies based on different injury types (e.g., prioritizing more accurate location data and shortening the interval between repeated alarms during a fall), improving adaptability to different hazardous scenarios.

[0042] Syncope has complex mechanisms and can be caused by cardiac factors (extremely fast or slow heart rate), neurological factors (decreased heart rate variability), or postural factors (hypotension + lying flat). A single heart rate threshold can easily miss detections. Therefore, in one embodiment, syncope identification includes: Heart rate variability calculation based on PPG optical signals; Syncope is defined as follows: when the heart rate is below the heart rate threshold for severe bradycardia or above the heart rate threshold for severe tachycardia and the heart rate variability is below the heart rate variability threshold; or when the blood oxygen saturation is below the threshold for severe hypoxia and the posture angle obtained based on the motion posture signal is below the supine posture angle threshold and continues to exceed the supine duration threshold.

[0043] Specifically, extracting heart rate sequences from PPG signals. (i=1,...,n, where n is the number of heartbeats in 1 minute). Calculate the heart rate variability (SDNN): Secondly, the heart rate thresholds for severe bradycardia are: <40 beats / minute for the elderly, <50 beats / minute for children, and <45 beats / minute for field workers. These values ​​refer to the definition of symptomatic bradycardia in the relevant bradycardia and cardiac conduction delay assessment and management guidelines (heart rate <40 beats / minute in a conscious state is severe bradycardia). The heart rate for children is based on pediatric standards, and the median value is used for field workers.

[0044] For severe tachycardia, the heart rate thresholds are: >150 beats / minute for the elderly, >160 beats / minute for children, and >155 beats / minute for outdoor workers. Referencing the relevant guidelines for the management of supraventricular tachycardia, the threshold for extreme tachycardia (>150 beats / minute) is adjusted upwards to 160 beats / minute to account for the faster heart rates in children.

[0045] The heart rate variability threshold was set to 20 milliseconds. This value is based on industry-recognized literature reports: a heart rate variability of less than 20 milliseconds indicates a severely reduced heart rate variability, which is closely related to autonomic nervous system failure and syncope precursors.

[0046] Judgment criteria: If the average heart rate within the window is less than 40 beats / minute or more than 150 beats / minute, and SDNN is less than 20 milliseconds, then it is judged as a syncope event.

[0047] In addition, the threshold for severe hypoxemia is set at 88%. According to relevant guidelines for the diagnosis and treatment of altitude sickness, an oxygen saturation level below 88% is considered severe hypoxemia and requires immediate intervention.

[0048] The supine posture angle threshold is set to 30° (i.e., pitch angle ≤ 30°). A pitch angle ≤ 30° indicates that the body is close to horizontal (lying flat or severely tilted), which is a common position after fainting.

[0049] The duration threshold for lying flat is set to 2 seconds. 2 seconds is an empirical value that can exclude brief head-down or bending-over movements.

[0050] Criteria for diagnosis: If blood oxygen saturation is <88%, and the pitch angle obtained from the posture signal is ≤30° and this state lasts for more than 2 seconds, it is considered a syncope event. This criterion is independent of heart rate and applies to loss of consciousness caused by orthostatic hypotension, vasovagal syncope, or epileptic seizures.

[0051] Based on the above, this method combines three different criteria—extreme heart rate values, heart rate variability, and blood oxygenation-positional combination—to cover various types of syncope, including cardiogenic (rapid / slow arrhythmias), neurogenic (autonomic dysfunction), and orthostatic (orthostatic hypotension). Compared to a single threshold method, it improves sensitivity while maintaining specificity, thus ensuring timely alarm triggering when syncope occurs.

[0052] Hypoxia detection needs to consider both acute severe hypoxia (such as drowning or suffocation) and progressive hypoxia (such as altitude ascent). Traditional devices only have fixed thresholds, which can easily miss slow declines or fail to provide early warnings. Therefore, in one embodiment, hypoxia detection includes: Calculation of blood oxygen saturation change rate and heart rate increase based on PPG optical signals; Hypoxia is defined as a condition where blood oxygen saturation is below the low blood oxygen threshold for an extended period beyond the low blood oxygen duration threshold, or when blood oxygen saturation is between the lower and upper limits of the low blood oxygen range, the decrease in blood oxygen saturation exceeds the blood oxygen decrease threshold, and the increase in heart rate exceeds the heart rate compensatory increase threshold.

[0053] Specifically, the blood oxygen saturation (SpO2) value is continuously recorded for 5 minutes, once every 1 second. Let the SpO2 value at the initial time t0 be... The value at time t5 after 5 minutes is Calculate the rate of change (decline) of blood oxygen saturation: (Unit: %, positive value indicates a decrease).

[0054] At the same time, record the heart rate value and calculate the increase in heart rate: (Unit: times / minute)

[0055] Secondly, the low oxygen threshold is set at 88%, the same as the aforementioned severe low oxygen threshold. The low oxygen duration threshold is set at 1 minute. The rationale is that in healthy individuals, brief breath-holding may cause SpO2 to temporarily drop below 90% but recover within 30 seconds. Hypoxemia lasting more than 1 minute is pathological, indicating serious respiratory or circulatory system disorders.

[0056] Criteria for determination: If SpO2 < 88% and this state is maintained for more than 60 seconds, it is considered a hypoxic event. This criterion applies to acute severe hypoxia, such as altitude sickness, drowning, and suffocation.

[0057] Secondly, the lower limit of the low-to-mid range for blood oxygen saturation is set at 88%, and the upper limit at 92%. This range corresponds to mild hypoxia (normal SpO2 is 95%~100%). The threshold for blood oxygen saturation decrease is set at 5% (i.e., a decrease of ≥5% within 5 minutes). This is based on the fact that in healthy adults at rest, the diurnal fluctuation of blood oxygen saturation is usually less than 2%, and a decrease of 5% exceeds the normal fluctuation range, serving as a warning sign. The threshold for compensatory increase in heart rate is set at 15 beats / minute. When the body is hypoxic, the heart rate increases compensatorily to maintain tissue oxygen supply. An increase of ≥15 beats / minute indicates a significant compensatory demand; if this is accompanied by a decrease in SpO2, it strongly suggests pathological hypoxia. Criteria for diagnosis: If SpO2 is between 88% and 92% (inclusive), and the SpO2 decreases by ≥5% within 5 minutes, and the heart rate increases by ≥15 beats / minute, then a hypoxic event is diagnosed. This criterion applies to slowly developing hypoxic environments (such as several hours after entering a high-altitude area or carbon dioxide accumulation in a confined space), and can provide early warning before severe hypoxia is reached.

[0058] Based on the above, this embodiment can identify both acute severe hypoxia and provide early warning of progressive hypoxia. By introducing the rate of decrease in blood oxygen and the compensatory increase in heart rate, the advance of hypoxia recognition is shifted forward by several minutes to tens of minutes from the point when the blood oxygen level falls below the lower limit of the low-to-mid range. This provides the wearer with time to respond (such as supplementing oxygen, evacuating from high-altitude areas, or opening ventilation), reducing the risk of brain injury caused by hypoxia.

[0059] Hypothermia and fever are two diametrically opposed abnormalities of body temperature, requiring graded responses. Existing devices often only detect absolute temperature values, ignoring the rate of temperature change and the difference between ambient temperature and body surface temperature, leading to delayed warnings of hypothermia or misdiagnosis of fever. Therefore, in one embodiment, the identification of hypothermia and fever includes: The rate of change of skin temperature is calculated based on the body surface temperature signal, and the difference between skin temperature and ambient temperature is also calculated. When the skin temperature is between the lower limit and the upper limit of the mild hypothermia range and the rate of decrease in skin temperature exceeds the hypothermia warning change rate threshold, it is judged as a mild hypothermia warning. When the skin temperature is between the lower limit and the upper limit of the moderate hypothermia range, a moderate hypothermia alarm is triggered. When the skin temperature is below the severe hypothermia threshold and the difference between the skin temperature and the ambient temperature is below the severe hypothermia temperature difference threshold, a severe hypothermia alarm is triggered. A fever is defined as a skin temperature exceeding the fever threshold.

[0060] Skin temperature T (unit: °C) is read directly from a flexible temperature sensor embedded inside the watch band. Ambient temperature. The ambient temperature is obtained from the watch's built-in sensor (or local weather data obtained via the internet, with preference given to the built-in sensor). The rate of change (decrease) in skin temperature over 10 minutes is calculated. (Unit: °C, positive values ​​indicate a decrease in temperature).

[0061] in The skin temperature at the start time. This represents the skin temperature after 10 minutes. Calculate the difference between the skin temperature and the ambient temperature: (Unit: ℃, can be negative).

[0062] Secondly, the range for mild hypothermia is 32℃ to 35℃ (inclusive). According to the Wilderness Medicine Association's guidelines for the prevention and management of hypothermia, a core temperature of 32 to 35℃ is considered mild hypothermia. Skin temperature is typically 2 to 3℃ lower than core temperature; therefore, a skin temperature of 32 to 35℃ corresponds to a core temperature of approximately 34 to 37℃, which falls on the borderline of mild hypothermia.

[0063] The threshold for the rate of change of hypothermia warning is set at 2℃ / 10 minutes. The basis for this setting is that when a healthy adult is resting in a cold environment, the rate of decrease in skin temperature is about 0.5℃ / 10 minutes. When the rate of decrease reaches 2℃ / 10 minutes, it indicates that heat loss is serious and early intervention is required (such as adding clothing or finding a sheltered place).

[0064] Judgment criteria: If the skin temperature is between 32℃ and 35℃, and the temperature drops by ≥2℃ within 10 minutes, a mild hypothermia warning will be triggered. At this time, the system will prompt the wearer to keep warm through vibration and on-screen text, but an emergency alarm has not yet been triggered.

[0065] Secondly, the moderate hypothermia range is 28℃ to 32℃ (excluding 32℃). This corresponds to a core temperature of approximately 30 to 34℃, which is considered moderate hypothermia. At this point, the person may already be experiencing dangerous symptoms such as confusion, decreased coordination, and cessation of shivering.

[0066] Judgment criteria: If the skin temperature is between 28℃ and 32℃ (inclusive of 28℃, excluding 32℃), a moderate hypothermia alarm will be triggered. At this time, a local alarm will be immediately activated (vibration + red screen flashing) and a warning message will be sent to the bound terminal.

[0067] In addition, the severe hypothermia threshold is set at 28°C. This corresponds to a core temperature below approximately 30°C, constituting severe hypothermia, which is life-threatening (potentially leading to cardiac arrest). The severe hypothermia temperature difference threshold is set at 5°C. When the difference between skin temperature and ambient temperature is less than 5°C, it indicates that the wearer has lost their ability to generate and retain heat, and their body temperature is about to reach equilibrium with the environment, a critical situation.

[0068] Judgment criteria: If the skin temperature is <28℃ and the difference between the skin temperature and the ambient temperature is <5℃, a severe hypothermia emergency alarm (highest priority) will be triggered. At this time, the system will activate the audible and visual alarm, vibration, and satellite positioning, and send the highest priority alarm message to the bound terminal.

[0069] Furthermore, the fever threshold is set at 38℃. Referring to the Chinese guidelines for the diagnosis and management of acute fever of unknown etiology in children aged 0-5 years, an axillary temperature ≥38℃ is defined as fever. Skin temperature is slightly lower than axillary temperature; 38℃ is used as the warning value, and exceeding this value indicates fever.

[0070] Judgment criteria: If the skin temperature is ≥38℃, a fever alarm is triggered. Fever may be caused by infection, heatstroke, etc., and it is necessary to inform a guardian or seek medical attention promptly. The system will trigger a level two alarm (vibration + information transmission).

[0071] Based on the above, hypothermia is categorized into three levels: mild, moderate, and severe, corresponding to early warning, alarm, and emergency alarm, respectively. It also provides a tiered response mechanism by predicting the rate of temperature drop in advance. Using the temperature difference between the body surface and the environment as an auxiliary criterion for severe hypothermia avoids misjudgments caused by individual sensor drift. Fever determination is simple and clear, forming a complete body temperature anomaly monitoring system together with hypothermia, enhancing safety and protection capabilities.

[0072] Both hypertension and hypotension are significant risk factors for cardiovascular and cerebrovascular accidents, but existing smart wearable devices lack sufficient accuracy in blood pressure measurement, making it difficult to provide graded warnings. Therefore, in one embodiment, the identification of abnormal blood pressure includes: The severity of low blood pressure and high blood pressure is classified according to the different ranges of systolic and diastolic blood pressure.

[0073] Each grading standard is based on the relevant Chinese guidelines for the prevention and treatment of hypertension and the expert consensus on the diagnosis and management of hypotension.

[0074] For example: Hypotension classification: Mild hypotension: Systolic blood pressure <90 mmHg Moderate hypotension: systolic blood pressure <80 mmHg Severe hypotension: Systolic blood pressure <70 mmHg Hypertension classification: Mild hypertension: 140 mmHg < systolic blood pressure ≤ 160 mmHg Moderate hypertension: 160 mmHg < systolic blood pressure < 180 mmHg Severe hypertension: systolic blood pressure ≥180 mmHg or diastolic blood pressure ≥120 mmHg Real-time systolic blood pressure (SBP) and diastolic blood pressure (DBP) were measured using the aforementioned dynamic calibration method (updated every 5 minutes). When any level of abnormality is detected, different levels of alarms are triggered according to their severity: Mild hypotension / mild hypertension: Local recording only, and a prompt message will be displayed on the screen (non-emergency).

[0075] Moderate hypotension / moderate hypertension: Triggers a Level 1 alarm (vibration + red flashing on the screen).

[0076] Severe hypotension / severe hypertension: Triggers a level 2 alarm (audio-visual + vibration + alarm message). In addition, the system monitors the duration of the abnormal blood pressure: if systolic blood pressure ≥180 mmHg persists for more than 5 minutes, the alarm level is automatically escalated to emergency, and the highest priority alarm is immediately sent to the bound terminal.

[0077] Based on dynamically calibrated, high-precision blood pressure data, clinical-level grading and identification of blood pressure abnormalities are achieved. This not only alerts wearers to seek medical attention promptly but also provides quantitative data for remote monitoring (such as recording the frequency, duration, and severity of blood pressure abnormalities), effectively preventing stroke, myocardial infarction, or shock caused by extreme blood pressure.

[0078] Poisoning from harmful gases (such as carbon monoxide and hydrogen sulfide) often occurs silently, initially manifesting as changes in respiratory rate, abnormal heart rate, and decreased blood oxygenation. Since environmental sensors are not always carried on the person, wearable devices lack a direct means of detecting toxic gases. Therefore, in one embodiment, the identification of harmful gas poisoning includes: Respiratory rate is acquired based on piezoelectric pressure signal, and heart rate and blood oxygen saturation are acquired based on PPG optical signal. The ratio of heart rate to respiratory rate is calculated. When the respiratory rate is below the bradycardia threshold or above the tachycardia threshold, and the heart rate is above the toxic tachycardia threshold or below the toxic bradycardia threshold, and the ratio of heart rate to respiratory rate is below the low heart rate-to-respiratory ratio threshold or above the high heart rate-to-respiratory ratio threshold, and the blood oxygen saturation is below the toxic blood oxygen threshold and continues to decrease beyond the toxic blood oxygen decrease rate threshold, it is determined to be toxic gas poisoning.

[0079] Specifically, respiratory rate (RR, in breaths per minute) is obtained by bandpass filtering the signal from the piezoelectric pressure sensor array (passband 0.1Hz–0.5Hz, corresponding to 6–30 breaths per minute) and extracting the respiratory waveform. The number of peaks (or zero-crossings) within a 1-minute time window is then counted to determine the respiratory rate. Heart rate and oxygen saturation (SpO2) are obtained from the PPG signal using the same method.

[0080] Secondly, the criteria for setting each threshold are as follows: Bradypneumonia threshold: 10 breaths / minute. The resting respiratory rate in healthy adults is 12–20 breaths / minute. A rate below 10 breaths / minute indicates severe respiratory depression, commonly seen in opioid poisoning, the later stages of carbon monoxide poisoning, or central respiratory failure. See also: Indicators of Respiratory Failure in Emergency Medicine; The threshold for tachypnea is 30 breaths per minute. A respiratory rate higher than 30 breaths per minute is considered severe tachypnea, seen in hypoxia compensation, metabolic acidosis, or poisoning by irritant gases. The threshold for tachycardia due to poisoning is 140 beats per minute (150 beats per minute for children). In the early stages of poisoning by carbon monoxide, hydrogen sulfide, etc., due to tissue hypoxia, the heart rate increases compensatorily, and a heart rate exceeding 140 beats per minute indicates severe stress. The threshold for bradycardia due to poisoning is 45 beats per minute. Certain nerve gases or high concentrations of hydrogen sulfide can directly inhibit the cardiac conduction system, leading to a decrease in heart rate. Heart rate to respiratory ratio low threshold: 2. A normal heart rate to respiratory ratio is approximately 4:1 (HR 75 bpm, RR 18 bpm). A ratio below 2 indicates rapid breathing relative to heart rate, suggesting abnormal respiratory drive (such as central hyperventilation). Heart rate to respiratory rate ratio (HRRR) threshold: 5. A ratio higher than 5 indicates that the heart rate is too fast relative to the respiratory rate, which is common during the hypoxia compensation period (the increase in heart rate exceeds the increase in respiratory rate). The blood oxygen threshold for poisoning is 90%. Blood oxygen levels below 90% indicate hypoxemia, requiring immediate intervention. The threshold for the rate of decrease in blood oxygen saturation due to poisoning is 2% per minute. A decrease rate exceeding 2% per minute (i.e., a decrease of ≥2% within 1 minute) indicates a rapidly progressing pathological process, such as inhalation of high concentrations of carbon monoxide or rapid depletion of oxygen in a confined space.

[0081] The condition for judgment is that all four of the following conditions (logical AND) must be met simultaneously: Abnormal respiratory rate: <10 breaths / minute or >30 breaths / minute; Abnormal heart rate: heart rate >140 beats / minute or <45 beats / minute; Abnormal heart rate-to-respiration ratio: The ratio of respiratory rate to heart rate is <2 or >5; Abnormal blood oxygenation: Blood oxygen concentration <90% and blood oxygen concentration decreased by ≥2% in the last minute (this can be determined by continuously comparing the current value with the value 1 minute ago).

[0082] For example, in carbon monoxide poisoning, the respiratory rate may increase compensatorily in the early stages (>30), the heart rate may increase (>140), and the heart rate-to-respiratory ratio may be close to normal or slightly low. However, blood oxygen saturation is normal or slightly decreased due to the formation of carboxyhemoglobin. This condition requires SpO2 <90% and a rapid decrease; in actual severe carbon monoxide poisoning, SpO2 can drop below 80%. Hydrogen sulfide poisoning is more typical: respiratory depression (<10), bradycardia (<45), abnormal heart rate-to-respiratory ratio, and a rapid decrease in blood oxygen. This combination of conditions can effectively cover the physiological patterns of poisoning by various common harmful gases.

[0083] Based on the above, by utilizing the respiration detection function of piezoelectric sensors and combining it with heart rate and blood oxygen data from PPG sensors, abnormal patterns across four dimensions—respiratory frequency, heart rate, heart rate-to-respiration ratio, absolute blood oxygen value, and its rate of decline—can be used to indirectly identify harmful gas poisoning. This indirect detection method expands the protective range of wearable devices, providing additional safety guarantees for scenarios such as fieldwork, mines, confined spaces, and firefighting. Once poisoning is detected, the highest-level emergency alarm will be triggered immediately, buying precious rescue time.

[0084] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps.

[0085] Secondly, based on the same inventive concept, this application also provides a watch for implementing the aforementioned intelligent safety monitoring method integrating multi-parameter health monitoring and emergency alarm location. The solution provided by this watch is similar to the implementation described in the above method; therefore, the specific limitations of one or more watch embodiments provided below can be found in the above-described limitations of the intelligent safety monitoring method integrating multi-parameter health monitoring and emergency alarm location, and will not be repeated here.

[0086] In one embodiment, such as Figure 3 As shown, the watch includes: a case, a strap, a main control module, a sensor module, a communication and positioning module, an emergency alarm module, and a power management module. Among them: Watch case 1, the bottom surface of which is arc-shaped and matches the direction of the radial artery (an OLED touch screen can be installed on watch case 1). Watch strap 2, installed on the watch case; The main control module, located inside the watch case, is used to execute the aforementioned intelligent safety monitoring method that integrates multi-parameter health monitoring and emergency alarm location (see above for details, which will not be repeated here). The sensor module includes a PPG optical sensor, a piezoelectric pressure sensor array, an accelerometer and a gyroscope mounted on the watch case, and a flexible temperature sensor embedded inside the watch strap. The communication and positioning module includes a satellite positioning submodule and a wireless communication submodule installed inside the casing; The emergency alarm module includes a physical SOS button 4, a speaker 5, and a vibration motor (not shown in the figure) located on the case. The power management module is located inside the watch case.

[0087] Specifically, the watch case is made of lightweight, wear-resistant engineering plastic with a curved bottom surface that matches the radial artery in the wrist. The case interior features a multi-layered stacked structure and multiple sensor interfaces. A physical SOS button is located on the side of the case. The curved bottom surface design ensures a close fit to the wrist when worn, maintaining good contact between the sensors and the radial artery.

[0088] The watch band is made of medical-grade silicone and is attached to both ends of the watch case. A flexible temperature sensor (such as a Kyocera KT15 series polyimide thermistor) is embedded inside the band. The band is adjustable to fit different users (children / adults).

[0089] Main control module: Located inside the watch case, it includes a processor and an AI acceleration unit (e.g., a Rockchip RV1103B integrated chip, containing an ARM Cortex-M7 MCU and a 0.5 TOPS NPU). The main control module is configured to execute any of the aforementioned intelligent safety monitoring methods for integrated multi-parameter health monitoring and emergency alarm location. The main control module also includes a storage unit (for storing the wearer's physiological baseline data, AI model parameters, and historical monitoring data) and a Bluetooth / StarFlash communication submodule.

[0090] The sensor module includes a PPG optical sensor (e.g., Silicon Labs SCL3300) mounted on the watch case, a piezoelectric pressure sensor array, an accelerometer and gyroscope (e.g., Bosch BMI160), and a flexible temperature sensor (e.g., Kyocera KT15) mounted on the strap. The piezoelectric pressure sensor array consists of six piezoelectric units arranged along the radial artery arc along the inner side of the watch case, with a 2mm spacing between adjacent units. Each piezoelectric unit is made of flexible piezoelectric material (commercially available PVDF piezoelectric film or MXene / BN / PVDF-TrFE nanocomposite piezoelectric material can be used). The accelerometer and gyroscope are mounted close to the main control module to reduce signal interference. The PPG optical sensor is located in the center of the back of the watch case, surrounded by a light-shielding structure.

[0091] Communication and positioning modules: Located inside the watch case, these include a satellite positioning submodule (BeiDou / GPS dual-mode, such as the Shanghai Hisilicon Hi332X, positioning sensitivity -162dBm, accuracy ≤5 meters) and a wireless communication submodule (including an eSIM cellular communication submodule supporting 4G Cat1 networks, such as the Hi2871W; and a Bluetooth / StarFlash 2.0 communication submodule with a transmission rate of 16Mbps). The satellite positioning submodule refreshes positioning data in seconds under dangerous conditions.

[0092] Emergency alarm module: Located on the case, it includes a physical SOS button (side), a miniature speaker (capable of emitting ≥80dB audible and visual alarms) and a vibration motor (to achieve local tactile alarm), supporting both automatic and manual trigger modes.

[0093] Power management module: Located inside the casing, it includes a 200-400mAh soft-pack lithium polymer battery, a fast-charging chip (e.g., SGM4056 from SG Micro), and an intelligent power management unit. It supports 5V / 1A fast charging and features an intelligent sleep mode and a time-segmented sampling mechanism. The communication module's standby power consumption is ≤180μA, and the device can standby for ≥7 days. The main control module is configured to support silent mode (wakes up every 10 minutes, using a low-power ADC to read data) and event-triggered mode (runs the NPU at full power when a fall or sudden heart rate change is detected, increasing the sampling frequency to over 100Hz).

[0094] The watch determines its fit by monitoring the substrate impedance or static pressure distribution of the piezoelectric pressure sensor array. If a loose fit is detected (due to impedance abrupt changes or uneven pressure), the watch automatically pauses blood pressure measurement and alerts the user.

[0095] The main control module, sensor module, communication and positioning module, emergency alarm system, and power management module are electrically connected via flexible circuit boards or internal wires, forming a closed-loop system for multi-parameter acquisition, local analysis, hazard identification, and emergency alarm. Specifically, the sensor module transmits the acquired signals to the main control module via I2C or SPI bus; the main control module processes the signals and controls the emergency alarm system via GPIO; the communication and positioning module interacts with the main control module via UART; and the power management module supplies power to all modules. The above connection methods are all conventional techniques in this field and will not be elaborated upon in this embodiment.

[0096] Based on the above, this watch ensures signal acquisition quality through an optimized structural layout (curved bottom surface, piezoelectric array adapted to the radial artery); achieves low-latency hazard identification through local AI processing; guarantees accurate transmission of alarm information through dual-mode positioning and cellular communication; and achieves long battery life through low-power management. Overall, this watch can accurately monitor multiple health parameters and automatically trigger emergency alarms when danger occurs, making it suitable for high-risk individuals.

[0097] In one embodiment, the main control module includes a fall detection unit and a fall detection unit.

[0098] The fall detection unit calculates synthetic acceleration and attitude angle change based on motion posture signals and extracts the peak impact acceleration. A fall is identified when the peak impact acceleration exceeds the impact acceleration fall threshold, the attitude angle change exceeds the attitude angle change threshold, and the post-fall attitude stabilization time exceeds the attitude stabilization time threshold. The fall detection unit identifies a fall when the peak impact acceleration exceeds the impact acceleration fall threshold and the duration of this excess exceeds the impact acceleration duration threshold.

[0099] In one embodiment, the main control module includes a syncope recognition unit.

[0100] The syncope recognition unit is used to calculate heart rate variability based on PPG optical signals. Syncope is determined when the heart rate is below the severe bradycardia threshold or above the severe tachycardia threshold and the heart rate variability is below the heart rate variability threshold, or when the blood oxygen saturation is below the severe hypoxemia threshold and the posture angle obtained based on the motion posture signal is below the supine posture angle threshold and persists for more than the supine duration threshold. In one embodiment, the main control module includes a hypoxia detection unit.

[0101] The hypoxia identification unit is used to calculate the rate of change of blood oxygen saturation and the magnitude of heart rate increase based on PPG optical signals. When blood oxygen saturation is below the low blood oxygen threshold for a period of time exceeding the low blood oxygen duration threshold, or when blood oxygen saturation is between the lower limit and the upper limit of the low blood oxygen range, the decrease in blood oxygen saturation exceeds the blood oxygen decrease threshold, and the increase in heart rate exceeds the heart rate compensation increase threshold, hypoxia is determined.

[0102] In one embodiment, the main control module includes a temperature loss detection unit and a fever detection unit.

[0103] The hypothermia detection unit calculates the skin temperature change rate based on the body surface temperature signal and the difference between the skin temperature and the ambient temperature. A mild hypothermia warning is issued when the skin temperature is between the lower and upper limits of the mild hypothermia range and the skin temperature decrease rate exceeds the hypothermia warning change rate threshold. A moderate hypothermia alarm is issued when the skin temperature is between the lower and upper limits of the moderate hypothermia range. A severe hypothermia alarm is issued when the skin temperature is below the severe hypothermia threshold and the difference between the skin temperature and the ambient temperature is below the severe hypothermia temperature difference threshold. The fever detection unit determines a fever when the skin temperature exceeds the fever threshold.

[0104] In one embodiment, the main control module includes a blood pressure abnormality detection unit.

[0105] The blood pressure abnormality identification unit is used to classify the severity of hypotension and hypertension based on different ranges of systolic and diastolic blood pressure.

[0106] In one embodiment, the main control module includes a hazardous gas poisoning detection unit.

[0107] The hazardous gas poisoning identification unit is used to collect respiratory rate based on piezoelectric pressure signal, collect heart rate and blood oxygen saturation based on PPG optical signal, and calculate the ratio of heart rate to respiratory rate. When the respiratory rate is lower than the bradycardia threshold or higher than the tachycardia threshold, and the heart rate is higher than the poisoning tachycardia threshold or lower than the poisoning bradycardia threshold, and the ratio of heart rate to respiratory rate is lower than the low heart rate-to-respiratory ratio threshold or higher than the high heart rate-to-respiratory ratio threshold, and the blood oxygen saturation is lower than the poisoning blood oxygen threshold and continues to decrease beyond the poisoning blood oxygen decrease rate threshold, it is determined to be hazardous gas poisoning.

[0108] The various modules in the aforementioned smart safety monitoring watch, which integrates multi-parameter health monitoring and emergency alarm positioning, can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0109] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4 As shown, the computer device includes a processor, memory, and network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores all data required for the watch's operation. The network interface communicates with external terminals via a network connection. When executed by the processor, the computer program implements an intelligent safety monitoring method integrating multi-parameter health monitoring and emergency alarm location.

[0110] Those skilled in the art will understand that Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0111] Thirdly, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of any of the methods described in the above method embodiments.

[0112] Fourthly, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of any of the methods described in the above method embodiments.

[0113] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0114] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0115] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. An intelligent safety monitoring method integrating multi-parameter health monitoring and emergency alarm location, characterized in that, The method includes: The wearer's multi-parameter signals are collected simultaneously, including PPG optical signals, piezoelectric force signals, motion posture signals, and body surface temperature signals. The mapping relationship between pulse wave conduction time and blood pressure is dynamically calibrated based on the wearer's historical resting data, and physiological parameters are calculated using the mapping relationship based on the currently collected multi-parameter signals. Physiological and motion features are extracted from the multi-parameter signals, and the physiological and motion features are input into a lightweight AI model for local real-time inference. A hazard index threshold algorithm is used to identify the probability of a dangerous state. The dangerous state includes at least one of the following: fall, collapse, fainting, hypoxia, abnormal blood pressure, hypothermia, fever, or poisoning by harmful gases. When the probability exceeds a preset threshold and the duration of the dangerous state exceeds a set time, or when an active triggering signal is detected, an emergency alarm process is triggered; the emergency alarm process is to activate local sound, light and vibration alarms, activate satellite positioning, and send alarm information to the bound terminal, the alarm information including positioning data and physiological data summary; If the emergency alarm process is completed and a preset false alarm cancellation operation is received, the local audible and visual alarm and the alarm information are stopped, and the current physiological parameters are input into the lightweight AI model for false alarm learning; if the emergency alarm process is completed, the dangerous state continues, and the false alarm cancellation operation is not received, the local audible and visual alarm and the alarm information are repeated at set intervals.

2. The method according to claim 1, characterized in that, The identification of the fall and the drop includes: Based on the motion attitude signal, the synthetic acceleration and attitude angle change are calculated, and the peak value of the impact acceleration is extracted from the synthetic acceleration. A fall is determined when the peak impact acceleration exceeds the impact acceleration fall threshold, the change in attitude angle exceeds the attitude angle change threshold, and the duration of attitude stabilization after the fall exceeds the attitude stabilization duration threshold. When the peak value of the impact acceleration exceeds the impact acceleration fall threshold and the duration of the excess exceeds the impact acceleration duration threshold, it is determined to be a fall.

3. The method according to claim 1, characterized in that, The identification of syncope includes: Heart rate variability is calculated based on the PPG optical signal; Syncope is defined as follows: when the heart rate is below the severe bradycardia heart rate threshold or above the severe tachycardia heart rate threshold and the heart rate variability is below the heart rate variability threshold; or when the blood oxygen saturation is below the severe hypoxemia threshold and the posture angle obtained based on the movement posture signal is below the supine posture angle threshold for a continuous period exceeding the supine duration threshold.

4. The method according to claim 1, characterized in that, The identification of hypoxia includes: The rate of change in blood oxygen saturation and the magnitude of heart rate increase were calculated based on the PPG optical signal. When the blood oxygen saturation is below the low blood oxygen threshold for a sustained period exceeding the low blood oxygen duration threshold, or when the blood oxygen saturation is between the lower limit and the upper limit of the low blood oxygen range, the decrease in blood oxygen saturation exceeds the blood oxygen decrease threshold, and the increase in heart rate exceeds the heart rate compensation increase threshold, it is determined to be hypoxia.

5. The method according to claim 1, characterized in that, The identification of the hypothermia and fever includes: The skin temperature change rate is calculated based on the surface temperature signal, and the difference between the skin temperature and the ambient temperature is calculated. When the skin temperature is between the lower limit and the upper limit of the mild hypothermia range and the rate of decrease in skin temperature exceeds the hypothermia warning change rate threshold, it is determined to be a mild hypothermia warning. When the skin temperature is between the lower limit and the upper limit of the moderate hypothermia range, a moderate hypothermia alarm is triggered. When the skin temperature is below the severe hypothermia threshold and the difference between the skin temperature and the ambient temperature is below the severe hypothermia temperature difference threshold, a severe hypothermia alarm is triggered. When the skin temperature exceeds the fever threshold, it is determined to be a fever.

6. The method according to claim 1, characterized in that, The identification of blood pressure abnormalities includes: The severity of low blood pressure and high blood pressure is classified according to the different ranges of systolic and diastolic blood pressure.

7. The method according to claim 1, characterized in that, The identification of the harmful gas poisoning includes: The respiratory rate is acquired based on the piezoelectric pressure signal, and the heart rate and blood oxygen saturation are acquired based on the PPG optical signal. The ratio of heart rate to respiratory rate is calculated. When the respiratory rate is below the bradycardia threshold or above the tachycardia threshold, and the heart rate is above the poisoning tachycardia threshold or below the poisoning bradycardia threshold, and the ratio of heart rate to respiratory rate is below the low heart rate-to-respiratory ratio threshold or above the high heart rate-to-respiratory ratio threshold, and the blood oxygen saturation is below the poisoning blood oxygen threshold and continues to decrease beyond the poisoning blood oxygen decrease rate threshold, it is determined to be poisoning by the harmful gas.

8. A smart safety monitoring watch integrating multi-parameter health monitoring and emergency alarm positioning, characterized in that, The watch includes: The watch case has an arc-shaped bottom surface that matches the direction of the radial artery; The watch strap is mounted on the watch case; The main control module is located inside the watch case and is used to execute the intelligent safety monitoring method integrating multi-parameter health monitoring and emergency alarm positioning as described in any one of claims 1 to 7. The sensor module includes a PPG optical sensor, a piezoelectric pressure sensor array, an accelerometer and a gyroscope disposed on the watch case, and a flexible temperature sensor embedded inside the watch strap. The communication and positioning module includes a satellite positioning submodule and a wireless communication submodule disposed within the casing; The emergency alarm module includes a physical SOS button, a speaker, and a vibration motor located on the case. The power management module is located inside the watch case.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the intelligent safety monitoring method for integrated multi-parameter health monitoring and emergency alarm positioning as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the intelligent safety monitoring method for integrated multi-parameter health monitoring and emergency alarm location as described in any one of claims 1 to 7.

Images