A smart watch for safety monitoring and alarming

Abstract

The application relates to the field of time measurement, and discloses an intelligent watch for safety monitoring and alarming, which comprises a data acquisition and preprocessing module, a risk feature extraction module, an environment evaluation module, a dynamic risk fusion module, a predictive buffer alarm control module and an alarm execution module which are connected in sequence. The system collects physiological and environmental data of a user through multiple sensors, calculates risk factors in four dimensions of heart rate variability, motion contradiction, skin electric response and abnormal stillness, dynamically adjusts the fusion weight of each factor in combination with the real-time evaluated environmental risk level, and generates a comprehensive risk score. The alarm control module executes a hierarchical buffer alarm process including early warning, local delayed alarm and remote alarm according to the score and its short-term prediction value in combination with a dynamic threshold. Through multi-dimensional information fusion and environment adaptive decision, the application can more accurately distinguish physiological interference from real risks, reduces false alarms, and ensures timely response to real risks.

Description

Technical Field

[0001] This invention relates to the field of time measurement technology, specifically to a smartwatch for safety monitoring and alarms. Background Technology

[0002] Currently, smartwatches with safety monitoring functions typically rely on data from one or a few sensors, such as heart rate and acceleration, to determine whether a user has experienced an accident (e.g., excessively high or low heart rate, or a fall). The core of this approach lies in the real-time monitoring of specific sensor signals and comparison with thresholds, triggering an alarm when the threshold is exceeded.

[0003] While the aforementioned technical solutions can provide basic risk alerts, they still have some shortcomings: First, due to their reliance on a single or limited judgment dimension, the robustness of their criteria is poor. For example, the increase in heart rate during daily exercise and pathological tachycardia show similarities in a single heart rate indicator, making it difficult to effectively distinguish them using only a fixed threshold. This directly leads to a high false alarm rate. Second, their "post-event triggered" response mode has inherent delays. The system only initiates the alarm process after physiological indicators exceed the threshold, resulting in an unavoidable time lag between the occurrence of the risk and the issuance of the rescue information, potentially missing the opportunity for early intervention. Furthermore, static judgment rules cannot adapt to complex real-world scenarios. Fixed sensitivity parameters and triggering logic cannot adaptively adjust to the risk differences in the user's environment (such as the safe home versus the unfamiliar outdoors), making it difficult to achieve a good balance between sensitivity and specificity in changing environments. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a smartwatch for security monitoring and alarm, which solves the problems of high false alarm rate, response delay, and lack of scene adaptability in existing smartwatch security monitoring solutions, which are difficult to optimize in a coordinated manner due to reliance on single-dimensional threshold judgment and static response logic.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a smartwatch for security monitoring and alarm, comprising: The data acquisition and preprocessing module is used to simultaneously acquire and process data from multiple sensors; The risk feature extraction module is used to calculate multiple risk factors in different dimensions based on the data processed by the data acquisition and preprocessing module. The environmental assessment module is used to assess the current environmental risk level based on the data processed by the data acquisition and preprocessing module. The dynamic risk fusion module connects the risk feature extraction module and the environmental assessment module. It is used to dynamically adjust the fusion weight of each risk factor according to the current environmental risk level, and fuse the weighted risk factors to generate a real-time comprehensive risk score. The predictive buffer alarm control module, connected to the dynamic risk fusion module and the environmental assessment module, is used to control the alarm state to migrate between multiple levels based on the real-time comprehensive risk score, its predicted value, and decision parameters dynamically determined by the current environmental risk level. The alarm state migration logic includes triggering a low-level warning based on the predicted value before reaching the high-level alarm threshold, and initiating a delayed confirmation process that can be stopped due to user operation or risk mitigation after reaching the high-level alarm threshold. The alarm execution module is connected to the predictive buffer alarm control module and is used to execute alarm actions corresponding to the current alarm state.

[0006] Preferably, the specific steps for the risk feature extraction module to calculate the heart rate variability factor include: Extracting continuous inter-heartbeat sequences from clean pulse wave signals; The standard deviation of the heart rate interval sequence is calculated within a sliding time window to obtain a time-domain index of heart rate variability. The heart rate variability time-domain index calculated in the current window is compared with the user's personal resting baseline value learned through long-term monitoring. The heart rate variability factor is obtained by normalization calculation, and its value increases as the current heart rate variability time-domain index decreases relative to the personal resting baseline value.

[0007] Preferably, the specific steps for the risk feature extraction module to calculate the motion contradiction factor include: Based on the preprocessed acceleration signal, the standard deviation of the acceleration vector amplitude is calculated within a short time window and used as the quantization value of the actual motion intensity. At the same time, the difference between the current instantaneous heart rate and the user's personal resting heart rate is obtained as the relative heart rate load value; The relative heart rate load value and the quantified actual exercise intensity value are input into a preset fusion function. The fusion function is configured to output a high value when the relative heart rate load value is high and the quantified actual exercise intensity value is low, thereby obtaining the exercise contradiction factor.

[0008] Preferably, the specific method for the comprehensive determination of the execution rules by the environmental assessment module is as follows: A lookup table is predefined to map location security level, time anomaly flag, meteorological change flag, and acoustic anomaly flag to discrete environmental risk level; The location security level is determined by comparing the current geographic location with a pre-stored geofence database; The time anomaly flag is determined by comparing the current time, current location, and the user's historical behavior habit model. The meteorological change indicator is determined by judging whether the absolute value of the recent air pressure change rate exceeds a first threshold. The acoustic anomaly indicator is determined by judging whether the current ambient sound pressure level continuously deviates from its short-term dynamic baseline by a certain amount.

[0009] Preferably, the preset weight configuration table in the dynamic risk fusion module is designed as follows: As the environmental risk level increases, the weights assigned to the skin conductance response factor and the abnormal rest factor increase accordingly, while the weights assigned to the heart rate variability factor and the motion paradox factor decrease accordingly, making the system more sensitive to changes in physiological indicators characterizing user stress and mobility in high-risk environments.

[0010] Preferably, the dynamic risk fusion module calculates the real-time comprehensive risk score. The specific formula is: In the formula, , , , These respectively represent the heart rate variability factor, the motion paradox factor, the skin conductance response factor, and the abnormal rest factor; , , , These represent the dynamic fusion weights obtained from the weight configuration table, corresponding to the current environmental risk level, and satisfying the following conditions: + + + =1.

[0011] Preferably, the predictive buffer alarm control module uses a first-order exponential smoothing method to predict the short-term trend of the real-time comprehensive risk score. The calculation formula is: In the formula, This provides a real-time comprehensive risk score for the current moment. The predicted value calculated at the previous time step. A smoothing factor between 0 and 1. This is the preset prediction step time.

[0012] Preferably, the finite state machine in the predictive buffer alarm control module includes at least a monitoring state, an early warning state, a local alarm state, and a remote alarm state; The monitoring status is maintained when the real-time comprehensive risk score is lower than the warning threshold. When the real-time comprehensive risk score reaches or exceeds the warning threshold, the warning state is entered. In the warning state, if the predicted score reaches or exceeds the alarm threshold, it will immediately transition to the local alarm state. Upon entering the local alarm state, a high-level local alarm control command is immediately triggered and the delayed confirmation timer is started; In the local alarm state, if a user cancellation signal is received before the timer expires or the real-time comprehensive risk score remains below the alarm threshold, the system will exit to the monitoring state or the early warning state. If the timer times out and the real-time comprehensive risk score is still higher than the alarm threshold, then the system will transition to the remote alarm state.

[0013] Preferably, the predictive buffer alarm control module dynamically sets the warning threshold and alarm threshold in the following way: A threshold lookup table is pre-stored, mapping different environmental risk levels to different warning and alarm thresholds; The duration of the delayed confirmation timer is also dynamically determined based on the current environmental risk level through a duration lookup table in a similar manner.

[0014] Preferably, the remote alarm data packet encapsulated by the alarm execution module further includes at least one of the following fields: The alarm is triggered by the heart rate value, the identification of the main triggering risk factors, the device battery level information, and a brief historical trend of the real-time comprehensive risk score over a recent period.

[0015] This invention provides a smartwatch for security monitoring and alarm functions. It has the following beneficial effects: 1. This invention, through the parallel calculation and fusion of multi-dimensional risk factors, can more comprehensively assess a user's safety status. Specifically, the system simultaneously calculates heart rate variability, motion paradox, skin conductance response, and abnormal resting factors, and dynamically adjusts the fusion weight of each factor based on the environmental risk level. This multi-dimensional information fusion and dynamic weighting mechanism enables the system to cross-validate different physiological signals, effectively distinguishing between instantaneous indicator fluctuations caused by physiological reasons such as exercise and genuine pathological risk signals, thereby reducing false alarms that may arise from judging by a single indicator.

[0016] 2. This invention optimizes the timeliness and user experience of alarms by introducing a tiered buffered alarm mechanism based on short-term trend prediction. The alarm control module predicts the real-time comprehensive risk score and triggers only a non-intrusive prompt when the score exceeds the warning threshold but the predicted risk does not increase sharply. When the score reaches the alarm threshold, the system initiates a delayed confirmation process that can be canceled by the user or terminated due to risk mitigation. This design enables the system to provide early warnings of risks and provides users with a time window to handle brief anomalies or false alarms, thereby reducing unnecessary disturbances while ensuring a final alarm response to real and ongoing risks.

[0017] 3. This invention achieves scene-adaptive sensitivity of the system by deeply embedding environmental risk assessment into the decision chain. The system assesses the environmental risk level based on multi-source information such as geographical location, time, and sound, and dynamically adjusts the weight of risk fusion, alarm trigger threshold, and confirmation delay accordingly. In high-risk environments, the system automatically increases its sensitivity to signals such as tension and stillness and shortens the response delay; in safe environments, a relatively lenient strategy is adopted to reduce interference. This allows the system to adapt to various scenarios such as home, outdoors, and unfamiliar areas, avoiding the problem of fixed parameters exhibiting unbalanced performance in different environments. Attached Figure Description

[0018] Figure 1 This is a block diagram of the overall system structure of the present invention.

[0019] The modules are as follows: 101. Data acquisition and preprocessing module; 102. Risk feature extraction module; 103. Environmental assessment module; 104. Fusion scoring module; 105. Alarm control module; 106. Alarm execution module. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0021] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0022] Example: Please see the appendix Figure 1 This invention provides a smartwatch for security monitoring and alarms, comprising: The data acquisition and preprocessing module 101 connects to various sensors built into the smartwatch to acquire and preprocess raw sensor data.

[0023] The sensor suite includes at least one photoplethysmography (PPG) sensor for detecting changes in blood flow volume, a triaxial accelerometer for detecting linear acceleration, a triaxial gyroscope for detecting angular rate, a skin conductance sensor for detecting changes in skin conductivity, a GPS receiver module for acquiring geographic coordinates, a barometer for measuring atmospheric pressure, and a microphone for collecting ambient sound. For sensor selection, the PPG sensor can be chosen from... 86150 or similar integrated optical sensors to support multi-wavelength measurements; triaxial accelerometers and gyroscopes are optional. or A six-axis inertial measurement unit (IMU) is used to provide synchronized acceleration and angular velocity data; a skin conductance sensor can be employed. Or, similar to a bioimpedance measurement front end, to measure skin conductance with high precision, the data acquisition and preprocessing module 101 performs synchronous acquisition of multi-sensor data and performs targeted preliminary processing on the acquired raw signals to generate a set of well-defined, time-aligned standardized data for use by subsequent analysis modules.

[0024] To achieve time synchronization of multi-source data, the data acquisition and preprocessing module 101 adopts a synchronous acquisition mechanism based on hardware timer interrupts. The system is configured with a high-precision timer with a fixed sampling period. An interrupt signal is generated. Typically, The sampling rate is set to 0.01 seconds (i.e., 100Hz). This frequency is sufficient to capture the highest frequency components of signals such as ECG, pulse wave, acceleration, and skin conductance, while also considering system power consumption. At each interruption... The processor reads the data registers of all sensors participating in the synchronous acquisition and marks this set of data as being from the same moment. The sampling. This moment is determined by the system clock reference. and sampling sequence number The decision, their relationship is as follows: In this way, raw readings from different sensors are assigned a unified timescale, forming an initial dataset. .

[0025] The data acquisition and preprocessing module 101 also includes a circular buffer for storing standardized data from a recent time period (e.g., 30 seconds). The design of this buffer allows the subsequent risk feature extraction module 102 and environmental assessment module 103 to flexibly access historical data windows of different time lengths according to their respective algorithm requirements, without having to repeatedly collect raw data. For example, the calculation of heart rate variability factor requires a 5-minute RR interval sequence, which can be accessed by accessing the clean pulse wave signal stored in the circular buffer. Perform dynamic extraction and calculation.

[0026] For the raw signal output by the photoplethysmography (PPG) sensor The data acquisition and preprocessing module 101 first performs bandpass filtering to retain the frequency band related to heart rate and remove DC offset and high-frequency noise. This digital filtering operation can be expressed as: in, and These are filter coefficients. and This refers to the filter order. The passband range of this filter is set to 0.5 Hz to 5 Hz, and its specific design is well-known in the art. To reduce artifacts caused by body movement on the pulse wave signal, the data acquisition and preprocessing module 101 employs an adaptive filtering algorithm. The preprocessed accelerometer signal... (As a motion reference) and the bandpass filtered pulse wave signal (As the main input) it is processed. One implementation is to use the normalized least mean square algorithm, which iteratively updates the filter weight vector through the following steps. And output a clean signal: Calculate the filter output, which is the estimate of the moving reference signal: .

[0027] Calculate the error signal, which is the difference between the estimated motion artifact and the original input: .

[0028] Update filter weights: .in, It is a convergent step size. It is a small positive number that prevents the denominator from being zero.

[0029] Initialization weight vector of adaptive filter It can be set to an all-zero vector, with a convergence step size. A typical value of 0.01 is chosen to ensure that the algorithm converges quickly within tens of milliseconds. The algorithm iteratively estimates the motion artifact components related to the pulse wave signal in the acceleration signal and subtracts them from the main input, ultimately obtaining a clean pulse wave signal with significantly suppressed motion artifacts. .

[0030] After adaptive filtering, a clean pulse wave signal with significantly suppressed motion artifacts is obtained. .

[0031] For a triaxial accelerometer, the raw output data is the acceleration values ​​on three orthogonal axes. The data acquisition and preprocessing module 101 needs to separate the linear acceleration generated by body motion from the signal. This is typically done by using a high-pass filter to remove the component representing the direction of static gravity. Linear acceleration It can be obtained through the following digital filter operation: in This represents the transfer function of a first-order or second-order high-pass filter with a cutoff frequency of approximately 0.1 Hz to 0.3 Hz. This filtering operation eliminates the long-term effects of gravity, making... Primarily reflects the user's active movements. Gyroscope raw data. After calibration and unit conversion, the output is directly given as an angular velocity value. Gyroscope data is primarily used to assist in determining user posture (e.g., estimating direction changes through angular velocity integral) or to provide additional motion information in specific scenarios such as fall detection, but it is not a mandatory input in the risk factor calculation of this invention.

[0032] For the skin conductance sensor, its raw signal This is a time-varying measurement of skin conductance. The data acquisition and preprocessing module 101 decomposes it into two components: slowly changing skin conductance levels. and rapidly changing skin conductance responses This can be achieved using a set of filters connected in parallel. Through a low-pass filter with an extremely low cutoff frequency (e.g., 0.01 Hz) extract: Then it can be passed through a bandpass filter Extract (e.g., with a passband of 0.05 Hz to 0.5 Hz), or by subtracting from the original signal. Then, bandpass filtering is performed to obtain: Skin conductance response The amplitude and frequency of sympathetic nerve activity are important indicators for measuring sympathetic nerve activity.

[0033] For environmental sensor data, the data acquisition and preprocessing module 101 performs basic parsing and smoothing. (Global Positioning System coordinates) After Kalman filtering and smoothing to suppress positioning jitter, stable coordinates are output. Barometer reading After passing through a moving average filter to reduce instantaneous noise, within the window length The average value within is: Here, the length of the sliding window The option to select 10 corresponds to a smoothing time of 1 second, which can effectively suppress high-frequency noise from the barometric pressure sensor and quickly respond to sudden changes in ambient air pressure.

[0034] microphone audio signal Sound pressure level converted to decibel scale Its calculation involves a sliding window average of the squared values ​​of the audio signal and taking the logarithm, the specific implementation of which is well-known in the field. Network connectivity status. The system timestamp is resolved to a recognizable service set identifier. The system timestamp is then converted to local time. .

[0035] After the above-mentioned synchronous acquisition and specialized preprocessing of each signal path, the data acquisition and preprocessing module 101 outputs a set of standardized, time-base aligned data. : This dataset serves as the direct input for subsequent risk feature extraction and environmental assessment.

[0036] The risk feature extraction module 102 is connected to the data acquisition and preprocessing module 101 and is used to calculate risk factors in multiple dimensions, including heart rate variability factor, motion paradox factor, skin conductance response factor and abnormal rest factor.

[0037] The core function of this module is to calculate four independent, normalized risk factors in real time from the input physiological and motor signals. These factors are designed to quantify the risk of physiological abnormalities in different dimensions. The output value of each factor is limited to a closed interval between 0 and 1, with a higher value indicating a higher risk in that dimension.

[0038] The first risk factor is heart rate variability. It is used to assess the balance state of the autonomic nervous system. The risk feature extraction module 102 first extracts features from clean pulse wave signals. Extract continuous Interval sequence. Specifically, this method involves detecting the peak points of the pulse waveform and calculating the time interval between adjacent peaks. The unit is milliseconds. Peak detection can employ an algorithm combining dynamic thresholding and adaptive windowing: firstly, ... Perform a first-order difference to find all positive zero-crossing points as candidate peaks; then, based on the most recent few valid peaks... The search window width is dynamically adjusted based on the average value of the intervals to eliminate false positives; finally, the final peak is confirmed by checking whether the amplitude of the candidate peak exceeds a specific proportion (e.g., 0.6) of the current signal's average amplitude. For a length of... Within a sliding time window (e.g., 5 minutes) Interphase series, calculate their standard deviation : in, It is inside the window Number of intervals These are The average value of the interval. It reflects the time-domain characteristics of heart rate variability. This factor is compared with the user's individual resting baseline. Calculations are performed by comparison. Calculated within multiple time windows when the user is in a calm, conscious state. The average value is then calculated, and the system can learn this value during the initialization phase. The specific learning process is as follows: Within 24 hours of the user's first use or manual triggering of "learning mode," the system automatically identifies and marks periods when the user is not exercising, not sleeping, and has a stable heart rate (e.g., determined by a combination of accelerometer and heart rate data), and calculates this every 10 minutes during these periods. After removing outliers exceeding the mean ± 2 standard deviations, the average of the remaining data is taken as the mean. This baseline value is stored in non-volatile memory and supports periodic (e.g., monthly) or user-triggered online updates. Heart rate variability factor The calculation formula is: The formula will Limited to the range [0, 1]. When When the factor value is close to or above the baseline value, it is close to 0, indicating low risk; when When the value is significantly lower than the baseline, the factor value approaches 1, indicating a high risk.

[0039] The second risk factor is the motion contradiction factor. It is used to identify mismatches between heart rate and physical activity levels. Module 102 simultaneously calculates two sub-features: actual motion intensity based on acceleration signals. and relative load based on heart rate signal The actual motion intensity is determined by the linear acceleration vector. The length of the module in a short time window The standard deviation is calculated within (e.g., 30 seconds) to obtain: Here, `std` represents the standard deviation operation. Standard deviation, rather than the mean, is used to quantify exercise intensity because it better reflects the magnitude of changes in acceleration signals, thus more accurately characterizing the dynamics of physical activity, such as walking, running, or trembling. Relative load. Current instantaneous heart rate user's resting heart rate The difference. Available from The interval series can be directly converted. These are individual baseline parameters. The acquisition method and Similarly, when the user is completely relaxed, not exercising, sitting, or lying down, the average heart rate over 5 minutes is recorded as the resting heart rate baseline. Exercise-contradictory factors. These two sub-features are combined using a fusion function: in, yes S A type function used to map the output to the interval (0, 1). and It is a positive scaling factor used to adjust the relative contributions of heart rate and exercise intensity to the factor. and Typical values ​​can be determined through fitting empirical data, such as calibration under typical scenarios where users are engaged in moderate-intensity running (high heart rate, high exercise intensity) and emotional stress (high heart rate, low exercise intensity). One approach is to set... =0.3 (times / minute) -1 , =0.8(m / s²) -1 This allows for an increase in heart rate of 50 beats per minute at an exercise intensity of 0.2 m / s², The value is close to 0.9; while at the same heart rate increase but with an exercise intensity of 0.8 m / s², The value drops below 0.5. This occurs when the heart rate is high and physical activity is low. The value is large, resulting in A value close to 1 indicates a high risk of cardiac load that is disproportionate to the amount of exercise.

[0040] The third risk factor is skin conductance factor. It reflects the excitation level of the sympathetic nervous system. The risk feature extraction module 102 obtains skin conductance levels from the preprocessed data. and skin conductance response The calculation of this factor is divided into two parts. The first part is the relative change in skin conductance levels. : It is a dynamic baseline, which can be adjusted by... It is obtained by low-pass filtering over a long time constant (e.g., tens of minutes). The introduction of a dynamic baseline is to accommodate the slow drift of skin conductance levels with environmental temperature, humidity, and individual physiological state, so that... It can reflect a short-term increase in stress. The second part is the activity of the skin's electrical conductance response. By time window Internal Statistics Exceeding the preset threshold Number of peaks To evaluate: in It is a constant used for normalization, for example, within a 60-second window. It can be set to 10, corresponding to a highly active state of skin conductance responses exceeding 10 times per minute. The skin conductance response factor is a weighted sum of these two parts: Weight It is a coefficient between 0 and 1. A typical value for this coefficient is 0.6, meaning that both long-term elevated levels and short-term frequent responses contribute to the final stress factor, but with greater emphasis on the dynamic indicator of response frequency. The calculation results are obtained through... The operation truncates to the interval [0, 1].

[0041] The fourth risk factor is the abnormal quiescent factor. It is used to detect situations where, in a non-sleep state, a user is unable to move due to an underlying problem but experiences physiological stress. Module 102 first determines whether the user is in a "stationary state." This determination is based on a relatively long time window. (For example, within 5 minutes) the actual exercise intensity Continuously below a low threshold If this condition is met, the system proceeds to a static state determination. Under the premise of a static state determination, the module checks for contradictory physiological signals, i.e., simultaneously satisfying... or ,in and It is a preset threshold. and Typical values ​​are 0.7 and 0.6, respectively. These two thresholds are set relatively high to filter out subtle physiological fluctuations in the resting state, triggering the abnormal resting factor only when significant autonomic dysfunction or stress response occurs. The calculation logic of the abnormal resting factor is as follows: This logic ensures that only when low levels of physical activity are accompanied by physiological signals of high stress or autonomic nervous system dysfunction, Only then will a higher risk value be output.

[0042] The risk feature extraction module 102 executes the above calculation process at a fixed period (e.g., once per second) and outputs four real-time updated risk factor values. , , and These factor values ​​are transmitted to the downstream fusion scoring module 104.

[0043] The environmental assessment module 103 is used to assess the safety level of the user's current environment. This environmental risk level is a discrete value.

[0044] The environmental assessment module 103 receives relevant data from the data acquisition and preprocessing module 101, including geographic location coordinates, local time, atmospheric pressure, ambient sound pressure level, and network connectivity information. This module comprehensively analyzes this information and outputs a discrete value representing the potential environmental risk level. . The value ranges from 1 to 4, with higher levels indicating greater environmental risk.

[0045] The environmental assessment module 103 makes a comprehensive judgment by evaluating four sub-rules. The first sub-rule is location security assessment. The module will assess the current... The system compares the geographic coordinates obtained from network positioning with a pre-installed or online geofencing database. The geofencing database is built based on the user's historical activity patterns and user-defined safe zones. For example, users can manually mark "home," "office," "gym," etc., as safe zones using a mobile application and define their geofencing (a circular area with a radius of 100-500 meters). The system also automatically learns the user's frequently occurring locations. Locations are grouped into frequently used areas. For unmarked areas, the location is determined based on distance from known safe zones and area type (via...). The location security level is comprehensively assessed based on data obtained (such as from hospitals, police stations, parks, shopping malls, etc.) and the length of time users stay in the area. The database typically categorizes areas into four levels, from 1 to 4. It classifies areas into categories such as "Safe Home," "Frequently Used Workplace," "Familiar Public Areas," "Unfamiliar Areas," and "Known High-Risk Areas." Based on the comparison results, this sub-rule outputs a location security level. For example, numbers 1 to 4 can be used to represent risks from low to high. If the exact location cannot be obtained, it can be combined with the current connection. Network identifier ( Is it a trusted network (such as a home network)? (This can be used to make auxiliary judgments.)

[0046] The second sub-rule is time anomaly assessment. This assessment combines the current local time (including day of the week and time of day) with the user's historical behavioral habit model. The habit model is built by recording the user's most frequent locations and activity states (e.g., at home, at work, commuting) on ​​different dates and times over the past few weeks. For example, the system can statistically determine that a user is within their home geofence 95% of the time between 8 PM and 10 PM on Tuesdays, thus marking that time period and location as a habitual pattern. The historical behavioral habit model uses a Bayesian estimation method for a given time... The probability distribution of a user's appearance at various locations is calculated, and the location with the highest probability is selected as the expected location. The temporal anomaly assessment outputs a binary flag. When the user's current location (by...) When the initial characterization (or initial characterization) deviates significantly from the typical location predicted by the habit model at the current time (e.g., appearing in an unfamiliar area late at night), it is a sign of a problem. Set to 1 otherwise set to 0.

[0047] The third sub-rule is the meteorological abrupt change index assessment. This assessment is based on barometer data. The module calculates the absolute value of atmospheric pressure changes over a recent period (e.g., the past 10 minutes). For example, it calculates the absolute value of the difference between the current pressure and the pressure 10 minutes ago. If the absolute value of this change exceeds a preset threshold (e.g., 2 hPa), it is considered to pose a potential risk of sudden weather change, at which point the meteorological sudden change index is activated. The output is 1 if the air pressure is high and 0 otherwise. Drastic changes in air pressure often indicate the arrival of severe weather such as storms and strong winds, which may increase the risk to users when they are outdoors.

[0048] The fourth sub-rule is the Acoustic Anomaly Index Assessment. This assessment analyzes the audio signals captured by the ambient microphone. The module continuously calculates the short-term average of the ambient sound pressure level as a dynamic baseline. One implementation method is to use a sliding time window (e.g., 5 minutes) to calculate the moving average of the sound pressure level as the baseline. Then, the current real-time sound pressure level... Compare with that baseline. If A sustained increase in noise level above the baseline (e.g., exceeding 20 decibels) or a sustained decrease in noise level below the baseline with abnormally quiet surroundings may indicate an abnormal environment. A sustained increase in noise level above the baseline may indicate sudden noise (e.g., a car accident, argument, alarm), while a sustained decrease in noise level below the baseline (e.g., silence below background noise levels) may indicate that the user is in abnormal isolation or that the device's audio acquisition is malfunctioning. This sub-rule outputs a binary flag. 1 indicates an abnormal acoustic environment, and 0 indicates normal.

[0049] After obtaining the output of the above four sub-rules , , , Subsequently, the environmental assessment module 103 generates the final environmental risk level through a comprehensive judgment logic. This logic is defined through a rule mapping table. The following table provides a concrete example: Table 1 Environmental Risk Level Determination Table This table defines the range from combinations of subrules to discrete risk levels. The mapping. For example, when a user is in a known high-risk area, regardless of other conditions, It is directly classified as the highest level, level 4. This applies when in an unfamiliar area and there are signs of time anomalies or sudden weather changes. Set to 3. In ordinary public areas where the acoustic environment is abnormal, Also set at 3. In a safe home or office environment, Set to the lowest level, 1. If the above special conditions are not met, It is classified as Level 2 (ordinary).

[0050] Environmental assessment module 103 performs a full assessment and updates the output at a low frequency (e.g., every minute). This environmental risk level It is transmitted to the subsequent fusion scoring module 104 and alarm control module 105 for dynamic adjustment of system parameters.

[0051] The fusion scoring module 104 is connected to both the risk feature extraction module 102 and the environmental assessment module 103. The fusion scoring module 104 dynamically adjusts the fusion weights of each risk factor based on the environmental risk level and calculates and generates a real-time comprehensive risk score.

[0052] The fusion scoring module 104 receives four risk factors output by the risk feature extraction module 102, and the environmental risk level output by the environmental assessment module 103. This module is based on environmental risk levels. The weights of each risk factor are dynamically determined, and the weighted risk factors are summed to generate a real-time comprehensive risk score between 0 and 1. . The higher the value, the higher the overall risk level of the current assessment.

[0053] This module first performs dynamic weight configuration. The system predefines the environmental risk levels for the four risk factors. The associated weighting coefficients are the weights of the heart rate variability factor. Weights of contradictory factors in motion Weights of skin electroreactivity factors and the weight of the abnormal static factor For any given The values ​​of these four weighting coefficients are all non-negative, and their sum is 1, which satisfies the following condition: The design principle of weight allocation is to adjust the system's focus on different physiological risk dimensions under different environmental risks. For example, when the environmental risk level... A higher E value indicates a complex or dangerous external environment for the user. In this case, the system should pay more attention to physiological indicators that may directly reflect the user's sudden tension, stress, or mobility. Therefore, the weights of skin conductance response factors and abnormal resting factors can be relatively increased. The theoretical basis for this design is that in high-risk environments, threats that users may encounter (such as being tracked or experiencing violence) will trigger immediate sympathetic nerve activation (manifested as increased skin conductance response) and stress responses such as "freezing" or inability to move (manifested as high stress in an abnormal resting state). In low-risk environments, the risks that users may face are more related to their own physiological functions (such as heart problems). In this case, heart rate variability and motion paradox are more reflective of the problem. A specific implementation method is to store the weight coefficients corresponding to different E values ​​in a lookup table. The following table is an example: Table 2. Dynamic Weighting of Risk Factors The integrated scoring module 104 is based on the environmental risk level input in real time. Find and obtain the corresponding set of weight coefficients from the table. , , , .

[0054] Subsequently, the module calculates a real-time comprehensive risk score. Suppose at some point... The input risk factor value is , , , The obtained weight is , , , Real-time comprehensive risk score Calculated using the following weighted sum formula: Since each risk factor has a value range of [0, 1] and a weight sum of 1, therefore The result also falls within the [0, 1] interval. This calculation is performed continuously at a high frequency (e.g., per second), and updated in real time. value.

[0055] Through the above process, the integrated scoring module 104 utilizes environmental risk levels. By dynamically modulating the contribution ratios of multiple physiological risk factors, the final risk score is improved. It can simultaneously reflect the combined effects of physiological abnormalities and environmental background. The calculated real-time comprehensive risk score The alarm is output to the alarm control module 105 for further decision-making.

[0056] The alarm control module 105 is connected to the fusion scoring module 104. The alarm control module 105 has a built-in finite state machine that controls the system alarm state transition based on the comprehensive risk score, its short-term predicted value, and dynamically adjusted decision thresholds.

[0057] Alarm control module 105 receives real-time comprehensive risk score from fusion scoring module 104. and environmental risk levels from environmental assessment module 103. This module is based on The current value, the predicted future trend, and the result of The dynamically determined decision parameters are used to control the alarm level, timing, and process through a finite state machine. The output of the alarm control module 105 is the control command sent to the alarm execution module 106.

[0058] Alarm control module 105 adjusts the alarm based on the current environmental risk level. Set dynamic decision-making parameters, including early warning thresholds. Alarm threshold and confirmation delay time These parameters vary. Value changes, usually The higher the value, the higher the threshold. and The lower the setting, the shorter the latency. The shorter the setting, the more sensitive and responsive the system becomes in high-risk environments. One specific implementation method is to obtain the value through a table lookup, such as the table below: Table 3 Alarm Parameter Configuration Table Based on Environmental Risk Level Alarm control module 105 according to the current The value is obtained from this table or its equivalent function, representing the currently effective threshold. , and delay time .

[0059] Alarm control module 105 performs comprehensive risk scoring A short-term forecast is performed. This forecast uses a first-order exponential smoothing method. Let the current time be... The current rating is The predicted value at the previous time step was Then for the future Predicted rating values ​​after time Calculated using the following formula: in, The smoothing factor is a constant between 0 and 1. A typical value is 0.3, which strikes a balance between smoothing historical trends and responding to recent changes. To predict the step size, it can be set to be the same as... Associated short timeframes, such as 3 seconds, to quickly anticipate risk trends during the delayed confirmation period.

[0060] A finite state machine includes the following states and transition logic: state 0: Monitoring status. This is the initial state.

[0061] Residency requirements: Real-time comprehensive risk assessment Below the warning threshold .

[0062] Action: No alarm command.

[0063] Migration conditions: when Rise and reach or exceed When, migrate to state 1.

[0064] state 1: Warning status.

[0065] Entry condition: From state 0, when ≥ .

[0066] In-state logic: Calculate the predicted value .

[0067] like < Send a command to the alarm execution module 106 to execute a Level 1 alert (e.g., display a warning icon on the dial and generate a short vibration).

[0068] like ≥ This means that the risk score is highly likely to exceed the alarm threshold in the near future. To avoid delays, the system will skip the warning state and immediately transition to the current state. 2. To address the impending high risks in a more urgent manner.

[0069] like ≥ : Do not execute Level 1 prompts, immediately migrate to the current state. 2.

[0070] Migration conditions: like Falling back to Next, migrate back to the previous state. 0.

[0071] like Rise to Or above, migrate to state 2.

[0072] If the predicted value ≥ , migrate to state 2.

[0073] state S 2: Local alarm status.

[0074] Entry action: Send a command to the alarm execution module 106 to execute a Level 2 local alarm (e.g., continuous strong vibration, screen flashing red and displaying a cancel button).

[0075] Start a countdown timer, with the set duration being the confirmation delay time under the current circumstances. .

[0076] In-state monitoring: Monitor whether the user clicks the cancel button.

[0077] Continuous monitoring .

[0078] Monitor whether the countdown timer has reached zero.

[0079] Migration conditions: User Cancellation: If a user clicks the cancel button during the countdown, the timer and local alarms are stopped, and the system transitions back to its previous state. 0.

[0080] Risk mitigation: During the countdown period, if For multiple consecutive cycles (e.g., 3 cycles), the value is below If the timer stops, the state transitions back to the previous state. 1. This reflects the "buffer" mechanism: if the risk resolves itself during the delay period, there is no need to escalate the alert.

[0081] Timeout: If the countdown ends and neither of the above two conditions has occurred, then transition to the current state. 3.

[0082] state 3: Remote alarm status.

[0083] Entry Action: Send a command to alarm execution module 106 to execute a Level 3 remote alarm. The command includes the data to be sent, such as timestamps and real-time data. Value, environmental level , Coordinates, etc.

[0084] Subsequent migration: After a remote alarm command is issued, the status can be migrated to. 4. Provide continuous prompts, or automatically return to the status after a preset time. 0.

[0085] state 4: Continuous prompt status (optional).

[0086] Action: Maintain the alarm information display on the screen and intermittently vibrate gently until the user actively confirms.

[0087] Migration conditions: Migration back to the previous state only after user confirmation. 0.

[0088] In addition, to ensure system stability and robustness, the alarm control module 105 also integrates an anti-oscillation mechanism. When the finite state machine is in monitoring state... 0 and warning status When switching frequently between 1 and 2 (e.g., more than 3 times within 1 minute), the system will temporarily raise the warning threshold. Adjust the risk score upward by a fixed offset (e.g., +0.05) to suppress frequent alarms caused by the score fluctuating around the threshold. This offset will automatically recover after the risk score has stabilized or after a period of time (e.g., 10 minutes).

[0089] The alarm control module 105 performs the above-mentioned logical judgment at a fixed frequency (e.g., every second) based on the latest input and internal state, drives state transition, and outputs corresponding control commands, thereby realizing hierarchical, buffered alarm control with trend prediction.

[0090] The alarm execution module 106 is connected to the alarm control module 105. The alarm execution module 106 executes corresponding local prompts, local alarms, or remote alarm actions according to control commands.

[0091] The alarm execution module 106 receives control commands from the alarm control module 105. This module parses and executes these commands, specifically driving the smartwatch's hardware output devices or communication modules to complete a series of actions, from local notifications and local alarms to remote network alarms. The alarm execution module 106 is the final execution unit for system decisions.

[0092] The alarm execution module 106 is connected to multiple hardware driver interfaces of the smartwatch. When a command is received, the alarm execution module 106 calls the corresponding underlying driver based on the opcode and parameters carried in the command. The command typically contains an operation type field and an optional parameter data field.

[0093] For commands with an operation type of "Level 1 Prompt," the alarm execution module 106 performs a local prompt action. This typically involves driving the linear motor and display screen of the device. The linear motor is controlled to generate a brief, moderate-intensity vibration, which can be a single pulse. Simultaneously, the display screen's control layer is invoked to change the content displayed in a specific area of ​​the screen. For example, a yellow warning icon may be displayed in a corner of a watch face, or the bezel of the watch face may flash a yellow backlight for several seconds before automatically resuming. During this process, the device's speaker remains silent to avoid unnecessary interference.

[0094] For commands with an operation type of "Level 2 Local Alarm", the alarm execution module 106 performs a stronger local alarm action. This includes driving the linear motor to vibrate continuously in a specific high-intensity, high-frequency mode until a stop command or user interaction is received. The display control layer is invoked, switching the screen background to red and displaying a clear alarm message in the center of the screen, such as "An anomaly detected, do you need help?", while providing two virtual buttons, "Confirm Safety" and "Request Help," for user touch interaction. The screen may flash at a certain frequency to enhance visual alertness. At this time, the alarm execution module 106 simultaneously begins listening for touch screen events, waiting for user interaction with the virtual buttons, and feeding back the operation result (such as clicking "Confirm Safety") to the alarm control module 105.

[0095] For commands with an operation type of "Level 3 Remote Alarm", the alarm execution module 106 performs remote alarm data transmission. The parameter data segment of the command contains the alarm information packet to be sent. The alarm execution module 106 first encapsulates this information packet according to a predefined data structure. An exemplary alarm information packet data structure is shown in the table below: Table 4. Data Structure Definition Table for Alarm Information Packets The alarm execution module 106 calls the smartwatch's wireless communication management interface (such as a cellular network module or a Bluetooth channel connected to a mobile phone) to establish a network connection. Then, the encapsulated alarm information packet is transmitted using a secure transmission protocol (such as...). The data is sent to a preset cloud server address. The cloud server address and the necessary authentication key are pre-stored in the device's secure storage area. During data transmission, the alarm execution module 106 can report the transmission status (such as "sending", "sending successful", "sending failed") to the alarm control module 105. If transmission fails, the alarm execution module 106 can attempt to retransmit according to a predetermined retry strategy (e.g., retrying every 30 seconds, up to 3 times).

[0096] The alarm execution module 106 also needs to process "stop alarm" commands from the alarm control module 105. Upon receiving such a command, the alarm execution module 106 immediately stops any ongoing local alarm actions, such as stopping motor vibration, restoring normal screen display, and terminating any ongoing network transmission tasks.

[0097] Through the above design, the alarm execution module 106 transforms abstract alarm control commands into specific, user-perceptible hardware actions and reliable data communication processes, thereby completing the final stage of security monitoring.

[0098] To more clearly illustrate the complete workflow of this invention, a typical timing example is provided below: Initial state: User is at home (location security level) =1, Environmental Risk Level =1). The system is under monitoring. 0, =0.2.

[0099] Risk evolution: The user experiences sudden cardiac discomfort, and heart rate variability decreases sharply. The value increased from 0.1 to 0.8; this was due to user inactivity. Keep it at 0.1. The value rose to 0.35, triggering the warning threshold. (1) = 0.4, not reached, actual =0.35, still in The state is at 0. However, the risk continues to rise. It rises to 0.45 within 5 seconds, reaching =0.4, the state machine transitions to the warning state. 1.

[0100] Early warning and forecasting: In State 1, the system calculates the predicted value. Assuming =0.3, the predicted value quickly rose to 0.65, approaching the alarm threshold. (1) = 0.7. Meanwhile... It continued to climb to 0.68. The system, based on "if..." ≥ The rule of "" will be migrated to the local alarm status. 2.

[0101] Local Alarms and Buffers: Enter After 2 seconds, the watch starts vibrating strongly and the screen turns red, displaying a cancel button, while simultaneously starting for 20 seconds. The countdown started. The user attempted to cancel but was unsuccessful due to weakness. During the countdown, It remains above 0.7.

[0102] Alarm escalation: If the countdown ends, the user has not canceled the alarm, and the risk has not been mitigated, the state machine will transition to remote alarm status. 3. The alarm execution module 106 sends an alarm packet containing information such as location, heart rate, and risk score to the preset emergency contact or service center to complete the final alarm.

[0103] This example provides an intuitive understanding of how the modules work together to achieve a complete closed loop from risk perception, trend prediction, tiered buffering to final alerts.

[0104] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0105] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0106] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0107] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A smartwatch for security monitoring and alarm, characterized in that, include: The data acquisition and preprocessing module is used to simultaneously acquire and process data from multiple sensors; The risk feature extraction module is used to calculate risk factors of multiple dimensions based on the data processed by the data acquisition and preprocessing module. The environmental assessment module is used to assess the current environmental risk level based on the data processed by the data acquisition and preprocessing module. The dynamic risk fusion module connects the risk feature extraction module and the environmental assessment module. It is used to dynamically adjust the fusion weight of each risk factor according to the current environmental risk level, and fuse the weighted risk factors to generate a real-time comprehensive risk score. The predictive buffer alarm control module, connected to the dynamic risk fusion module and the environmental assessment module, is used to control the alarm state to migrate between multiple levels based on the real-time comprehensive risk score, its predicted value, and decision parameters dynamically determined by the current environmental risk level. The alarm state migration logic includes triggering a low-level warning based on the predicted value before reaching the high-level alarm threshold, and initiating a delayed confirmation process that can be stopped due to user operation or risk mitigation after reaching the high-level alarm threshold. The alarm execution module is connected to the predictive buffer alarm control module and is used to execute alarm actions corresponding to the current alarm state.

2. A smartwatch for security monitoring and alarm according to claim 1, characterized in that, The specific steps for the risk feature extraction module to calculate the heart rate variability factor include: Extracting continuous inter-heartbeat sequences from clean pulse wave signals; The standard deviation of the heart rate interval sequence is calculated within a sliding time window to obtain a time-domain index of heart rate variability. The heart rate variability time-domain index calculated in the current window is compared with the user's personal resting baseline value learned through long-term monitoring. The heart rate variability factor is obtained by normalization calculation, and its value increases as the current heart rate variability time-domain index decreases relative to the personal resting baseline value.

3. A smartwatch for security monitoring and alarm according to claim 1, characterized in that, The specific steps for the risk feature extraction module to calculate the motion contradiction factor include: Based on the preprocessed acceleration signal, the standard deviation of the acceleration vector amplitude is calculated within a short time window and used as the quantization value of the actual motion intensity. At the same time, the difference between the current instantaneous heart rate and the user's personal resting heart rate is obtained as the relative heart rate load value; The relative heart rate load value and the quantified actual exercise intensity value are input into a preset fusion function. The fusion function is configured to output a high value when the relative heart rate load value is high and the quantified actual exercise intensity value is low, thereby obtaining the exercise contradiction factor.

4. A smartwatch for security monitoring and alarm according to claim 1, characterized in that, The specific method for the comprehensive judgment of the execution rules in the environmental assessment module is as follows: A lookup table is predefined to map location security level, time anomaly flag, meteorological change flag, and acoustic anomaly flag to discrete environmental risk level; The location security level is determined by comparing the current geographic location with a pre-stored geofence database; The time anomaly flag is determined by comparing the current time, current location, and the user's historical behavior habit model. The meteorological change indicator is determined by judging whether the absolute value of the recent air pressure change rate exceeds a first threshold. The acoustic anomaly indicator is determined by judging whether the current ambient sound pressure level continuously deviates from its short-term dynamic baseline by a certain amount.

5. A smartwatch for security monitoring and alarm according to claim 1, characterized in that, The preset weight configuration table in the dynamic risk fusion module is designed as follows: As the environmental risk level increases, the weights assigned to the skin conductance response factor and the abnormal rest factor increase accordingly, while the weights assigned to the heart rate variability factor and the motion paradox factor decrease accordingly, making the system more sensitive to changes in physiological indicators characterizing user stress and mobility in high-risk environments.

6. A smartwatch for security monitoring and alarm according to claim 1, characterized in that, The specific formula for calculating the real-time comprehensive risk score RS by the dynamic risk fusion module is as follows: In the formula, , , , These respectively represent the heart rate variability factor, the motion paradox factor, the skin conductance response factor, and the abnormal rest factor; , , , These represent the dynamic fusion weights obtained from the weight configuration table, corresponding to the current environmental risk level, and satisfying the following conditions: + + + =1.

7. A smartwatch for security monitoring and alarm according to claim 1, characterized in that, The predictive buffer alarm control module uses a first-order exponential smoothing algorithm to predict the short-term trend of the real-time comprehensive risk score. The calculation formula is: In the formula, This provides a real-time comprehensive risk score for the current moment. The predicted value calculated at the previous time step. A smoothing factor between 0 and 1. This is the preset prediction step time.

8. A smartwatch for security monitoring and alarm according to claim 1, characterized in that, The finite state machine in the predictive buffer alarm control module includes at least a monitoring state, an early warning state, a local alarm state, and a remote alarm state. The monitoring status is maintained when the real-time comprehensive risk score is lower than the warning threshold. When the real-time comprehensive risk score reaches or exceeds the warning threshold, the warning state is entered. In the warning state, if the predicted score reaches or exceeds the alarm threshold, it will immediately transition to the local alarm state. Upon entering the local alarm state, a high-level local alarm control command is immediately triggered and the delayed acknowledgment timer is started; In the local alarm state, if a user cancellation signal is received before the timer expires or the real-time comprehensive risk score remains below the alarm threshold, the system will exit to the monitoring state or the early warning state. If the timer times out and the real-time comprehensive risk score is still higher than the alarm threshold, then the system will transition to the remote alarm state.

9. A smartwatch for security monitoring and alarm according to claim 1, characterized in that, The predictive buffer alarm control module dynamically sets the warning threshold and alarm threshold in the following way: A threshold lookup table is pre-stored, mapping different environmental risk levels to different warning and alarm thresholds; The duration of the delayed confirmation timer is also dynamically determined based on the current environmental risk level through a duration lookup table in a similar manner.

10. A smartwatch for security monitoring and alarm according to claim 1, characterized in that, The remote alarm data packet encapsulated by the alarm execution module also includes at least one of the following fields: The alarm is triggered by the heart rate value, the identification of the main triggering risk factors, the device battery level information, and a brief historical trend of the real-time comprehensive risk score over a recent period.

Images