User abnormal event recognition method and system based on wearable device and radar

Abstract

The application relates to the technical field of information processing, and particularly discloses a user abnormal event identification method and system based on a wearable device and a radar, which comprises a wearable device, a radar and a server, the radar is arranged in a target space, the wearable device is worn on a monitored user, and physiological data collected by the wearable device and space behavior data collected by the radar are received; a behavior abnormality score, a physiological abnormality score and a scene risk score are determined based on the physiological data and the space behavior data; a historical risk score is acquired; an abnormal event score is determined based on the behavior abnormality score, the physiological abnormality score, the scene risk score and the historical risk; an abnormal event level of the monitored user is determined according to the abnormal event score, and service response information is generated; through the fusion identification of the wearable device and the radar, multi-dimensional risk scoring based on physiological data and space behavior data is realized, the accuracy of user abnormal event identification is effectively improved, and the false positive rate is reduced.

Description

Technical Field

[0001] This invention relates to the field of information processing technology, and in particular to a method and system for identifying abnormal user events based on wearable devices and radar. Background Technology

[0002] With the accelerating aging of the population and the increasing demand for health management, health monitoring solutions based on wearable devices have received widespread attention. Currently, wearable devices such as smart bracelets, smartwatches, and smart jewelry can collect physiological data such as heart rate, blood oxygen, and HRV (heart rate variability) through photoplethysmography (PPG) sensors.

[0003] Existing wearable devices only collect physiological data and compare the collected user's physiological data with normal values ​​to determine whether the user has any abnormal risks. This general threshold method for identifying abnormal risks cannot distinguish whether the user is resting normally or experiencing pathological silence. For example, it cannot distinguish whether the user is resting quietly or experiencing sudden fainting silence. When the heart rate rises, it cannot distinguish whether the rise is pathological or caused by a sudden fall. This results in low accuracy in identifying abnormal events and a high false alarm rate. Summary of the Invention

[0004] The technical problem solved by this invention is to provide a method and system for identifying abnormal user events based on wearable devices and radar, so as to solve the problem that the accuracy of identifying abnormal user events is low and the false alarm rate is high in the prior art due to the use of only physiological data.

[0005] In a first aspect, the present invention provides a user anomaly event identification method based on wearable devices and radar, applied to an anomaly event identification system including wearable devices, radar, and a server, wherein the radar is disposed within a target space, and the wearable device is worn by a monitored user within the target space, and the user anomaly event identification method includes: Receive physiological data of the monitored user collected by the wearable device, and receive spatial behavior data of the monitored user within the target space collected by the radar; Based on the physiological data and spatial behavior data of the monitored user, determine the monitored user's behavioral abnormality score, physiological abnormality score, and scenario risk score; Obtain the historical risk score of the monitored user; Based on the behavioral anomaly score, the physiological anomaly score, the scenario risk score, and the historical risk score, an anomaly event score of the monitored user in the target space is determined. The level of abnormal events for the monitored user is determined based on the abnormal event score. Service response information is generated based on the level of the abnormal event.

[0006] Secondly, the present invention provides a user abnormal event identification system based on wearable devices and radar, including wearable devices, radar and server, wherein both the wearable devices and the radar communicate with the server, the radar is set in a target space, and the wearable devices are worn on the monitored user in the target space; The wearable device is configured to collect physiological data of the monitored user and send the physiological data to the server; The radar is configured to collect spatial behavior data of the monitored user within the target space and send the spatial behavior data to the server; The server is configured to receive the physiological data and the spatial behavior data, and execute the user abnormal event identification method based on wearable devices and radar as described in this invention.

[0007] This invention collects physiological data of monitored users through wearable devices and simultaneously collects spatial behavior data of monitored users through radar. Based on the physiological and spatial behavior data, it determines the monitored user's behavioral anomaly score, physiological anomaly score, and scene risk score. Furthermore, based on the behavioral anomaly score, physiological anomaly score, scene risk score, and historical risk score, it determines the abnormal event score of the monitored user in the target space, thereby determining the anomaly event level and generating corresponding service response information. On the one hand, this invention determines the anomaly event score and matches graded responses from four dimensions: behavioral anomaly, physiological anomaly, scene risk, and historical risk. This upgrades anomaly event identification from a single threshold comparison of physiological data to a multi-dimensional joint judgment based on user behavior, scene risk, and long-term historical data. This makes user anomaly event identification more scientific and comprehensive, able to distinguish whether the user's physiological state is abnormal during various spatial behaviors, significantly improving the accuracy and reliability of user anomaly event identification and reducing the false alarm rate. On the other hand, the radar is placed in the space where the monitored user is located. Compared with wearing motion sensors on the user's body, it can achieve unobtrusive real-time collection of user behavior. Moreover, compared with analyzing user behavior by collecting video streams through cameras, it offers better privacy and reduces the monitored user's resistance to being monitored.

[0008] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a flowchart of a user anomaly event identification method based on wearable devices and radar provided by an embodiment of the present invention; Figure 2 This is a structural diagram of a user abnormal event recognition system based on wearable devices and radar provided in an embodiment of the present invention. Detailed Implementation

[0011] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0012] Figure 1 A flowchart of a user anomaly event identification method based on wearable devices and radar provided in an embodiment of the present invention is shown below. Figure 1 As shown, this user anomaly event identification method based on wearable devices and radar includes: S10, Receive physiological data of the monitored user collected by the wearable device, and receive spatial behavior data of the monitored user in the target space collected by the radar.

[0013] The spatial user abnormal event recognition system of this invention is used to identify abnormal events of users within a target space. Abnormal events can include abnormal physiological data and abnormal behavior. Abnormal behavior can include events such as falling, sudden fainting, and prolonged sitting. The target space can be a hospital ward, a private residence, various areas of a nursing home, or the living room, bedroom, and bathroom of a family home, etc. Figure 2 As shown, the space user abnormal event identification system includes wearable devices, radar, and a server. The radar is set up in the target space, and the wearable devices are worn by the monitored users in the target space.

[0014] In one alternative embodiment, the wearable device can be a smart ring, smart bracelet, smartwatch, smart band, or similar device. The wearable device can integrate a PPG optical sensing module, a three-axis accelerometer, a body temperature sensor, and a low-power Bluetooth communication module. The PPG optical sensing module is based on the principle of photoplethysmography (PPG). It emits light waves of a specific wavelength (e.g., green light wavelength 500-600nm or infrared light wavelength 800-1000nm) and receives the light signals reflected or transmitted through human tissue. It then uses the periodic changes in the light signals caused by the heartbeat to extract physiological parameters such as heart rate, blood oxygen saturation, and HRV (heart rate variability).

[0015] Specifically, the PPG optical sensing module includes at least one LED light source and one photodetector. The LED light source emits light signals towards the finger tissue. After absorption and scattering by the skin, tissue, and blood vessels, a portion of the light signal reaches the photodetector. Since blood flow pulsates with the periodic beating of the heart, the degree of light signal absorption also changes periodically. By measuring the intensity changes of the light signal received by the photodetector, a pulse waveform synchronized with the heartbeat can be extracted. Peak detection of the pulse waveform and calculation of the time interval between adjacent peaks yield heart rate data. Further time-domain and frequency-domain analysis of the pulse waveform, calculating the standard deviation of the successive heartbeat interval (RR interval), yields HRV data. Furthermore, by emitting light signals of at least two different wavelengths (e.g., red light 660nm and infrared light 940nm), and utilizing the absorption differences of oxyhemoglobin and deoxyhemoglobin at different wavelengths, the absorption ratio of the light signal at the two wavelengths can be calculated, thus obtaining blood oxygen saturation data.

[0016] A triaxial accelerometer is used to collect motion acceleration signals of the finger parts in three-dimensional space (X-axis, Y-axis, Z-axis) to help identify the user's hand activity status and body activity intensity. When the user is stationary, the accelerometer outputs a static component close to gravitational acceleration (approximately 9.8 m / s²). When the user is active, the accelerometer outputs a dynamic acceleration change signal. By calculating the amplitude and analyzing the frequency of the acceleration signal, it is possible to determine whether the user is in motion, the level of motion intensity, and whether there has been a sudden impact (such as a fall).

[0017] Low-power Bluetooth communication modules (such as BLE5.0 or BLE5.3) are used to transmit collected physiological data to the server in real time. For example, the low-power Bluetooth communication module can upload the collected physiological data to the local gateway, and then upload it to the server via the Internet through the gateway. BLE5.3 technology supports long-distance transmission (line-of-sight distance of more than 100 meters) and low-power operation mode, making it suitable for all-weather continuous monitoring scenarios.

[0018] Preferably, the wearable device can be a smart ring. The smart ring can adopt the appearance design of smart jewelry, with the sensing module embedded inside the jewelry ring setting. The ring setting material can be ceramic, stainless steel or titanium alloy, etc. The appearance is no different from that of a regular decorative ring, realizing the aesthetic integration of medical sensing and daily wear, reducing users' resistance to medical devices, and improving users' compliance with long-term wear.

[0019] The smart ring supports IP68 waterproof and dustproof rating, allowing normal use in everyday water-related scenarios such as showering and handwashing. A single charge provides at least 7 days of battery life, making it suitable for continuous monitoring around the clock and minimizing the disruption to daily life caused by frequent charging. The smart ring also features a miniature positioning module (such as a UWB or Bluetooth AOA positioning tag) that supports indoor electronic fence functionality, triggering an alert when the monitored user leaves the designated area. Additionally, the smart ring has a one-button SOS physical button that triggers an emergency alarm when pressed by the user.

[0020] The radar can be millimeter-wave radar or lidar, preferably millimeter-wave radar, specifically 60GHz FMCW (Frequency Modulated Continuous Wave) radar. It includes an antenna array and a signal processing unit. FMCW radar transmits a continuous wave signal whose frequency changes linearly with time (e.g., a chirped signal linearly scanning from 57GHz to 64GHz) and receives the echo signal reflected from the human body. Using the frequency difference between the transmitted and received signals at the same moment (i.e., the difference frequency), the target distance is calculated using the formula d = c × f_diff / (2 × K), where c is the speed of light, f_diff is the difference frequency, and K is the frequency modulation slope. By performing Doppler processing on the received signal, the phase change of the echo signal is extracted, and the radial velocity of the target relative to the radar is calculated. Using the phase difference of the received signals from different antenna elements in the antenna array, an angle-of-arrival estimation algorithm (such as the MUSIC algorithm or the ESPRIT algorithm) is used to calculate the angle information of the target relative to the radar, ultimately generating point cloud data containing three-dimensional spatial coordinates (x, y, z).

[0021] Radar processes point cloud data using clustering algorithms (such as DBSCAN density clustering) and tracking algorithms (such as Kalman filtering or particle filtering). Clustering algorithms group spatially adjacent point cloud points as the same target, distinguishing human targets from environmental clutter. Tracking algorithms associate and track human targets in consecutive frames, extracting the centroid coordinates, motion trajectory, and posture features of the human target. Posture recognition is achieved by classifying the spatial distribution of the point cloud. For example, when the point cloud is distributed in a horizontal strip shape and close to the ground, it is determined to be a fallen posture; when the point cloud is distributed in a vertical elliptical shape and the height is between 1.2 and 1.8 meters, it is determined to be a standing posture; when the point cloud is compressed and the height is less than 0.5 meters, it is determined to be a sitting or lying posture.

[0022] Spatial behavior data includes at least one of the following: user posture (standing, sitting, lying down, falling), fall events (including the time and duration of the fall), activity level (quantified by the amplitude and frequency of movement per unit time, such as the cumulative distance of movement per hour or the integral of acceleration amplitude), bed-leaving events (the time and duration at which the user leaves the bed), prolonged static state (continuous sitting or lying down for more than a preset threshold, such as more than 2 hours), duration of stay in risk areas (such as balconies, bathrooms, kitchens, etc.), number of times the user gets up at night (the number of times the user gets up from the bed during the night), and indoor activity radius (the maximum distance the user moves within the target space).

[0023] Wearable devices and radar upload the collected data to the server in real time via Bluetooth Low Energy and Wi-Fi, respectively. The local gateway acts as a data aggregation node, connecting the wearable devices and radar to achieve millisecond-level timestamp synchronization of data from both ends. The gateway is equipped with a high-precision real-time clock (RTC) and keeps synchronized with the server time via the NTP protocol. Each piece of data from the wearable devices and radar is accompanied by a timestamp. The gateway aligns the timestamps of the received data and then packages and uploads it.

[0024] After receiving physiological and spatial behavior data, the server associates the received data with the user ID of the monitored user and stores it in the database.

[0025] This embodiment takes a nursing home scenario as an example. Each resident wears a smart ring, and a 60GHz FMCW millimeter-wave radar is installed on the ceiling of each individual's room. All devices are connected to the nursing home's local gateway and then to a cloud server. The smart ring collects physiological data such as the resident's heart rate, HRV, blood oxygen, and body surface temperature around the clock. The millimeter-wave radar collects spatial behavior data such as the resident's posture, fall events, bed-leaning events, nighttime toilet visits, and prolonged static states 24 hours a day. The data from the smart ring and the millimeter-wave radar are synchronized with the gateway using millisecond-level timestamps and then uploaded to the server.

[0026] S20. Based on the physiological data and spatial behavior data of the monitored users, determine the abnormal behavior score, abnormal physiological score, and scenario risk score of the monitored users.

[0027] In an optional embodiment, before step S20, a baseline establishment process is further included: acquiring historical physiological data and historical spatial behavior data of the monitored user within a preset time period in the target space; determining the physiological baseline of the monitored user based on the historical physiological data; the physiological baseline includes at least one of the following: mean resting heart rate, sleep heart rate range, normal HRV range, blood oxygen baseline range, and average daily sleep duration baseline; and determining the behavioral baseline of the monitored user based on the historical spatial behavior data; the behavioral baseline includes at least one of the following: sleep onset time, wake-up time, average number of nighttime awakenings, duration of stay in a single risk area, daytime quiet threshold duration, and daily activity range baseline.

[0028] A baseline refers to a personalized reference benchmark generated by statistical analysis of historical data collected from users in their normal daily lives. Unlike existing technologies that use fixed universal thresholds, this invention establishes a user-specific physiological and behavioral baseline through historical data learning. It uses the user's own historical data as a reference standard for anomaly detection, thereby eliminating the influence of individual differences on the detection results and significantly reducing the false alarm rate.

[0029] For example, the heart rate values ​​at the same time each day (such as daytime waking hours) over 14 days are statistically analyzed, and the average value is taken after removing abnormally high values ​​caused by exercise, emotions, etc., as the baseline for the average resting heart rate; the heart rate values ​​during sleep over 14 days are statistically analyzed, and the 5th percentile and 95th percentile are used as the lower and upper limits of the sleep heart rate range; the HRV values ​​during waking hours over 14 days are statistically analyzed, and the 5th percentile and 95th percentile are used as the normal HRV range; the blood oxygen values ​​over 14 days are statistically analyzed, and the 5th percentile and 95th percentile are used as the baseline range for blood oxygen; the daily sleep duration over 14 days is statistically analyzed, and the average value is used as the baseline for the average daily sleep duration.

[0030] Specifically, the sleep and wake-up times over 14 days were statistically analyzed to identify stable sleep and wake-up times using clustering. The number of times people woke up at night over 14 days was statistically analyzed, and the average was used as the baseline for the average number of times people woke up at night. The duration of a single stay in a risk area (e.g., time spent in the toilet or smoking area) over 14 days was statistically analyzed, and the average or median was used as the baseline for the duration of a single stay in a risk area. The continuous quiet time during the day (i.e., the length of a continuous period of time during which users do not move or have any activity) over 14 days was statistically analyzed, and the 90th percentile was used as the critical quiet time during the day. The amount of activity during the day over 14 days was statistically analyzed, and the 5th and 95th percentiles were used as the baselines for the daily activity range.

[0031] Taking the monitoring of elderly people in nursing homes as an example: For 14 consecutive days, the server automatically learns each elderly person's daily behavior patterns and physiological rhythms, generating personalized behavioral and physiological baselines. For example, Elderly Person A usually falls asleep at 22:00, gets up at 07:00, gets up 1-2 times per night, and spends 3-8 minutes in the toilet each time; resting heart rate is 58-68 beats / min, and blood oxygen is 97%-100%.

[0032] In an optional embodiment, step S20 includes the following sub-steps: Step S201: Compare each behavioral indicator in the spatial behavioral data with the behavioral baseline, calculate the deviation of each behavioral indicator, and perform a weighted summation of the deviations of each behavioral indicator to obtain the behavioral anomaly score.

[0033] Specifically, spatial behavior data includes at least one of the following: posture data, activity level data, number of nighttime urinations, and duration of stay in risk areas. The comparison process is as follows: The posture data from the real-time acquired spatial behavior data is compared with the posture reference in the behavior baseline to obtain the posture deviation value. Posture recognition is achieved through the morphological feature classification of point cloud data generated by millimeter-wave radar. Specifically, the server performs principal component analysis on the point cloud data to extract the distribution characteristics of the point cloud in three-dimensional space: if the dispersion of the point cloud in the horizontal direction (XZ plane) is greater than the dispersion in the vertical direction (Y axis), and the centroid height is less than 0.3 meters, it is classified as a falling posture; if the dispersion of the point cloud in the vertical direction is greater than that in the horizontal direction, and the centroid height is between 1.2 and 2.0 meters, it is classified as a standing posture; if the point cloud is compressed and the centroid height is between 0.3 and 0.8 meters, it is classified as a sitting posture. If the real-time posture is a falling posture while the normal posture in the reference is a standing, sitting, or lying posture, the posture deviation value is assigned a high value (e.g., 90-100 points); if the real-time posture is consistent with the reference posture, the posture deviation value is 0 points.

[0034] The activity level data from real-time collected spatial behavior data is compared with the activity level baseline in the behavior baseline to obtain the activity level deviation value. For example, if a user typically has an activity level of more than 200 steps per hour during the day (corresponding to an acceleration amplitude integral greater than a certain threshold), and real-time detection shows that the activity level is 0 for 2 consecutive hours (i.e., the acceleration amplitude is continuously lower than the resting threshold), then the activity level deviation value is high (e.g., 80-100 points). The deviation value is calculated based on the ratio of real-time activity level to the baseline value. Deviation range = |Real-time activity - Benchmark value| ÷ Benchmark value × 100%.

[0035] The number of times a user gets up at night in the real-time collected spatial behavior data is compared with the number of times a user gets up at night in the behavior baseline to obtain the night-time deviation value. For example, if a user usually gets up 1-2 times a night, and the real-time detection shows 5 times a user gets up at night in a single night, then the night-time deviation value is high (e.g., 70-90 points).

[0036] The time spent in risk areas in the real-time collected spatial behavior data is compared with the baseline time spent in risk areas in the behavior data to obtain the risk area stay deviation value. For example, if a user usually spends 5-10 minutes in the toilet, and the real-time detection shows that the user stays in the toilet for more than 30 minutes, the risk area stay deviation value is high (e.g., 90-100 points).

[0037] The deviation values ​​of posture, activity level, nighttime wake-up, and risk area stay are multiplied by their respective preset weights and then summed to obtain the abnormal behavior score. Specifically, the deviation values ​​of each behavioral indicator are first normalized to the range of 0-100 points, and then multiplied by their weights to obtain the abnormal behavior score of the monitored user.

[0038] For example, the formula for calculating the Behavioral Abnormality Score (BAS) is as follows: BAS=W 姿态 ×Attitude deviation value + W 活动量 × Activity level deviation + W 夜起 ×Nighttime deviation value + W 风险区域停留 × Deviation value of stay in risk areas.

[0039] Among them, W 姿态 W 活动量 W 夜起 W 风险区域停留 Assuming preset weights, and the sum of the four weights equals 1, in one optional embodiment, W 姿态 =0.35, W 活动量 =0.25, W 夜起 =0.20, W 风险区域停留 =0.20.

[0040] Step S202: Compare each physiological indicator in the physiological data with the physiological baseline, calculate the deviation of each physiological indicator, and perform a weighted summation of the deviations of each physiological indicator to obtain the physiological abnormality score.

[0041] The heart rate data collected in real time is compared with the mean resting heart rate baseline to obtain the heart rate deviation value. For example, if the user's mean resting heart rate is 72 beats / min and the real-time heart rate is 105 beats / min, then the heart rate deviation value is high. The deviation value is calculated based on the difference between the real-time heart rate and the baseline heart rate. Deviation range = |Real-time heart rate - Baseline heart rate| ÷ Baseline heart rate × 100%; The deviation needs to be normalized to the range of 0-100 points.

[0042] The HRV data from real-time collected physiological data is compared with the normal HRV interval baseline in the physiological data set to obtain the HRV deviation value. For example, the normal HRV interval is defined by the 5th percentile and the 95th percentile. If the real-time HRV value is lower than the lower limit of the normal HRV interval, the deviation value is positive; if the real-time HRV value is within the normal HRV interval, the deviation value is 0. If the user's normal HRV interval is 30-50ms and the real-time HRV is 15ms, the HRV deviation value is approximately 90 points.

[0043] The real-time blood oxygen saturation data is compared with the baseline blood oxygen saturation range to obtain the blood oxygen deviation value. Specifically, the baseline blood oxygen saturation range is defined by the 5th percentile and the 95th percentile. If the real-time blood oxygen saturation value is lower than the lower limit of the baseline blood oxygen saturation range, the deviation value is positive. If the user's baseline blood oxygen saturation range is 96%-100% and the real-time blood oxygen saturation is 92%, the blood oxygen deviation value is approximately 90 points.

[0044] The physiological abnormality score (PAS) is obtained by multiplying the heart rate deviation, HRV deviation, and blood oxygen deviation by their respective preset weights and then summing the results. For example, the formula for calculating the PAS is as follows: PAS=W 心率 ×Heart rate deviation + W HRV ×HRV deviation + W 血氧 × Blood oxygen deviation value; Among them, W 心率 W HRV W 血氧 The weights are preset, and the sum of the three weights equals 1. In an optional embodiment, W 心率 =0.40, W HRV =0.35, W 血氧 =0.25.

[0045] It should be noted that in S201-S202, the deviation value is calculated as follows: the difference between the real-time data and the baseline value is calculated, and the ratio of the difference to the baseline value is used as the deviation magnitude to eliminate the dimensional differences between different indicators.

[0046] Step S203: Extract the real-time spatial location coordinates of the monitored user from the spatial behavior data, and obtain the current system time. Determine the scenario risk score based on the real-time spatial location coordinates and the current system time.

[0047] The scenario risk score is used to characterize the degree of impact of external environmental factors (the user's current location and time period) on health risks. In one embodiment, the process of determining the scenario risk score is as follows: First, the risk coefficient for a given period is determined based on the current system time. The server retrieves the current system time from the system clock and determines whether it falls within a preset nighttime or daytime period. The preset nighttime period is from 11:00 PM to 5:00 AM the next day (covering the period of deep sleep), while the preset daytime period is from 5:00 AM to 11:00 PM. If the current system time falls within the preset nighttime period, the risk coefficient is the first period coefficient (e.g., 1.5); if the current system time falls within the preset daytime period, the risk coefficient is the second period coefficient (e.g., 0.8). The first period coefficient is greater than the second period coefficient because the probability of sudden cardiovascular and cerebrovascular events is higher at night (related to nighttime blood pressure fluctuations and circadian rhythms), and family members or caregivers are less alert, making delayed detection more likely.

[0048] Then, based on the real-time spatial location coordinates and a pre-defined spatial region attribute mapping table, the risk coefficient of the region corresponding to the real-time spatial location coordinates is determined. Different spatial regions in the spatial region attribute mapping table correspond to different regional risk coefficients. The point cloud data output by the radar in real time contains the three-dimensional spatial coordinates (x, y, z) of the user's centroid. The server matches these coordinates with the electronic map of the target space (the electronic map pre-marks the boundary coordinates of each region during system deployment) to determine which spatial region the user is currently in. For example, the bathroom floor is slippery and easy to fall in, and the space is enclosed, making it difficult to be discovered after fainting, with a risk coefficient of 1.8; the kitchen has the risk of secondary injuries from knives, gas, etc., with a risk coefficient of 1.5; the living room or dining room is an open space, easy to observe and rescue, with a risk coefficient of 1.0; the risk premium in the bedroom is low, with a risk coefficient of 0.6. The above are just examples. In actual applications, different scenarios can pre-configure the attribute mapping tables of each region in each target space to obtain the corresponding regional risk coefficients.

[0049] Finally, the time-period risk coefficient and the regional risk coefficient are multiplied and normalized to a range of 0-100 points to obtain the scenario risk score.

[0050] For example, in user A's room at a nursing home, when the radar detects that user A is currently in the bathroom and the system time is 2:30 AM, the time-based risk coefficient is 1.5, the area risk coefficient is 1.8, the product is 2.7, and the normalized scenario risk score is approximately 95 points (high-risk scenario). As another example, when user A is in bed in the bedroom during a daytime nap (2:00 PM), the time-based risk coefficient is 0.8, the area risk coefficient is 0.6, the product is 0.48, and the normalized scenario risk score is approximately 25 points (low-risk scenario).

[0051] S30. Obtain the historical risk score of the monitored user.

[0052] Historical risk scores are used to characterize the inherent vulnerability of monitored users and the cumulative deterioration trend of their recent health status. The server obtains the basic disease information of the monitored users, such as past medical history entered by the users through the client: hypertension, diabetes, stroke history or coronary heart disease history, etc., and further matches the basic disease information with a preset disease risk level table to determine the risk coefficient of the basic disease. For example, the risk coefficients corresponding to various diseases can be pre-configured. For example, the risk coefficient is 20 for no underlying disease, 40 for grade 1 hypertension, 55 for grade 2-3 hypertension, 60 for diabetes with hypertension, 75 for a history of stroke, and 80 for a history of coronary heart disease / heart failure. More risk coefficients corresponding to other diseases can also be configured.

[0053] Furthermore, the historical abnormal event score sequence of the monitored user within the first preset time window (e.g., within the last 7 days) is obtained, and the changing trend of the historical abnormal event score sequence is calculated. If the changing trend is an upward trend (i.e., the risk score is on the rise), the trend decay risk coefficient is determined according to the slope of the changing trend. The larger the slope, the higher the trend decay risk coefficient. If the changing trend is a non-upward trend (i.e., the risk score is stable or declining), the trend decay risk coefficient is zero. The underlying disease risk coefficient and the trend decay risk coefficient are weighted and summed to obtain the historical risk score.

[0054] For example, the server obtains the daily abnormal event score sequence of the monitored user within the first preset time window (e.g., but not limited to the last 7 days) (i.e., the daily RiskScore calculated and stored when this method was last executed). The daily RiskScore is stored in the database after the last calculation of the day, associated with the user ID and date. The server reads the RiskScore sequence [R1, R2, ..., R7] of the last 7 days, and the RiskScore sequence [R8, R9, ..., R14] of the previous 7 days (i.e., the previous 14 to the previous 8 days), and calculates the mean of the two sequences respectively: the mean of the last 7 days μ1=(R1+...+R7) / 7, the mean of the previous 7 days μ2=(R8+...+R14) / 7, and the slope of the trend k=(μ1-μ2) / 7.

[0055] If the slope k is positive (i.e., μ1 > μ2, indicating an upward trend in the risk score), the trend decay risk coefficient is determined based on the slope magnitude. A larger slope indicates a faster deterioration in health, resulting in a higher trend decay risk coefficient. The trend decay risk coefficient = k × λ, where λ is a preset amplification factor (e.g., λ = 2.0) used to map the slope to a range of 0-100 points. If k ≤ 0 (i.e., the risk score is stable or declining), the trend decay risk coefficient is 0.

[0056] The historical risk score (HRS) is obtained by weighting and summing the risk coefficients of underlying diseases and the trend decay risk coefficients. HRS = Risk coefficient of underlying disease × W 基础 + Trend decay risk coefficient × W 趋势 ; Among them, W 基础 +W 趋势 =1, in an alternative embodiment, W 基础 =0.6, W 趋势 =0.4, with underlying diseases reflecting the user's long-term health status and given a higher weight, while trend decay reflects recent dynamic deterioration and is given a lower weight, in order to avoid excessive short-term fluctuations affecting the historical risk score.

[0057] Example calculation: The user has no underlying diseases, the risk coefficient of the underlying diseases is 20, the average risk score of the past 7 days is 58 points, the average score of the window of the previous 7 days is 42 points, the slope k=(58-42) / 7≈2.29, the trend decay risk coefficient=2.29×2.0≈4.58, then the historical risk score HRS=20×0.6+4.58×0.4=12+1.83≈14 points.

[0058] If a user has a history of stroke (underlying disease risk coefficient 75), the average risk score for the past 7 days is 58, and the average score for the previous 7 days is 42, then the historical risk score HRS = 75 × 0.6 + 4.58 × 0.4 = 45 + 1.83 ≈ 47.

[0059] Historical risk scores are used to characterize the inherent vulnerability of monitored users and the cumulative deterioration trend of their recent health status. The underlying disease risk coefficient reflects the impact of the user's underlying disease background (such as hypertension, diabetes, history of stroke, etc.) on health risk. The trend decay risk coefficient reflects the deterioration slope of the user's recent abnormal event scores, capturing the progressive health decline trend.

[0060] S40. Based on behavioral anomaly scores, physiological anomaly scores, scenario risk scores, and historical risk scores, determine the abnormal event scores of the monitored user within the target space.

[0061] The RiskScore is calculated using the following formula: RiskScore=BAS×α+PAS×β+SRS×γ+HRS×δ; Wherein, BAS is the behavioral abnormality score, PAS is the physiological abnormality score, SRS is the scenario risk score, HRS is the historical risk score, and α, β, γ, and δ are the weights of the behavioral abnormality score, physiological abnormality score, scenario risk score, and historical risk score, respectively, and the sum of α, β, γ, and δ is equal to 1. In an optional embodiment, α=0.35, β=0.35, γ=0.20, and δ=0.10.

[0062] The above weighting is based on the following: behavioral abnormalities and physiological abnormalities are direct indicators of the risk of abnormal events. Behavioral abnormalities directly reflect whether a user has experienced objective abnormal events such as falls or prolonged periods of stillness, while physiological abnormalities directly reflect whether a user has experienced physiological deterioration such as a sudden increase in heart rate or a sudden decrease in blood oxygen. Each constitutes an independent risk dimension, and therefore each is assigned a relatively high weight of 0.35. Scenario risk, as an amplifier of the external environment (the same abnormality occurring in a bathroom at night has a higher degree of urgency than occurring in a living room during the day), is assigned a weight of 0.20. Historical risk, as a long-term background correction factor (reflecting the user's inherent vulnerability and gradual decline trend), is assigned a weight of 0.10. The above weighting values ​​are merely illustrative; those skilled in the art can adjust these weighting parameters according to actual application scenarios and monitoring accuracy requirements.

[0063] In one example, a user has BAS=75, PAS=80, SRS=90, and HRS=45. Then: RiskScore=75×0.35+80×0.35+90×0.20+45×0.10=76.75.

[0064] S50. Determine the level of abnormal events for the monitored user based on the abnormal event score.

[0065] The server matches the calculated anomaly event score to the corresponding anomaly event level. The criteria for classifying anomaly event levels are as follows: When the abnormal event score is greater than or equal to 0 and less than 40, it is judged as normal level; when the abnormal event score is greater than or equal to 40 and less than 60, it is judged as information reminder level; when the abnormal event score is greater than or equal to 60 and less than 80, it is judged as attention warning level; when the abnormal event score is greater than or equal to 80 and less than or equal to 100, it is judged as emergency alarm level.

[0066] The above scoring intervals are based on the normalization of scores for each dimension, such as behavioral abnormality score and physiological abnormality score, to 0-100 points. This ensures that the weighted fusion result of the scores for each dimension has a clear physical meaning. In practical applications, the boundary values ​​of each level interval can be adjusted according to the monitoring accuracy requirements and the tolerance for false alarm rate.

[0067] S60. Generate service response information based on the level of the abnormal event.

[0068] Based on the determined level of the abnormal event, the server generates corresponding service response information and pushes it to the user client, family client, and / or doctor's client / institutional management backend. The service response information can be medical service response information.

[0069] When the abnormal event level is normal (L0), the server generates a daily health briefing, which is pushed to the user's client in the form of an in-app message or a WeChat mini-program message. The daily health briefing includes the abnormal event score for the day, an overview of the scores for each dimension, and a trend summary of key indicators such as heart rate / blood oxygen / activity level, providing users with daily health references without any medical intervention.

[0070] When the abnormal event level is the information alert level (L1), the server generates a weekly report summarizing the health change trend information and pushes it in the form of a weekly summary report. The weekly report includes the abnormal event score curve for the past 7 days, the weekly change trend of scores in each dimension, the comparative analysis with the previous week, and textual suggestions (such as the slight decline in sleep quality at night this week, suggesting maintaining a regular schedule). The weekly report is for informational reference only and does not force users to take action.

[0071] When an abnormal event reaches the Level 2 (Level 2) of Attention Warning, the server generates appointment booking information and chronic disease follow-up reminders. Specifically, the server pushes an attention reminder notification to the user's client and / or family member's client, stating: "Your abnormal event score has been detected as XX points, which falls under the Attention level. We recommend seeking medical attention and examination promptly." The notification includes a one-click link that leads to the online appointment booking page of a partner medical institution. Users can book online text or video consultations through this portal. Simultaneously, the server pushes a reminder to the doctor, informing them of the user's Attention-Level abnormal event risk and automatically retrieving the user's structured health trend data for the past 14 days (including heart rate curves, blood oxygen curves, activity level curves, and nighttime sleep analysis) for the doctor to review before consultation, assisting in accurate diagnosis and treatment decisions.

[0072] When the abnormal event level is Emergency Alarm Level (L3), the server triggers a multi-channel triple alarm command and simultaneously generates an emergency green channel linkage request and a home visit trigger command. Specifically: App push notifications: Emergency alarm messages are pushed to user clients, family member clients, caregiver clients / institutional management backends via APNs (Apple Push Notification Service) and FCM (Firebase Cloud Push Messaging). The alarm messages include the alarm type (such as fall alarm or abnormal vital signs), the time of occurrence, and the user's current location.

[0073] SMS notification: Send an SMS message containing the user's name, alarm type, and location information to preset emergency contacts (such as family members or caregiver team leaders) via the SMS gateway.

[0074] Telephone notification: Call emergency contacts via automated voice call system (TTS technology) to play alarm messages and ensure that alarm information is received in a timely manner (avoid missing push notifications or text messages when the phone is on silent mode).

[0075] Simultaneously, the server sends an emergency green channel linkage request to the information system of the cooperating medical institution, including the user's identity information, abnormal event score, real-time physiological data, and behavioral data summary, so that the medical institution can prepare for receiving the patient in advance (such as pre-registration, pre-arrangement of emergency beds, and notification of emergency doctors). The server also generates a home visit trigger instruction to the institution's management backend, notifying caregivers or visiting doctors to immediately go to the user's location for on-site examination and treatment.

[0076] Taking a user's RiskScore of 76.75 as an example, since the abnormal event level is a warning level, the server generates online consultation appointment information and chronic disease follow-up reminder information, and pushes them to the user's client. The user can make an appointment for online consultation or chronic disease follow-up services with one click through the client. The doctor's side simultaneously receives the user's health trend data to assist in accurate diagnosis and treatment.

[0077] This embodiment achieves a complete process from daily monitoring to emergency medical intervention by matching the level of abnormal events with the tiered medical service measures. Under the normal level, only daily briefings are pushed to avoid excessive disturbance. Under the information reminder level, health trends are summarized for users' reference. Under the attention warning level, access to consultation appointments and chronic disease follow-up is provided to facilitate users to seek medical treatment in a timely manner. Under the emergency alarm level, information is delivered in a timely manner through multi-channel triple alarms, and rapid medical response is achieved through linkage requests via the emergency green channel. This solves the industry pain point of existing technologies that only monitor but do not provide services.

[0078] In one optional embodiment, when the wearable device is offline or disconnected (e.g., battery depleted, outside Bluetooth range, device malfunction), the radar independently performs fall detection and prolonged stationary state detection using the edge computing capabilities of the local gateway, triggering an emergency alarm. Specifically, the radar continuously collects spatial behavior data, and the local gateway has pre-configured rule-based fall detection logic (e.g., a sudden drop in point cloud centroid height and a sudden increase in horizontal scattering are considered falls). When a fall or prolonged stationary state is detected exceeding a preset threshold (e.g., 10 minutes), the gateway directly sends an alarm to the server via WiFi or 4G / 5G communication modules, without relying on data from the wearable device.

[0079] When the radar loses power or goes offline (e.g., due to power outage or device malfunction), the wearable device triggers an emergency alarm as a backup measure by detecting abnormal physiological indicators (such as heart rate consistently exceeding a preset threshold or blood oxygen levels falling below a preset threshold) and / or by the user actively pressing the SOS physical button on the wearable device. The wearable device's Bluetooth communication module directly sends an alarm to the server via the gateway when it detects an anomaly or the SOS button being pressed.

[0080] Through the aforementioned dual-end backup mechanism, it is ensured that the system can still provide basic security alarm functions through the other terminal when any sensing terminal fails, thereby improving the robustness and reliability of the system in complex real-world environments. The following examples illustrate the user anomaly event identification method based on wearable devices and radar according to embodiments of the present invention: Example: Specific applications in elderly safety monitoring scenarios This embodiment uses a nursing home scenario as an example to illustrate the application of the present invention in the safety monitoring of the elderly. Each resident of the nursing home wears a smart ring, and a 60GHz FMCW millimeter-wave radar is installed on the ceiling of each individual's room. All devices are connected to the local gateway of the nursing home and then connected to the cloud server through the gateway.

[0081] The smart ring collects physiological data such as heart rate, HRV, blood oxygen, and body surface temperature of the elderly around the clock, while the millimeter-wave radar collects spatial behavior data such as the elderly's posture, fall events, bed-leaning events, number of nighttime toilet visits, and prolonged static states around the clock. The data from the smart ring and the millimeter-wave radar are synchronized with the gateway at the millisecond-level timestamp and then uploaded to the server.

[0082] Baseline establishment: For 14 consecutive days, the server automatically learns each elderly person's daily behavior patterns and circadian rhythms, generating personalized behavioral and physiological baselines. For example, Elderly Person A typically falls asleep at 10:00 PM, wakes up at 7:00 AM, gets up 1-2 times per night, and spends 3-8 minutes each time in the toilet; resting heart rate is 58-68 beats / min, and blood oxygen saturation is 97%-100%.

[0083] Scenario 1 (Emergency Event): Radar detected abnormal posture data of elderly person A in the bathroom area (consistent with fall posture characteristics). Simultaneously, the heart rate suddenly increased from 62 beats / min to 115 beats / min. Therefore, the behavioral abnormality score is determined to be 92 points, the physiological abnormality score to be 85 points, the nighttime bathroom scenario risk score to be 95 points, and the historical risk score to be 35 points. The abnormal event score = 92 × 0.35 + 85 × 0.35 + 95 × 0.20 + 35 × 0.10 = 87.95 points, which is classified as an emergency alarm level. The server immediately triggers a triple alarm through App push notifications, SMS notifications, and telephone contact, notifying the family and caregiver station respectively, and simultaneously coordinating with partner medical institutions' emergency green channels to arrange home visits.

[0084] Scenario 2 (Non-urgent concern): For 7 consecutive days, the radar detected that Elder A's nighttime awakening frequency increased from 1-2 times to 4-5 times, and the radar also detected that blood oxygen levels remained below the individual baseline (below 94%) during nighttime sleep. The behavioral abnormality score was determined to be 65 points, the physiological abnormality score to be 70 points, the scenario risk score to be 55 points, and the historical risk score to be 45 points. The abnormal event score = 65×0.35+70×0.35+55×0.20+45×0.10=63 points, which was determined to be a concern warning level. The server pushed a concern reminder to the family's client and generated a consultation appointment portal. The family could make an online consultation with one click. The doctor simultaneously obtained the elderly's 14-day health trend data and provided follow-up suggestions.

[0085] Figure 2 This is a schematic diagram of a user anomaly event recognition system based on wearable devices and radar according to an embodiment of the present invention, as shown below. Figure 2 As shown, the user anomaly event recognition system based on wearable devices and radar includes wearable devices, radar, and a server. Both the wearable devices and radar communicate with the server (via Bluetooth, WiFi, or a gateway device). The radar is set in the target space (such as the ceiling or wall of a bedroom or living room). The wearable devices are worn by the monitored users in the target space. One server can communicate with wearable devices and radars in multiple target spaces.

[0086] The wearable device is configured to collect physiological data of the monitored user (including heart rate, HRV, blood oxygen saturation, body surface temperature, and motion impact signals) and send the physiological data to the server. The radar is configured to collect spatial behavior data of the monitored user in the target space, including posture, fall events, activity level, bed-leaning events, prolonged static state, bathroom stay duration, number of nighttime urinations, and indoor activity radius, and send the spatial behavior data to the server. The server is configured to receive the physiological data and spatial behavior data and execute the user abnormal event identification method based on wearable device and radar according to the embodiments of the present invention.

[0087] In an optional embodiment, the system further includes a local gateway configured to connect the wearable device and the radar, enabling millisecond-level timestamp synchronization of data between the two ends, and caching data and executing local emergency rule triggering in the event of a network outage. The local gateway adopts an edge computing architecture, enabling it to independently trigger local alarms for high-risk events such as falls and extreme abnormalities in heart rate and blood oxygenation when the network is interrupted, without relying on cloud services, thereby improving the system's reliability in complex network environments.

[0088] It should be understood that the various forms of processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this invention does not impose any limitations on them.

[0089] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for identifying abnormal user events based on wearable devices and radar, characterized in that, An abnormal event identification system is applied to a system comprising wearable devices, radar, and servers, wherein the radar is positioned within a target space, the wearable devices are worn by monitored users within the target space, and the user abnormal event identification method includes: Receive physiological data of the monitored user collected by the wearable device, and receive spatial behavior data of the monitored user within the target space collected by the radar; Based on the physiological data and spatial behavior data of the monitored user, determine the monitored user's behavioral abnormality score, physiological abnormality score, and scenario risk score; Obtain the historical risk score of the monitored user; Based on the behavioral anomaly score, the physiological anomaly score, the scenario risk score, and the historical risk score, an anomaly event score of the monitored user in the target space is determined. The level of the abnormal event for the monitored user is determined based on the abnormal event score. Service response information is generated based on the level of the abnormal event.

2. The method according to claim 1, characterized in that, Before determining the behavioral anomaly score, physiological anomaly score, and scene risk score of the monitored user based on the monitored user's physiological data and spatial behavior data, the process also includes: Acquire historical physiological data and historical spatial behavior data of the monitored user within a preset time period in the target space; The physiological baseline of the monitored user is determined based on the historical physiological data. The physiological baseline includes at least one of the monitored user's mean resting heart rate, sleep heart rate range, normal HRV range, blood oxygen baseline range, and average daily sleep duration baseline. The behavioral baseline of the monitored user is determined based on the historical spatial behavior data. The behavioral baseline includes at least one of the following: sleep time, wake-up time, average number of nighttime awakenings per day, duration of stay in a single risk area, critical duration of quiet during the day, and daily activity interval benchmark.

3. The method according to claim 2, characterized in that, Based on the physiological data and spatial behavior data of the monitored user, determine the monitored user's behavioral anomaly score, physiological anomaly score, and scene risk score, including: Each behavioral indicator in the spatial behavioral data is compared with the behavioral baseline, the deviation of each behavioral indicator is calculated, and the deviation of each behavioral indicator is weighted and summed to obtain the behavioral anomaly score. Each physiological indicator in the physiological data is compared with the physiological baseline, the deviation of each physiological indicator is calculated, and the deviation of each physiological indicator is weighted and summed to obtain the physiological abnormality score. The real-time spatial location coordinates of the monitored user are extracted from the spatial behavior data, and the current system time is obtained. The scenario risk score is determined based on the real-time spatial location coordinates and the current system time.

4. The method according to claim 3, characterized in that, Extracting the real-time spatial coordinates of the monitored user from the spatial behavior data, and obtaining the current system time, and determining the scenario risk score based on the real-time spatial coordinates and the current system time, includes: The risk coefficient for a specific time period is determined based on the current system time. If the current system time is in a preset nighttime period, the risk coefficient for that time period is a first time period coefficient. If the current system time is in a preset daytime period, the risk coefficient for that time period is a second time period coefficient. The first time period coefficient is greater than the second time period coefficient. Based on the real-time spatial location coordinates, a preset spatial region attribute mapping table is matched to determine the regional risk coefficient corresponding to the real-time spatial location coordinates. Different spatial regions in the spatial region attribute mapping table correspond to different regional risk coefficients. The scenario risk score is obtained by multiplying the time period risk coefficient and the regional risk coefficient and then normalizing the results.

5. The method according to claim 3, characterized in that, Each behavioral indicator in the spatial behavior data is compared with the behavioral baseline, the deviation of each behavioral indicator is calculated, and the deviations of each behavioral indicator are weighted and summed to obtain the behavioral anomaly score, including: The attitude data in the real-time collected spatial behavior data is compared with the attitude reference in the behavior baseline to obtain the attitude deviation value; The activity data in the real-time collected spatial behavior data is compared with the activity benchmark in the behavior baseline to obtain the activity deviation value; The number of nighttime awakenings in the real-time collected spatial behavior data is compared with the nighttime awakenings baseline in the behavior data to obtain the nighttime awakening deviation value; The dwell time in the risk area in the real-time collected spatial behavior data is compared with the dwell time benchmark in the risk area in the behavior baseline to obtain the dwell time deviation value in the risk area; The abnormal behavior score is obtained by multiplying the posture deviation value, activity level deviation value, nighttime wake-up deviation value, and risk area stay deviation value by their respective preset weights and then summing them.

6. The method according to claim 3, characterized in that, Each physiological indicator in the physiological data is compared with the physiological baseline, the deviation of each physiological indicator is calculated, and the deviations of each physiological indicator are weighted and summed to obtain the physiological abnormality score, including: The heart rate data in the real-time collected physiological data is compared with the mean resting heart rate in the physiological baseline to obtain the heart rate deviation value. The HRV data in the real-time collected physiological data is compared with the normal HRV interval benchmark in the physiological baseline to obtain the HRV deviation value; The blood oxygen data in the real-time collected physiological data is compared with the blood oxygen reference interval in the physiological baseline to obtain the blood oxygen deviation value. The physiological abnormality score is obtained by multiplying the heart rate deviation, HRV deviation, and blood oxygen deviation by their respective preset weights and then summing them.

7. The method according to claim 1, characterized in that, Obtain the historical risk score of the monitored user, including: Obtain the basic disease information of the monitored user, match the basic disease information with a preset disease risk level table, and determine the basic disease risk coefficient; Obtain the historical abnormal event score sequence of the monitored user within the first preset time window, and calculate the changing trend of the historical abnormal event score sequence; If the trend is upward, the trend decay risk coefficient is determined based on the slope of the trend; if the trend is not upward, the trend decay risk coefficient is zero. The historical risk score is obtained by weighted summing of the underlying disease risk coefficient and the trend decay risk coefficient.

8. The method according to any one of claims 1-7, characterized in that, The abnormal behavior score, the abnormal physiological score, the scenario risk score, and the historical risk score are used to determine the abnormal event score of the monitored user within the target space, including: The RiskScore is calculated using the following formula: RiskScore=BAS×α+PAS×β+SRS×γ+HRS×δ; Wherein, BAS is the behavioral abnormality score, PAS is the physiological abnormality score, SRS is the scenario risk score, HRS is the historical risk score, and α, β, γ, and δ are the weights of the behavioral abnormality score, physiological abnormality score, scenario risk score, and historical risk score, respectively, and the sum of α, β, γ, and δ is equal to 1. The abnormal event level of the monitored user is determined based on the abnormal event score, including: When the abnormal event score is greater than or equal to 0 and less than 40, it is judged as normal. When the abnormal event score is greater than or equal to 40 and less than 60, it is determined to be an information alert level; When the score of the abnormal event is greater than or equal to 60 and less than 80, it is determined to be at the level of attention and warning. When the abnormal event score is greater than or equal to 80 and less than or equal to 100, it is determined to be an emergency alarm level.

9. The method according to claim 8, characterized in that, Service response information is generated based on the aforementioned anomaly event level, including: When the abnormal event level is normal, a daily health briefing is generated. When the abnormal event is at the information alert level, a weekly report summarizing health change trends is generated. When the abnormal event level is a warning level, information on appointment booking and chronic disease follow-up reminders are generated. When the abnormal event level is an emergency alarm level, a triple alarm command is generated, consisting of an App push notification, an SMS notification, and a telephone notification. Simultaneously, an emergency green channel linkage request and a home visit trigger command are generated.

10. A user anomaly event recognition system based on wearable devices and radar, characterized in that, It includes wearable devices, radar, and a server. Both the wearable devices and the radar communicate with the server. The radar is located within the target space, and the wearable devices are worn by the monitored users within the target space. The wearable device is configured to collect physiological data of the monitored user and send the physiological data to the server; The radar is configured to collect spatial behavior data of the monitored user within the target space and send the spatial behavior data to the server; The server is configured to receive the physiological data and the spatial behavior data, and to execute the user abnormal event identification method based on wearable devices and radar as described in any one of claims 1-9.

Images