A warning system and method of implementation for monitoring unexpected falls and wanderings

Abstract

This invention discloses an early warning system and implementation method for monitoring accidental falls and getting lost. The method includes: Step 1, system initialization and device binding; Step 2, data acquisition; Step 3, abnormal behavior analysis; Step 4, tiered early warning. This invention distinguishes postures through a basic posture recognition unit, establishes a normal gait baseline through a normal gait baseline modeling unit, and determines fall actions based on posture and the normal gait baseline through a fall determination unit, effectively reducing false alarms and missed alarms. A personalized travel modeling unit constructs a personal travel behavior baseline model, and a multi-dimensional missing person anomaly detection unit performs four-dimensional anomaly detection based on space, time, behavior, and device status. Combined with adaptive dynamic geofencing, this solves the problem of rigid missing person determination mechanisms in existing systems. The tiered early warning module adopts a three-level differentiated early warning system, matching corresponding response strategies according to the level of abnormal risk, rationally allocating monitoring resources, and improving the user experience.

Description

Technical Field

[0001] This invention relates to the field of smart wearable device technology, specifically to an early warning system and implementation method for monitoring accidental falls and getting lost. Background Technology

[0002] With the accelerating aging process, accidents such as falls and getting lost among the elderly have become a major challenge in the field of public health. Currently, the main technical solutions for safety monitoring of the elderly rely on wearable devices such as smart bracelets and smartwatches. However, smart bracelets require users to actively wear them, and due to the poor memory of the elderly, there is a problem of them forgetting to wear them, leading to monitoring failure. Some solutions use insole-type monitoring devices, which use built-in sensors to detect position, angle, and acceleration data to determine whether the elderly have fallen or gotten lost. However, existing insole-type monitoring devices generally have the following shortcomings and cannot meet actual monitoring needs: 1. Existing systems cannot distinguish between active, controllable movements and passive, uncontrolled falls: Existing system algorithms only rely on posture angles. 1. The two static conditions of exceeding the limit and acceleration limit are used to judge all actions that meet the threshold as falls, which cannot distinguish daily actions that the elderly actively perform, such as crossing their legs or kicking their legs for exercise; 2. The missing person detection mechanism is rigid and has extremely low accuracy: The existing system only judges missing persons by crossing the static geofence. When the static geofence is set too large, it is easy for people to get lost without crossing the boundary. When the static geofence is set too small, it is easy to generate frequent warnings, causing warning fatigue for family members; 3. The existing system's warning mechanism is crude and lacks hierarchical design, which cannot reasonably allocate monitoring resources and affects the user experience. Summary of the Invention

[0003] The purpose of this invention is to provide an early warning system and implementation method for monitoring accidental falls and getting lost, so as to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: an early warning system for monitoring accidental falls and getting lost, comprising a device terminal, a cloud terminal, and a mobile terminal, wherein the cloud terminal establishes data connections with the device terminal and the mobile terminal respectively, and the cloud terminal includes a device information input module, an abnormal behavior analysis module, a graded early warning module, and a data storage module, wherein the data storage module establishes data connections with the device information input module, the abnormal behavior analysis module, and the graded early warning module respectively.

[0005] The device includes a master shoe monitoring end and a slave shoe monitoring end. The master shoe monitoring end includes a first six-axis sensor module, a first Bluetooth module, a communication module, a GPS module, a first power module, and a first pressure sensor array module. The first six-axis sensor module is used to acquire quaternion, triaxial acceleration, and angular velocity data. The master shoe monitoring end establishes a data connection with the slave shoe monitoring end through the first Bluetooth module, establishes a data connection with the cloud through the communication module, and acquires GPS data through the GPS module. The first power module is used to supply power to the master shoe monitoring end, and the first pressure sensor array module is used to detect the plantar pressure distribution and pressure change curve.

[0006] The slave shoe monitoring terminal includes a second six-axis sensor module, a second Bluetooth module, a second power supply module, and a second pressure sensor array module. The second six-axis sensor module is used to acquire quaternion, triaxial acceleration, and angular velocity data. The slave shoe monitoring terminal establishes a data connection with the master shoe monitoring terminal using the second Bluetooth module. The second power supply module is used to supply power to the slave shoe monitoring terminal. The second pressure sensor array module is used to detect the plantar pressure distribution and pressure change curve.

[0007] The abnormal behavior analysis module includes a transmission disconnection early warning unit, a basic posture recognition unit, a normal gait baseline modeling unit, a fall determination unit, a personalized travel modeling unit, a multi-dimensional missing person anomaly detection unit, and a dynamic geofence management unit. The transmission disconnection early warning unit monitors the communication status between the device and the cloud. The basic posture recognition unit identifies basic human postures in real time based on plantar pressure data of both feet. Basic postures include standing, walking, and sitting. The normal gait baseline modeling unit establishes a normal gait baseline model based on bipedal data when standing, walking, and going up and down stairs. The fall determination unit determines whether a fall has occurred. The personalized travel modeling unit constructs a personalized multi-dimensional travel behavior baseline model, specifically including a temporal baseline and a spatial baseline. The temporal baseline includes daily travel time periods and single travel duration, while the spatial baseline includes a set of frequently visited locations, a set of travel routes, and a maximum activity radius. The multi-dimensional missing person anomaly detection unit detects travel anomalies. The dynamic geofence management unit automatically adjusts the geofence range according to time and weather, and also supports custom adjustments.

[0008] The judgment rule of the fall judgment unit is as follows: when the basic posture is either standing or walking, if a normal gait does not appear within a preset time after an abnormal gait, it is judged as a fall.

[0009] The multi-dimensional missing person anomaly detection unit comprehensively judges travel anomalies based on four dimensions: spatial features, temporal features, behavioral features, and device status. If any two of these four dimensions show anomalies and the duration exceeds a preset threshold, the status is judged as suspected missing. If any three of these four dimensions show anomalies and the duration exceeds a preset threshold, the status is judged as confirmed missing. The spatial feature anomaly judgment criteria include route anomalies and regional anomalies. Route anomalies are judged by route deviation, which is represented by the Frescher distance similarity between the current route and historical fixed routes. If the Frescher distance similarity exceeds a threshold and the continuous walking time exceeds the threshold, the status is judged as route anomaly. Regional anomalies are judged by regional unfamiliarity, which is represented by the Euclidean distance between the current location and historically frequently visited locations. Let historically frequently visited locations be... Current location Calculate the Euclidean distance between frequently visited historical locations and the current location: If the minimum Euclidean distance exceeds a preset multiple of the individual's maximum activity radius, it is judged as an area anomaly; the criteria for judging time feature anomalies include abnormal travel duration and abnormal travel time period; the criteria for judging behavioral feature anomalies is that the number of consecutive time trajectory repetitions exceeds a threshold; the criteria for judging device status anomalies include GPS signal loss exceeding a threshold and the user continuing to move, and communication signal strength being lower than a threshold and the user continuing to move.

[0010] The data storage module includes a raw database, an abnormal behavior database, an early warning database, a personal database, and an early warning response log database. The raw database stores the collected raw data, the abnormal behavior database stores the analysis results and feature data of abnormal behavior, the early warning database stores the time, type, location, and response records of all early warning events, the personal database stores personal basic information, travel behavior baseline models, normal gait benchmark models, and model update logs, and the early warning response log database stores the entire record of the early warning response process.

[0011] A method for implementing an early warning system for monitoring accidental falls and getting lost includes: Step 1, system initialization and device binding; Step 2, data acquisition; Step 3, abnormal behavior analysis; and Step 4, tiered early warning. In step one above, the device information input module (21) in the cloud is accessed through the mobile terminal (3) to input the device terminal number, basic information of the elderly, contact information of family members, and emergency contact information, and to complete the binding of the user account and the device. After the main shoe monitoring terminal and the slave shoe monitoring terminal are powered on, a synchronous communication link between the main and slave devices is established, and the main shoe monitoring terminal establishes communication with the cloud. In step two above, after the elderly wear the insole that integrates the main shoe monitoring terminal and the secondary shoe monitoring terminal, the device collects the elderly's GPS positioning data and foot data. The foot data includes the quaternion of the feet, triaxial acceleration, triaxial angular velocity data, foot pressure distribution and pressure change curve data, and uploads them to the cloud. In step three above, the cloud performs abnormal behavior analysis and detection based on the received data through the abnormal behavior analysis module. In step four above, the graded early warning module executes graded early warning based on the output of the abnormal behavior analysis module.

[0012] In step three, the abnormal behavior analysis and detection specifically includes: performing transmission disconnection detection through the transmission disconnection early warning unit; generating a transmission disconnection abnormality flag if no valid data packet is received from the device after a preset time; performing basic posture recognition based on bipedal pressure data through the basic posture recognition unit and outputting the elderly person's current posture label through the fall judgment unit; performing fall judgment through the fall judgment unit; generating a fall flag if a normal gait does not appear within a preset time after an abnormal gait in a standing or walking posture; performing missing anomaly detection by the multi-dimensional missing anomaly detection unit in combination with real-time GPS data and travel behavior baseline model; and detecting fence crossing events based on real-time GPS data by the dynamic geofence management unit and generating a fence crossing flag.

[0013] In step four, the tiered early warning system is as follows: Level 1 warning trigger conditions include device battery level below 20%, communication outage time exceeding 5 minutes, single-dimensional travel anomaly, and fence crossing exceeding 5 minutes. When a Level 1 warning is triggered, a warning message is pushed to the mobile device. Level 2 warning trigger conditions include communication outage time exceeding 15 minutes, two-dimensional travel anomaly, and continuous fence crossing exceeding 10 minutes. When a Level 2 warning is triggered, a warning message is pushed to the mobile device, and an SMS notification is sent to the emergency contact to remind them to verify the elderly person's status. Level 3 warning trigger conditions include communication outage time exceeding 30 minutes, three-dimensional travel anomaly, a fall, and fence crossing exceeding 20 minutes. When a Level 3 warning is triggered, a warning message is pushed to the mobile device, and an emergency contact is called via voice call.

[0014] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention distinguishes postures through a basic posture recognition unit, establishes a normal gait baseline through a normal gait baseline modeling unit, and determines fall actions based on posture and the normal gait baseline through a fall determination unit, effectively reducing false alarm rates. It constructs a personal travel behavior baseline model through a personalized travel modeling unit, and performs four-dimensional anomaly detection based on space, time, behavior, and device status through a multi-dimensional missing person anomaly detection unit. Combined with adaptive dynamic geofencing, this solves the problem of rigid missing person determination mechanisms in existing systems. The tiered early warning module adopts a three-level differentiated early warning system, matching corresponding response strategies according to the anomaly risk level, rationally allocating monitoring resources, and improving the user experience. Attached Figure Description

[0015] Figure 1 This is a system structure block diagram of the present invention; Figure 2 This is a structural block diagram of the main shoe monitoring terminal of the present invention; Figure 3 This is a structural block diagram of the shoe monitoring end of the present invention; Figure 4 This is a cloud-based structural block diagram of the present invention; Figure 5 This is a structural block diagram of the abnormal behavior analysis module of the present invention; Figure 6 This is a block diagram of the data storage module structure of the present invention; Figure 7 This is a flowchart of the method of the present invention.

[0016] In the diagram: 1. Device end; 11. Master shoe monitoring end; 111. First six-axis sensor module; 112. First Bluetooth module; 113. Communication module; 114. GPS module; 115. First power module; 116. First pressure sensor array module; 12. Slave shoe monitoring end; 121. Second six-axis sensor module; 122. Second Bluetooth module; 123. Second power module; 124. Second pressure sensor array module; 2. Cloud; 21. Device information input module; 22. Abnormal behavior analysis module; 221. Transmission Disconnection Early Warning Unit; 222. Basic Posture Recognition Unit; 223. Normal Gait Baseline Modeling Unit; 224. Fall Detection Unit; 225. Personalized Travel Modeling Unit; 226. Multi-Dimensional Loss Anomaly Detection Unit; 227. Dynamic Geofencing Management Unit; 23. Hierarchical Early Warning Module; 24. Data Storage Module; 241. Raw Database; 242. Abnormal Behavior Database; 243. Early Warning Database; 244. Personal Database; 245. Early Warning Response Log Database; 3. Mobile Terminal. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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 are within the scope of protection of the present invention.

[0018] Please see the appendix Figure 1 - Appendix Figure 6This invention provides an embodiment of an early warning system for monitoring accidental falls and getting lost, comprising a device terminal 1, a cloud terminal 2, and a mobile terminal 3. The cloud terminal 2 establishes data connections with both the device terminal 1 and the mobile terminal 3. The cloud terminal 2 includes a device information input module 21, an abnormal behavior analysis module 22, a graded early warning module 23, and a data storage module 24, and the data storage module 24 establishes data connections with the device information input module 21, the abnormal behavior analysis module 22, and the graded early warning module 23, respectively. The device terminal 1 includes a master shoe monitoring terminal 11 and a slave shoe monitoring terminal 12. The master shoe monitoring terminal 11 includes a first six-axis sensor module 111, a first Bluetooth module 112, a communication module 113, a GPS module 114, a first power module 115, and a second... A pressure sensor array module 116 includes a first six-axis sensor module 111 for acquiring quaternion, triaxial acceleration, and angular velocity data. The master shoe monitoring terminal 11 establishes a data connection with the slave shoe monitoring terminal 12 via a first Bluetooth module 112, establishes a data connection with the cloud 2 via a communication module 113, and acquires GPS data via a GPS module 114. A first power module 115 supplies power to the master shoe monitoring terminal 11. The first pressure sensor array module 116 is used to detect the plantar pressure distribution and pressure change curve. The slave shoe monitoring terminal 12 includes a second six-axis sensor module 121, a second Bluetooth module 122, a second power module 123, and a second pressure sensor array module 124. The second six-axis sensor module 121 is used to acquire quaternion data. The device monitors the foot using three-axis acceleration and angular velocity data. A second Bluetooth module 122 establishes a data connection between the slave shoe monitoring terminal 12 and the master shoe monitoring terminal 11. A second power module 123 powers the slave shoe monitoring terminal 12. A second pressure sensor array module 124 detects the plantar pressure distribution and pressure change curve. The abnormal behavior analysis module 22 includes a transmission disconnection warning unit 221, a basic posture recognition unit 222, a normal gait benchmark modeling unit 223, a fall detection unit 224, a personalized travel modeling unit 225, a multi-dimensional missing person anomaly detection unit 226, and a dynamic geofence management unit 227. The transmission disconnection warning unit 221 monitors the communication status between the device 1 and the cloud 2. The basic posture recognition unit 222 is based on bipedal foot... The system includes: a pressure data unit for real-time identification of basic human postures, including standing, walking, and sitting postures; a normal gait baseline modeling unit 223 for establishing a normal gait baseline model based on bipedal data during standing, walking, and climbing stairs; a fall detection unit 224 for determining whether a fall has occurred; a personalized travel modeling unit 225 for constructing a personalized multi-dimensional travel behavior baseline model, specifically including a temporal baseline and a spatial baseline. The temporal baseline includes daily travel time periods and single travel duration, while the spatial baseline includes a set of frequently visited locations, a set of travel routes, and a maximum activity radius; a multi-dimensional missing person anomaly detection unit 226 for detecting travel anomalies; and a dynamic geofence management unit 227 for automatically adjusting the geofence range according to time and weather, while also supporting custom adjustments.The judgment rule of the fall judgment unit 224 is as follows: when the basic posture is either standing or walking, if a normal gait does not appear within a preset time after an abnormal gait, it is judged as a fall; the multi-dimensional missing anomaly detection unit 226 comprehensively judges the travel anomaly by considering four dimensions: spatial features, temporal features, behavioral features, and device status. If any two of the above four dimensions are abnormal and the duration exceeds a preset threshold, it is judged as a suspected missing state. If any three of the above four dimensions are abnormal and the duration exceeds a preset threshold, it is judged as a confirmed missing state. Among them, the spatial feature anomaly judgment criteria include route anomaly and regional anomaly. Route anomaly is judged by route deviation, which is represented by the Frescher distance similarity between the current route and the historical fixed route. If the Frescher distance similarity exceeds the threshold and the continuous walking time exceeds the threshold, it is judged as a route anomaly. Regional anomaly is judged by regional unfamiliarity, which is represented by the Euclidean distance between the current location and the historical frequently visited location. Let the historical frequently visited location be... Current location Calculate the Euclidean distance between frequently visited historical locations and the current location: If the minimum Euclidean distance exceeds a preset multiple of the individual's maximum activity radius, it is determined to be an area anomaly. The criteria for judging time feature anomalies include abnormal travel duration and abnormal travel time period. The criteria for judging behavioral feature anomalies is that the number of consecutive time trajectory repetitions exceeds a threshold. The criteria for judging device status anomalies include GPS signal loss exceeding a threshold and the user continuing to move, and communication signal strength being lower than a threshold and the user continuing to move. The data storage module 24 includes a raw database 241, an abnormal behavior database 242, an early warning database 243, a personal database 244, and an early warning response log database 245. The raw database 241 is used to store the collected raw data. The abnormal behavior database 242 is used to store the abnormal behavior analysis results and feature data. The early warning database 243 is used to store the time, type, location, and response records of all early warning events. The personal database 244 is used to store personal basic information, travel behavior baseline model, normal gait benchmark model, and model update log. The early warning response log database 245 is used to store the entire record of the early warning response process.

[0019] Please see the appendix Figure 7 The present invention provides an embodiment of an early warning system for monitoring accidental falls and getting lost, comprising: step one, system initialization and device binding; step two, data acquisition; step three, abnormal behavior analysis; and step four, tiered early warning. In step one above, the device information input module 21 of the cloud 2 is accessed through the mobile terminal 3 to input the device terminal 1 number, the elderly's basic information, family contact information, and emergency contact information, and complete the binding of the user account and the device. After the main shoe monitoring terminal 11 and the slave shoe monitoring terminal 12 are powered on, a master-slave device synchronous communication link is established, and the main shoe monitoring terminal 11 establishes communication with the cloud 2. In step two above, after the elderly wear the insole with integrated main shoe monitoring terminal 11 and secondary shoe monitoring terminal 12, the device terminal 1 collects the elderly's GPS positioning data and foot data. The foot data includes the quaternion of the feet, triaxial acceleration, triaxial angular velocity data, foot pressure distribution and pressure change curve data, and uploads them to the cloud 2. In step three above, cloud 2 performs abnormal behavior analysis and detection based on the received data through abnormal behavior analysis module 22. The abnormal behavior analysis and detection specifically includes: performing transmission disconnection detection through transmission disconnection early warning unit 221. If a valid data packet is not received from device 1 for a preset time, a transmission disconnection abnormality mark is generated; performing basic posture recognition based on bipedal pressure data through basic posture recognition unit 222 and outputting the elderly's current posture label; performing fall judgment through fall judgment unit 224. If a normal gait does not appear within a preset time after an abnormal gait in a standing or walking posture, a fall mark is generated; performing loss anomaly detection by multi-dimensional missing person anomaly detection unit 226 in combination with real-time GPS data and travel behavior baseline model and generating a corresponding level of travel anomaly mark; and performing loss anomaly detection by dynamic geofence management unit 227 based on real-time GPS data and generating a geofence crossing mark. In step four above, the graded early warning module 23 executes graded early warnings based on the output of the abnormal behavior analysis module 22. The graded early warnings are as follows: Level 1 early warning trigger conditions include device battery level below 20%, communication disconnection time exceeding 5 minutes, single-dimensional travel abnormality, and fence crossing exceeding 5 minutes. When a Level 1 early warning is triggered, an early warning message is pushed to the mobile terminal 3. Level 2 early warning trigger conditions include communication disconnection time exceeding 15 minutes, two-dimensional travel abnormality, and continuous fence crossing exceeding 10 minutes. When a Level 2 early warning is triggered, an early warning message is pushed to the mobile terminal 3, and an SMS notification is sent to the emergency contact to remind them to verify the elderly person's status. Level 3 early warning trigger conditions include communication disconnection time exceeding 30 minutes, three-dimensional travel abnormality, falling, and fence crossing exceeding 20 minutes. When a Level 3 early warning is triggered, an early warning message is pushed to the mobile terminal 3, and an emergency contact is called via voice call.

[0020] Experimental example: To verify the practical effectiveness of this invention in fall detection, missing person determination, and early warning resource allocation, 60 healthy adults aged 20-40 years were selected as experimental subjects and randomly divided into an experimental group (30 people, using the system described in this invention) and a control group (30 people, using the mainstream static threshold fall detection + static geofencing system) for comparative experiments. During the experiment, various fall scenarios were simulated 120 times and various missing person scenarios were simulated 80 times according to the actual accidental behavior characteristics of the elderly. At the same time, all false triggering events and early warning responses in daily activities were recorded. The experimental results are shown in the table below: Comparison Projects Experimental group (system of this invention) Control group (existing system) Fall detection false alarm rate 0.8% 42.7% Fall detection false negative rate 0.5% 18.3% False alarm rate for missing persons determination 1.7% 51.2% Missed reporting rate for missing persons 0.3% 12.5% Average lead time for missing persons warning 12.3 minutes 0 minutes Level 1 warning rate 82.5% 15.2% Average response time for Level III early warning 2.1 minutes 5.7 minutes Average daily monitoring resource usage time per person 1.2 minutes 8.7 minutes Based on the above, the advantages of this invention are that, when in use, the user account is bound to the device terminal 1 by accessing the device information input module 21 of the cloud terminal 2 via the mobile terminal 3. After the main shoe monitoring terminal 11 and the slave shoe monitoring terminal 12 are powered on, the first power module 115 and the second power module 123 provide continuous power to them respectively. The main shoe monitoring terminal 11 automatically establishes a Bluetooth synchronous communication link with the second Bluetooth module 122 of the slave shoe monitoring terminal 12 through the first Bluetooth module 112, and establishes communication with the cloud terminal 2 through the communication module 113. After the elderly wear the insole integrating the main shoe monitoring terminal 11 and the slave shoe monitoring terminal 12, the first six-axis sensor module 111 and the second six-axis sensor module 121 synchronously collect the quaternion, triaxial acceleration, and angular velocity of both feet. The pressure sensor array module 116 and the second pressure sensor array module 124 synchronously collect data on the plantar pressure distribution and pressure change curves of both feet. The GPS module 114 of the main shoe monitoring terminal 11 collects real-time GPS positioning data. The collected data from the shoe monitoring terminal 12 is transmitted to the main shoe monitoring terminal 11 via the second Bluetooth module 122 for timestamp alignment and encryption. Then, it is uploaded to the cloud 2 via the communication module 113. After receiving the data and completing the integrity verification, the cloud 2 stores all the raw data in the raw database 241 of the data storage module 24 and calls the abnormal behavior analysis module 22 to perform multi-dimensional anomaly detection: the transmission disconnection early warning unit 221 monitors the communication link status between the device terminal 1 and the cloud 2 in real time, and the basic posture recognition unit 2... 22. Based on bipedal pressure data, the system identifies three basic postures: standing, walking, and sitting. The normal gait baseline modeling unit 223 constructs a personal normal gait baseline model based on daily standing, walking, and stair climbing data from 7-14 consecutive days, and stores it in the personal database 244 of the data storage module 24. The fall detection unit 224, upon detecting a sudden interruption of the normal gait while the basic posture is standing / walking, opens a 10-second recovery time window. If no gait matching the personal normal gait baseline appears within the window, it is determined to be an accidental fall. The personalized travel modeling unit 225 constructs a personalized multi-dimensional travel behavior baseline model including temporal and spatial baselines based on 7-14 consecutive days of GPS data, and stores it in the personal database 244. Multi-dimensional missing person anomalies are also addressed. The detection unit 226 combines real-time GPS data with travel behavior baselines to comprehensively determine the level of travel anomalies from four dimensions: space, time, behavior, and device status. The dynamic geofence management unit 227 automatically adjusts the fence range according to time and weather and detects fence boundary crossing events in real time. For example, at night, the fence range is reduced to the building range, and it returns to normal during the day. The fence range remains unchanged on sunny / cloudy days and is reduced to the community range on rainy / snowy days. After receiving all the abnormal identifiers output by the abnormal behavior analysis module 22, the graded early warning module 23 divides them into three levels of early warning and executes differentiated response strategies. All abnormal analysis results are stored in the abnormal behavior database 242, early warning events are stored in the early warning database 243, and the entire early warning response process is stored in the early warning response log database 245.Guardians can use a mobile device to view the elderly person's location, movement status, and device battery level in real time, receive tiered alerts, view historical tracks and alert records, and confirm or cancel false alarms. The system automatically optimizes anomaly detection parameters based on false alarm feedback submitted by the guardian.

[0021] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. An early warning system for monitoring accidental falls and getting lost, comprising a device (1), a cloud (2), and a mobile terminal (3), characterized in that: The cloud (2) establishes data connections with the device (1) and the mobile terminal (3) respectively. The cloud (2) includes a device information input module (21), an abnormal behavior analysis module (22), a graded early warning module (23) and a data storage module (24). The data storage module (24) establishes data connections with the device information input module (21), the abnormal behavior analysis module (22) and the graded early warning module (23) respectively.

2. The early warning system for monitoring accidental falls and getting lost according to claim 1, characterized in that: The device end (1) includes a master shoe monitoring end (11) and a slave shoe monitoring end (12). The master shoe monitoring end (11) includes a first six-axis sensor module (111), a first Bluetooth module (112), a communication module (113), a GPS module (114), a first power module (115), and a first pressure sensor array module (116). The first six-axis sensor module (111) is used to acquire quaternion, triaxial acceleration, and angular velocity data. The master shoe monitoring end (11) establishes a data connection with the slave shoe monitoring end (12) using the first Bluetooth module (112), establishes a data connection with the cloud (2) using the communication module (113), acquires GPS data using the GPS module (114), the first power module (115) is used to supply power to the master shoe monitoring end (11), and the first pressure sensor array module (116) is used to detect the foot pressure distribution and pressure change curve.

3. The early warning system for monitoring accidental falls and getting lost according to claim 2, characterized in that: The foot monitoring terminal (12) includes a second six-axis sensor module (121), a second Bluetooth module (122), a second power module (123), and a second pressure sensor array module (124). The second six-axis sensor module (121) is used to acquire quaternion, triaxial acceleration, and angular velocity data. The foot monitoring terminal (12) establishes a data connection with the master foot monitoring terminal (11) using the second Bluetooth module (122). The second power module (123) is used to supply power to the foot monitoring terminal (12). The second pressure sensor array module (124) is used to detect the foot pressure distribution and pressure change curve.

4. The early warning system for monitoring accidental falls and getting lost according to claim 1, characterized in that: The abnormal behavior analysis module (22) includes a transmission disconnection early warning unit (221), a basic posture recognition unit (222), a normal gait benchmark modeling unit (223), a fall judgment unit (224), a personalized travel modeling unit (225), a multi-dimensional missing person anomaly detection unit (226), and a dynamic geofence management unit (227). The transmission disconnection early warning unit (221) is used to monitor the communication status between the device (1) and the cloud (2). The basic posture recognition unit (222) identifies the basic posture of the human body in real time based on the plantar pressure data of both feet. The basic posture includes standing, walking, and sitting postures. The normal gait benchmark modeling unit... (223) is used to establish a normal gait baseline model based on bipedal data when standing, walking and going up and down stairs. The fall judgment unit (224) determines whether a fall has occurred. The personalized travel modeling unit (225) is used to construct a personalized multi-dimensional travel behavior baseline model, which specifically includes a time baseline and a spatial baseline. The time baseline includes daily travel time and single travel duration. The spatial baseline includes a set of frequently visited locations, a set of travel routes and the maximum activity radius. The multi-dimensional missing anomaly detection unit (226) is used to detect travel anomalies. The dynamic geofence management unit (227) is used to automatically adjust the fence range according to time and weather, and also supports custom adjustments.

5. An early warning system for monitoring accidental falls and getting lost according to claim 4, characterized in that: The judgment rule of the fall judgment unit (224) is: when the basic posture is either standing or walking, if a normal gait does not appear within a preset time after the abnormal gait, it is judged as a fall.

6. The early warning system for monitoring accidental falls and getting lost according to claim 4, characterized in that: The multi-dimensional missing person anomaly detection unit (226) comprehensively judges travel anomalies based on four dimensions: spatial features, temporal features, behavioral features, and device status. If any two of the above four dimensions show anomalies and the duration exceeds a preset threshold, it is judged as a suspected missing person state. If any three of the above four dimensions show anomalies and the duration exceeds a preset threshold, it is judged as a confirmed missing person state. Among them, the spatial feature anomaly judgment criteria include route anomalies and regional anomalies. Route anomalies are judged by route deviation, which is represented by the Frescher distance similarity between the current route and the historical fixed route. If the Frescher distance similarity exceeds the threshold and the continuous walking time exceeds the threshold, it is judged as a route anomaly. Regional anomalies are judged by regional unfamiliarity, which is represented by the Euclidean distance between the current location and the historical frequently visited locations. Let the historical frequently visited locations be... Current location Calculate the Euclidean distance between frequently visited historical locations and the current location: If the minimum Euclidean distance exceeds a preset multiple of the individual's maximum activity radius, it is judged as an area anomaly; the criteria for judging time feature anomalies include abnormal travel duration and abnormal travel time period; the criteria for judging behavioral feature anomalies is that the number of consecutive time trajectory repetitions exceeds a threshold; the criteria for judging device status anomalies include GPS signal loss exceeding a threshold and the user continuing to move, and communication signal strength being lower than a threshold and the user continuing to move.

7. An early warning system for monitoring accidental falls and getting lost according to claim 1, characterized in that: The data storage module (24) includes a raw database (241), an abnormal behavior database (242), an early warning database (243), a personal database (244), and an early warning response log database (245). The raw database (241) is used to store the collected raw data. The abnormal behavior database (242) is used to store the abnormal behavior analysis results and feature data. The early warning database (243) is used to store the time, type, location, and response records of all early warning events. The personal database (244) is used to store personal basic information, travel behavior baseline model, normal gait benchmark model, and model update log. The early warning response log database (245) is used to store the entire record of the early warning response process.

8. A method for implementing an early warning system for monitoring accidental falls and getting lost, comprising: Step 1, system initialization and device binding; Step 2, data acquisition; Step 3, abnormal behavior analysis; Step 4, tiered early warning; characterized in that: In step one above, the device information input module (21) of the cloud (2) is accessed through the mobile terminal (3) to input the device terminal (1) number, the elderly’s basic information, family contact information, and emergency contact information, and complete the binding of the user account and the device. After the main shoe monitoring terminal (11) and the slave shoe monitoring terminal (12) are powered on, a synchronous communication link between the main and slave devices is established, and the main shoe monitoring terminal (11) establishes communication with the cloud (2). In step two above, after the elderly wear the insole of the integrated main shoe monitoring terminal (11) and the shoe monitoring terminal (12), the device terminal (1) collects the elderly's GPS positioning data and foot data. The foot data includes the quaternion of the feet, triaxial acceleration, triaxial angular velocity data, foot pressure distribution and pressure change curve data, and uploads them to the cloud (2). In step three above, the cloud (2) performs abnormal behavior analysis and detection based on the received data through the abnormal behavior analysis module (22); In step four above, the graded early warning module (23) executes graded early warning based on the output of the abnormal behavior analysis module (22).

9. A method for implementing an early warning system for monitoring accidental falls and getting lost, as described in claim 8, characterized in that: In step three, the abnormal behavior analysis and detection specifically includes: performing transmission disconnection detection through the transmission disconnection early warning unit (221), generating a transmission disconnection abnormality mark if a valid data packet is not received from the device (1) within a preset time; performing basic posture recognition based on bipedal pressure data through the basic posture recognition unit (222) and outputting the current posture label of the elderly through the fall judgment unit (224); performing fall judgment through the fall judgment unit (224), generating a fall mark if a normal gait does not appear within a preset time after an abnormal gait in a standing or walking posture; performing a multi-dimensional missing anomaly detection unit (226) by combining real-time GPS data and travel behavior baseline model to perform missing anomaly detection and generate a travel anomaly mark of the corresponding level; and performing a dynamic geofence management unit (227) by detecting fence crossing events based on real-time GPS data and generating a fence crossing mark.

10. A method for implementing an early warning system for monitoring accidental falls and getting lost, as described in claim 8, characterized in that: In step four, the tiered early warning is as follows: the triggering conditions for a level one early warning include the device battery level being below 20%, communication disconnection time exceeding 5 minutes, abnormal travel in a single dimension, and fence crossing exceeding 5 minutes. When a level one early warning is triggered, an early warning message is pushed to the mobile terminal (3); the triggering conditions for a level two early warning include the communication disconnection exceeding 15 minutes, abnormal travel in two dimensions, and fence crossing exceeding 10 minutes. When a level two early warning is triggered, an early warning message is pushed to the mobile terminal (3), and an SMS notification is sent to the emergency contact to remind them to verify the elderly person's status. The triggering conditions for a Level 3 warning include communication disconnection time exceeding 30 minutes, abnormal travel in three dimensions, falling, and fence crossing exceeding 20 minutes. When a Level 3 warning is triggered, the warning information is pushed to the mobile terminal (3), and an emergency contact is called via voice call.

Images