An automated external defibrillator active rescue guidance system

Abstract

The application relates to an automatic external defibrillator active rescue guiding system based on multi-source information fusion and augmented reality navigation, which comprises an emergency discovery layer, an intelligent dispatch center, a rescuer AR navigation implementation layer and an AED box operation guidance implementation layer; the emergency discovery layer provides multi-source emergency help information; the intelligent dispatch center is in communication connection with the emergency discovery layer, receives emergency information collected from the multiple information sources, makes AED matching, rescuer screening and path planning; the rescuer AR navigation implementation layer is in communication connection with the intelligent dispatch center, sends emergency rescue notification to the optimal rescuer, realizes AR real scene navigation or map navigation; the AED box operation guidance implementation layer synchronously guides the optimal rescuer through animation and voice. The application provides a campus-based AED active rescue guiding system and can be applied to campus emergency rescue.

Description

Technical Field

[0001] This invention relates to the field of smart medical emergency technology, specifically to an active rescue guidance system for automated external defibrillators based on multi-source information fusion and augmented reality navigation. Background Technology

[0002] Sudden cardiac arrest (SCA) is one of the leading causes of death worldwide. The golden window for resuscitation is within 4 minutes of onset; for every minute defibrillation is delayed, the patient's survival rate decreases by 7% to 10%. Automated external defibrillators (AEDs), as the most effective device for resuscitating cardiac arrest, are widely used in public places such as airports, shopping malls, schools, and stadiums.

[0003] However, the current deployment and use of AEDs have the following prominent problems:

[0004] (1) Delay in discovering emergency situations

[0005] The current AED rescue system relies heavily on eyewitnesses to discover and report incidents. In complex environments such as large campuses, shopping malls, and stadiums, eyewitnesses may be unable to recognize a cardiac arrest event and initiate rescue procedures immediately due to panic, lack of medical knowledge, or obstructed vision. Traditional surveillance cameras are only used for post-incident review and lack real-time intelligent analysis capabilities, resulting in an average time of several minutes from when the patient collapses to when rescue is initiated, far exceeding the golden window for rescue.

[0006] (2) Difficulty in locating and obtaining AEDs

[0007] Although some cities have established AED map lookup systems, rescuers are often in a state of high tension in emergency situations, making it difficult to quickly locate AEDs. In indoor environments (such as school buildings and gymnasiums), GPS signals are severely attenuated, and traditional positioning is not accurate enough. Rescuers can easily get lost in complex building structures, delaying the time to obtain AEDs.

[0008] (3) Lack of intelligent matching of rescuers

[0009] Existing systems typically use broadcast alarms (such as public address systems or WeChat group notifications), which cannot accurately match rescuers based on their real-time location, first aid qualifications, and response history. Faculty members with first aid certificates may be far from the scene, while ordinary students nearby may lack confidence and dare not provide assistance, resulting in the optimal rescue resources not being effectively deployed.

[0010] (4) Insufficient AED operation instructions

[0011] For rescuers with no experience using AEDs, steps such as opening the box, attaching electrode pads, analyzing heart rhythm, and defibrillation / CPR (cardiopulmonary resuscitation) still present operational obstacles. Most existing AED boxes are passive storage devices, lacking interactive operating instructions. Rescuers are prone to operational errors under high-pressure environments, which can affect the effectiveness of rescue efforts.

[0012] (5) Multi-source information silos

[0013] In campus security systems, surveillance cameras, wearable devices, mobile apps, and voice pickups operate independently, and their data is not integrated or interconnected. A single information source may result in false alarms (e.g., a camera misinterpreting a fall as a stumble) or missed alarms (e.g., wearable devices not covering all personnel). The lack of cross-validation mechanisms leads to insufficient system reliability.

[0014] In summary, the existing AED rescue system suffers from five core defects: slow detection, difficulty in location, poor matching, weak guidance, and scattered information. These defects result in excessively long total time from cardiac arrest to the first defibrillation, low operational accuracy, and significant psychological stress on rescuers, ultimately impacting the patient's resuscitation success rate. Therefore, there is an urgent need for a proactive rescue guidance system that can actively detect emergencies, intelligently allocate rescue resources, accurately guide rescuers, and provide full assistance in AED operation. Summary of the Invention

[0015] The technical problem solved by this invention is to provide an active rescue guidance system for automated external defibrillators (AEDs); based on multi-source information fusion and augmented reality navigation, it realizes active rescue guidance for AEDs.

[0016] The technical solution of the present invention to solve the above-mentioned technical problems is:

[0017] The system includes an emergency detection layer, an intelligent dispatch center, an AR navigation implementation layer for rescuers, and an AED box operation guidance implementation layer.

[0018] The emergency detection layer includes: an AI posture analysis module for campus surveillance cameras, used to detect the posture of a person falling in the surveillance footage; a heart rate abnormality detection module for smart wearable devices, used to detect sudden cardiac arrest or severe arrhythmia in the wearer; an SOS emergency call module for mobile terminals, used to receive emergency distress signals triggered by users; and a voice distress call recognition module, used to recognize preset distress keywords such as "help" and "someone has fainted" in the environment.

[0019] The intelligent dispatch center includes a server, a multi-source information fusion verification module, an AED matching module, a rescuer screening module, and a path planning module. The intelligent dispatch center is communicatively connected to the emergency detection layer, receives emergency information from multiple information source collection devices, performs multi-source information fusion verification, and determines a genuine emergency situation when at least two information sources trigger simultaneously or when the confidence level of a single information source exceeds a preset threshold. After determining a genuine emergency situation, the center locates the patient's position and automatically matches the optimal AED device and the optimal rescuer based on the patient's position, AED device position, AED device status, rescuer's position, and capability information. It then plans the optimal rescue path for the optimal rescuer, from their current position to the optimal AED device and then from the optimal AED device to the patient's position.

[0020] The rescuer AR navigation implementation layer is connected to the intelligent dispatch center to achieve the following: sending an emergency rescue notification to the mobile terminal of the optimal rescuer, the notification including the patient's location, the AED device's location, and the rescue route; and guiding the optimal rescuer to the optimal AED device in real time through AR real-scene navigation or map navigation.

[0021] The aforementioned AED box operation guidance layer is set in the AED device box and enables the following: when the optimal rescuer arrives at the AED device, the optimal rescuer is guided through touch screen animation and voice synchronously to complete the operation steps of opening the AED box, taking out the device, attaching electrode pads, analyzing heart rhythm, and performing defibrillation; after the optimal rescuer arrives at the patient's location with the AED device, the AED operation is continued to be completed through voice guidance.

[0022] The multi-source information fusion verification module is used to perform fusion analysis and cross-verification on emergency alarms from different information sources to determine whether they are genuine cardiac arrest emergencies. The multi-source information fusion verification includes: time window alignment, which aligns emergency information from different information sources by timestamp, and only determines a genuine emergency when multiple information sources trigger simultaneously within a preset time window; spatial location association, which increases the confidence of the emergency when the locations of multiple information sources are within a preset distance range; and single information source confidence assessment, which calculates the confidence based on the historical accuracy of the information source when only one information source triggers the alarm, and determines a genuine emergency when the confidence exceeds 0.85.

[0023] The AED matching module is used to comprehensively evaluate and select the most suitable AED device for the current emergency dispatch among multiple available AED devices. This includes calculating the distance from each available AED device to the patient's location, querying the consumable status of each AED device, and querying the device health status of each AED device. Based on the distance, consumables, and health status, and according to the corresponding weighting coefficients, the AED device with the highest comprehensive score is selected as the optimal AED device.

[0024] The rescuer screening module is used to comprehensively evaluate and select the most suitable rescuer from multiple potential rescuers, including: obtaining the location information of all registered users on campus, screening registered users within a preset range of the patient's location, prioritizing registered users with first aid training certificates, prioritizing registered users with short historical response times, excluding registered users who have participated in other emergency rescue missions, and selecting the registered user with the highest comprehensive score as the optimal rescuer.

[0025] The path planning module plans a reasonable path based on the determined location of the rescuer and the matched AED location, according to map data. This includes: a first path from the rescuer's current location to the optimal AED device, taking into account the floor, elevator / staircase availability, and real-time pedestrian density; a second path from the optimal AED device to the patient's location, taking into account the portability of the AED device and the accessibility of the patient's location; and dynamic path updates, automatically replanning the path and pushing it to the rescuer when obstacles are detected on the path.

[0026] Implementation of AR navigation for rescuers:

[0027] When the rescuer turns on their phone camera, the system overlays directional arrows, distance indicators, and AED device location markers onto the camera screen.

[0028] Precise indoor positioning via Bluetooth Beacon;

[0029] Provide voice prompts at key locations such as corners, stairwells, and elevator entrances;

[0030] It displays the estimated arrival time and remaining distance in real time.

[0031] The AED box operation guidance implementation layer is as follows:

[0032] The AED device's touchscreen is activated: When a rescuer reaches within 2 meters of the device, the touchscreen will automatically highlight and play a voice prompt saying "Please open the box and take out the AED."

[0033] Unboxing animation guidance: The touch screen displays an animation demonstration of opening the box and taking out the AED, while the voice prompts "Please press the button on the box to open the door and take out the AED device";

[0034] Electrode placement guidance: After the AED device is powered on, the touch screen displays a human body outline and uses flashing arrows to indicate the electrode placement position, with voice guidance "Please place the electrode as shown in the diagram";

[0035] Defibrillation safety confirmation: When the AED analyzes the heart rhythm and recommends defibrillation, the camera in the enclosure detects people in the vicinity. If someone is detected approaching, a voice announcement will say "Everyone move away from the patient" until it is confirmed that it is safe to perform defibrillation.

[0036] The AI ​​posture analysis module for the campus surveillance camera implements the following: human detection, using the YOLOv8-nano model to detect human targets in the surveillance footage; posture estimation, using the MediaPipe Pose model to extract 33 human key points; fall detection, triggering an emergency signal when the torso tilt angle is greater than 60 degrees, the key point remains still for more than 5 seconds, and the human body rapidly changes from a standing state to a fallen state; and multi-camera collaboration, automatically calling adjacent cameras to confirm the viewpoint when a single camera detects a fall, improving detection accuracy.

[0037] The AED box operation guidance implementation layer also includes a CPR auxiliary guidance module, which enables: when the AED analyzes the heart rhythm and recommends CPR, it plays a metronome sound through a speaker to guide the rescuer's compression rhythm; it provides voice guidance on compression depth and rebound requirements, and guides the ratio of compressions to artificial respiration; and it prompts the rescuer to take turns every 2 minutes.

[0038] The AED box operation guidance implementation layer also includes an emergency coordination module, which enables: automatically dialing 120 emergency number, sending a distress message containing the precise location, pushing emergency notifications to the school clinic and security office, and providing voice guidance for the division of labor when multiple rescuers arrive at the scene.

[0039] The AED active rescue guidance system provided by this invention has the following significant beneficial effects:

[0040] 1. Multi-dimensional proactive perception enables second-level detection of emergencies.

[0041] This invention achieves proactive and comprehensive monitoring of emergencies through a four-layer sensing network (AI video analysis, smart wearable device heart rate monitoring, mobile phone SOS emergency call, and voice call recognition). Compared to the traditional passive discovery mode that relies on human witnessing, the system can automatically issue an alarm within an average of 8 seconds after a patient falls to the ground, reducing the discovery time by 82% and ensuring that rescue is initiated within the golden rescue window, fundamentally solving the problem of "slow discovery".

[0042] 2. Cross-validation of multi-source information significantly improves system reliability.

[0043] This invention designs a multi-source information fusion verification mechanism, which cross-verifies multi-source alarms through time windows (e.g., 30 seconds) and spatial correlations (e.g., a 50-meter threshold). A single information source must achieve a high confidence level (≥0.85) to trigger an alarm, while multi-source information must satisfy spatiotemporal consistency. This mechanism effectively reduces the false alarm rate of single information sources (e.g., camera misjudgments, wristband anomalies), keeping the system's false alarm rate at an extremely low level, ensuring that rescue resources are not ineffectively allocated, and solving the problems of "scattered information and poor reliability" in existing technologies.

[0044] 3. Optimal resource intelligent matching to maximize rescue efficiency.

[0045] This invention designs an optimal AED matching algorithm and an optimal rescuer matching algorithm, respectively:

[0046] AED matching: Taking into account three factors, namely distance, adequacy of consumables, and health of the device, to ensure that the nearest and best-condition AED device is dispatched;

[0047] Rescuer matching: Taking into account distance, first aid qualification certificates (1.5 times weighted), and historical average response time, priority is given to dispatching rescuers with professional qualifications and rapid response.

[0048] Compared to traditional broadcast alarms, this invention enables precise and intelligent dispatch of rescue resources, reducing rescuer response time by 83% and AED acquisition time by 64%.

[0049] 4. Augmented Reality (AR) navigation solves the challenges of indoor positioning and wayfinding.

[0050] This invention develops AR navigation functionality based on ARKit / ARCore, combining Bluetooth Beacon (iBeacon protocol, accuracy 1-3 meters) to achieve high-precision indoor positioning. Rescuers can use their mobile phone camera to overlay 3D directional arrows (green / yellow / red markers), distance indicators, and AED location markers onto the real-world scene, along with voice prompts for immersive guidance. Compared to traditional 2D map navigation, AR navigation is more intuitive, conforms to human spatial cognition habits, and effectively solves the problems of "difficult positioning and navigation" in complex indoor environments.

[0051] 5. Interactive AED operation guide significantly improves the accuracy of operation.

[0052] This invention upgrades the AED box into a smart interactive terminal, integrating a touchscreen, speaker, camera, and microwave radar sensor. Through automatic activation via human body detection, step-by-step animated demonstrations (opening the box to retrieve the AED → powering on → attaching electrode pads → analyzing heart rhythm → defibrillation / CPR), intelligent voice prompts, and a metronome-assisted CPR (110 beats / minute), it provides comprehensive "hands-on" guidance for inexperienced rescuers. Real-world testing shows that the accuracy rate increases from 45% with traditional methods to 92%, a 104% improvement, effectively solving the problem of "weak guidance."

[0053] 6. Reduce the psychological pressure on rescuers and encourage more people to participate in rescue efforts.

[0054] By eliminating navigation anxiety with AR, alleviating fear of operation with animated voice guidance, and ensuring priority response from professional rescuers through intelligent dispatching, this invention reduces rescuers' psychological stress scores from 4.5 / 5 to 2.8 / 5 (a reduction of 38%). This reduced psychological burden helps improve rescuers' operational stability and encourages more non-professionals to lend a helping hand in emergencies, thus expanding the scope of social emergency response.

[0055] 7. Significantly shortens the time to first defibrillation and improves patient survival rate.

[0056] Based on the combined effects of the above technologies, this invention reduces the total time from simulated collapse to the first defibrillation from 280 seconds to 95 seconds, a reduction of 66%. Considering that the survival rate of cardiac arrest patients decreases by 7% to 10% for every minute the defibrillation is delayed, this invention can buy patients approximately 3 precious minutes, theoretically increasing the patient's survival rate by 21% to 30%, thus possessing significant clinical value and social significance.

[0057] 8. The system architecture is scalable and applicable to a wide range of scenarios.

[0058] The four-layer sensing network, intelligent dispatch center, AR navigation module, and interactive AED box of this invention all adopt a modular design, which can flexibly adjust the equipment configuration and algorithm parameters according to different scenarios (such as shopping malls, airports, communities, and factories). For example, shopping malls can add elevator sensor data to optimize path planning, and factories can connect to industrial safety systems to realize linkage alarms, which has good versatility and scalability. Attached Figure Description

[0059] The present invention will be further described below with reference to the accompanying drawings:

[0060] Figure 1 This is a system framework diagram of the present invention;

[0061] Figure 2 This is a system flowchart of the present invention. Detailed Implementation

[0062] The following embodiments are descriptions of specific implementations of the present invention. These descriptions are intended to facilitate understanding of the technical solutions of this application by those skilled in the art and are not intended to limit the scope of protection of the technical solutions of the present invention. All equivalent transformations that can be obtained from the description of the embodiments of the present invention should be within the scope of protection of the present invention.

[0063] See Figure 1 , 2 As shown, the AED active rescue guidance system of the present invention is illustrated using a middle school campus with an area of ​​80,000 square meters, including 3 teaching buildings, 1 gymnasium, 1 library, 2 playgrounds, and approximately 3,000 teachers and students as an example. The system of the present invention includes an emergency detection layer 1, an intelligent dispatch center 2, an AR navigation implementation layer 3 for rescuers, and an AED box operation guidance implementation layer 4.

[0064] Emergency Detection Layer 1 includes: a campus surveillance camera AI analysis module, a smart wearable device heart rate detection module, a mobile phone SOS emergency call module, and a voice call recognition module.

[0065] (1) Campus surveillance camera AI analysis module

[0066] 120 high-definition network cameras (Hikvision DS-2CD3T86FWDV2-I3S, 8 megapixels, supporting H.265 encoding) were deployed throughout the school.

[0067] Two AI analytics servers (NVIDIA RTX A4000 graphics cards, Intel Xeon W-2245 processors) were deployed in the monitoring center to run YOLOv8+MediaPipe inference.

[0068] The cameras cover all public areas of the school, including corridors, stairwells, playgrounds, gymnasiums, and canteens, ensuring no blind spots;

[0069] The AI ​​analysis server receives camera video streams via the campus private network (gigabit fiber optic), performs human posture recognition on the video streams from 120 cameras, analyzes human postures in real time, and determines whether there are abnormal behaviors such as falling to the ground.

[0070] (2) Heart rate detection module for smart wearable devices

[0071] All physical education teachers, school doctors, and security personnel will be equipped with smart bracelets (Huawei Band 9, which supports heart rate monitoring and blood oxygen detection).

[0072] The wristband connects to a mobile app via Bluetooth 5.0. When it detects an abnormal heart rate (such as a heart rate <40 or >180 for 10 seconds, or an abnormal heart rate variability index), it automatically sends an alarm to the smart dispatch center.

[0073] Students may voluntarily wear smartwatches that support heart rate monitoring to join the campus security network.

[0074] (3) Mobile SOS emergency call module

[0075] Develop a campus safety app that integrates an SOS emergency call function;

[0076] Users can press the power button three times in a row or trigger SOS with one click in the APP to automatically send location information (GPS + Bluetooth Beacon fusion positioning) and on-site recordings to the intelligent dispatch center.

[0077] The app supports offline triggering. When the network is unavailable, information is cached and automatically resent when the network is restored.

[0078] (4) Voice distress call recognition module

[0079] Deploy microphone arrays (iFlytek SR-101, 6-microphone circular array, supporting 5-meter far-field pickup) in open areas such as stadiums and playgrounds.

[0080] The microphone is connected to the edge computing box (Huawei Atlas 200I DK A2) to run a speech recognition model;

[0081] The keyword database includes 20 preset words such as "help", "someone has fainted", "call 120 quickly", and "need AED", with an accuracy rate of >95%.

[0082] The intelligent dispatch center 2 includes a server, a multi-source information fusion verification module, an AED matching module, a rescuer screening module, and a route planning module.

[0083] The server includes a primary server and a backup server;

[0084] Main server: Dell PowerEdge R750, 2×Intel Xeon Gold 6348, 256GB memory, 4×NVIDIA A30 GPU;

[0085] Backup server: hot standby with the same configuration to ensure high availability;

[0086] Storage: Dell PowerStore 500T, 20TB of available capacity, for storing event logs, video recordings, and system data;

[0087] Network: Dual 10 Gigabit fiber optic access to the campus core network ensures low-latency communication.

[0088] The multi-source information fusion verification module is used to perform fusion analysis and cross-verification on emergency alarms from different information sources to determine whether they are real cardiac arrest emergencies, thus avoiding ineffective rescue dispatch caused by false alarms from a single information source.

[0089] The core parameters considered by this module include time, space, and single-source confidence; the module presets three key thresholds:

[0090] Time window: 30 seconds, used to determine whether multiple alarms occur within a similar time frame;

[0091] Spatial correlation threshold: 50 meters, used to determine whether multiple alarms originate from similar geographical locations;

[0092] Single source confidence threshold: 0.85, used to assess the credibility of a single information source.

[0093] The module verification process is as follows:

[0094] Step 1: Time Grouping Processing

[0095] After receiving all alarm information, the system first sorts them according to their chronological order. Starting from the time of the first alarm, all alarms arriving within the next 30 seconds are grouped into the same group. If the time difference between an alarm and the first alarm in the current group exceeds 30 seconds, a new time grouping is initiated. This ultimately forms several alarm sets within different time windows.

[0096] Step 2: Multi-source cross-validation

[0097] For time groups containing two or more alarms, perform a spatial correlation check:

[0098] Extract the geographic location information (latitude and longitude coordinates) of all alarms in this group;

[0099] Calculate the distance between any two alarm locations within the group;

[0100] If the distance between all pairs of locations does not exceed 50 meters, then "multi-source cross-validation passed" is determined, confirming that an emergency has occurred;

[0101] If any two locations are more than 50 meters apart, the verification is considered to have failed.

[0102] Step 3: Single-source high-confidence verification

[0103] For time groups containing only a single alarm, the algorithm directly checks the confidence level of that alarm:

[0104] If the confidence level of the alarm is greater than or equal to 0.85, it is determined to be "single source high confidence" and an emergency has been confirmed.

[0105] If the confidence level is below 0.85, the verification is considered unsuccessful.

[0106] Step 4: Distance Calculation Method

[0107] The positional distance is calculated using the Haversine Formula in spherical geometry:

[0108] Convert the latitude and longitude coordinates of two points to radians;

[0109] Using the Earth's average radius of 6371 kilometers as a benchmark;

[0110] The distance between two points on the Earth's surface arc length is obtained by using trigonometric functions, with the unit being meters.

[0111] The AED matching module is used to comprehensively evaluate and select the most suitable AED device for the current emergency situation from among multiple available AED devices.

[0112] The screening criteria for AED devices are as follows:

[0113] First, rule out unusable AED devices: devices that are currently occupied; devices with insufficient consumables (electrode pads, batteries); and devices that have failed the health self-test.

[0114] For each AED device that passed the screening, the following three indicators were calculated:

[0115] First item: Distance rating

[0116] Calculate the distance between the patient's location and the location of the AED device;

[0117] The closer the distance, the higher the score. The reciprocal of the distance is used for normalization (adding 1 to avoid division by zero).

[0118] Second item: Consumables adequacy score

[0119] The value ranges from 0 to 1, where 1 indicates that the consumables are fully available.

[0120] It reflects the remaining availability of key consumables such as electrode sheets and batteries.

[0121] Third item: Equipment health score

[0122] The value ranges from 0 to 1, with 1 indicating that the device is in optimal condition.

[0123] An assessment is conducted based on factors such as the equipment's self-inspection history and maintenance records.

[0124] The overall score is calculated based on the following weights for the three evaluation criteria:

[0125] Distance rating: 50% (highest weight, ensuring the nearest device is prioritized);

[0126] Consumable adequacy: 30%;

[0127] Equipment health status: 20%;

[0128] Iterate through all available AED devices, calculate their comprehensive scores, select the device with the highest score as the optimal scheduling target, and return the matching score value of that device.

[0129] The rescuer screening module is used to comprehensively evaluate and select the most suitable rescuer from multiple potential rescuers, and supports returning multiple candidate rescuers for the system's reference.

[0130] First, exclude unavailable rescuers: rescuers who are currently busy (performing other rescue missions); rescuers currently marked as unavailable; and rescuers who are more than 500 meters away from the patient (beyond the maximum response radius).

[0131] For each rescuer who passed the screening, the following three indicators were calculated:

[0132] First item: Distance rating

[0133] Calculate the distance between the patient's location and the rescuer's current location;

[0134] The closer the distance, the higher the score. The reciprocal of the distance is used for normalization (adding 1 to avoid division by zero).

[0135] Second item: First aid qualification assessment

[0136] If the rescuer holds a valid first aid certificate (such as a Red Cross first aid certificate, AHA first aid certificate, etc.), the score is multiplied by a weighting factor of 1.5; if there is no certificate, the base score of 1.0 is maintained.

[0137] Third item: Historical response score

[0138] Calculated based on the rescuer's past average response time;

[0139] The shorter the average response time, the higher the score. The response time is normalized by the inverse (adding 1 to avoid division by zero).

[0140] The overall score is calculated based on the following weights for the three evaluation criteria:

[0141] Distance rating: 40%;

[0142] First aid qualification rating: 40% (equally important as distance, ensuring priority for professionals);

[0143] Historical response score: 20%.

[0144] Iterate through all available rescuers, calculate their overall scores, and sort them from highest to lowest. Return the rescuer with the highest score as the optimal match; if no suitable rescuer is found, return null.

[0145] The path planning module, based on the determined location of the rescuer and the matched AED location, plans a reasonable path according to map data. In this invention, the path planning uses the A-Star algorithm to search for the optimal path on the campus graph; including:

[0146] Nodes: Rooms, corridor intersections, stairwells, elevator entrances;

[0147] Edge weight: Distance × Congestion coefficient × Floor coefficient;

[0148] Dynamic updates: The path is recalculated every 5 seconds to avoid newly appearing obstacles.

[0149] Rescuer AR navigation implementation layer 3 includes:

[0150] 1. Technical Architecture

[0151] Mobile App: Developed based on ARKit (iOS) or ARCore (Android);

[0152] Indoor positioning: Bluetooth Beacon (iBeacon protocol, Estimote Beacon), one beacon deployed every 10 meters, positioning accuracy 1-3 meters;

[0153] Map data: Campus indoor map (converted from CAD drawings), containing semantic information such as rooms, corridors, stairs, and elevators.

[0154] 2. AR navigation interface

[0155] Open your phone's camera; a 3D arrow will be overlaid on the screen to indicate the direction.

[0156] Arrow colors: green (go straight), yellow (turn), red (arrive);

[0157] Distance markers: Above the arrows, it displays "120 meters to AED" and "Turn right in 50 meters ahead";

[0158] AED marking: A flashing green icon is displayed at the location of the AED in the camera view;

[0159] Voice prompt: Every 30 seconds, it will announce "Go straight for 100 meters ahead, then turn right".

[0160] The AED cabinet operation guide for Layer 4 includes the following hardware features:

[0161] Single-board computer: Raspberry Pi 4B as the main control board; Lina touchscreen, speakers, etc.

[0162] Touchscreen: 10.1-inch capacitive screen with a resolution of 1280×800, supports glove touch;

[0163] Speakers: 2×5W, adjustable volume, maximum 95dB;

[0164] Camera: 2 megapixels, used to detect the arrival of rescuers and people in the vicinity;

[0165] Human body detection: Microwave radar sensor (24GHz) detects human approach within a 2-meter range.

[0166] Based on the aforementioned requirements, the activation logic is as follows:

[0167] When the rescuer enters a 2-meter range, the microwave radar is triggered → the screen brightens and the voice says "Please open the box and take out the AED";

[0168] When the rescuer opens the box (detected by the Hall sensor), the screen switches to the "Retrieve AED" animation;

[0169] When the AED is removed (detected by the weight sensor), the screen switches to the "electrode attachment" animation.

[0170] The on-screen animation shows the operation steps as follows:

[0171] Step 1: Open the box and retrieve the AED;

[0172] Animation: Tap the button on the box → The door opens → The AED is retrieved with both hands;

[0173] Voice prompt: "Please press the green button to open the case door and take out the AED device with both hands."

[0174] Duration: 10 seconds. You can click "Next" to skip.

[0175] Step 2: Power on

[0176] Animation: Press the AED power button with your finger → Screen lights up → Self-test passes;

[0177] Voice prompt: "Please press the green power button on the AED and wait for the self-test to complete."

[0178] Duration: 15 seconds (synchronized with the actual self-test time of the AED).

[0179] Step 3: Attach electrode pads

[0180] Animation: Human body outline, flashing above the right chest → flashing below the left chest;

[0181] Voice prompt: "Please tear open the electrode packaging and apply it to the upper right chest and lower left chest as shown in the diagram."

[0182] Duration: 20 seconds.

[0183] Step 4: Analyze heart rhythm

[0184] Animation: AED screen displays "Analyzing" → Defibrillation recommended / CPR recommended;

[0185] Voice prompt: "Please ensure everyone leaves the patient's area. The AED is analyzing the heart rhythm."

[0186] Safety Confirmation: The camera inside the enclosure detects people in the vicinity. If anyone approaches, a voice message repeats, "Everyone, please leave the patient's location."

[0187] Step 5: Defibrillation / CPR;

[0188] Defibrillation: Voice prompt "Press the flashing orange button to defibrillate";

[0189] CPR: The voice says, "Please begin chest compressions, following the rhythm of the metronome," while the speaker plays a "beep-beep-beep" beat (110 times / minute).

[0190] The above-mentioned active rescue guidance system was tested.

[0191] Test scenario: A middle school conducted a simulated cardiac arrest emergency drill, with a total of 10 drills.

[0192] Test sample: 50 faculty and students (none of whom had experience using AEDs).

[0193] Test metrics:

[0194] (1) Emergency detection time: the time from the simulated fall to the system issuing an alarm;

[0195] (2) Rescuer response time: The time from when the system issues an alarm to when the rescuer confirms their arrival;

[0196] (3) AED acquisition time: The time from when the rescuer confirms their arrival to when they retrieve the AED;

[0197] (4) First defibrillation time: The total time from the simulated fall to the completion of the first defibrillation;

[0198] (5) Operation accuracy rate: The accuracy rate of the rescuer in completing the AED operation steps.

[0199] Control group: Traditional method (witnesses find the AED on their own, without systematic guidance).

[0200] Experimental group: Using the AED active rescue guidance system of the present invention.

[0201] The test results are as follows:

[0202] index experimental group control group Increase Emergency discovery time 8 seconds 45 seconds -82% Rescuer response time 15 seconds 90 seconds -83% AED acquisition time 65 seconds 180 seconds -64% First defibrillation time 95 seconds 280 seconds -66% Operation accuracy 92% 45% +104% Rescuer psychological stress score 2.8 / 5 4.5 / 5 -38%

[0203] Therefore, based on the above tests, the AED active rescue guidance system significantly shortens the rescue response time, improves the accuracy of operation, reduces the psychological pressure on rescuers, and effectively improves the success rate of rescuing patients with cardiac arrest.

Claims

1. An active rescue guidance system for an external defibrillator, characterized in that: The system includes an emergency detection layer, an intelligent dispatch center, an AR navigation implementation layer for rescuers, and an AED box operation guidance implementation layer. The emergency detection layer includes: an AI posture analysis module for campus surveillance cameras, used to detect the posture of a person falling in the surveillance footage; The intelligent wearable device's heart rate abnormality detection module is used to detect sudden cardiac arrest or severe arrhythmia in the wearer; the mobile terminal's SOS emergency call module is used to receive emergency distress signals triggered by the user; and the voice distress call recognition module is used to recognize preset distress call keywords such as "help" and "someone has fainted" in the environment. The intelligent dispatch center includes a server, a multi-source information fusion verification module, an AED matching module, a rescuer screening module, and a path planning module. The intelligent dispatch center is communicatively connected to the emergency detection layer, receives emergency information from multiple information source collection devices, performs multi-source information fusion verification, and determines a genuine emergency situation when at least two information sources trigger simultaneously or when the confidence level of a single information source exceeds a preset threshold. After determining a genuine emergency situation, the center locates the patient's position and automatically matches the optimal AED device and the optimal rescuer based on the patient's position, AED device position, AED device status, rescuer's position, and capability information. It then plans the optimal rescue path for the optimal rescuer, from their current position to the optimal AED device and then from the optimal AED device to the patient's position. The rescuer AR navigation implementation layer is connected to the intelligent dispatch center to achieve the following: sending an emergency rescue notification to the mobile terminal of the optimal rescuer, the notification including the patient's location, the AED device's location, and the rescue route; and guiding the optimal rescuer to the optimal AED device in real time through AR real-scene navigation or map navigation. The aforementioned AED box operation guidance layer is set in the AED device box and enables the following: when the optimal rescuer arrives at the AED device, the optimal rescuer is guided through touch screen animation and voice synchronously to complete the operation steps of opening the AED box, taking out the device, attaching electrode pads, analyzing heart rhythm, and performing defibrillation; after the optimal rescuer arrives at the patient's location with the AED device, the AED operation is continued to be completed through voice guidance.

2. The active rescue guidance system for external defibrillators according to claim 1, characterized in that: The multi-source information fusion verification module is used to perform fusion analysis and cross-verification on emergency alarms from different information sources to determine whether they are genuine cardiac arrest emergencies. The multi-source information fusion verification includes: time window alignment, which aligns emergency information from different information sources by timestamp, and only determines a genuine emergency when multiple information sources trigger simultaneously within a preset time window; spatial location association, which increases the confidence of the emergency when the locations of multiple information sources are within a preset distance range; and single information source confidence assessment, which calculates the confidence based on the historical accuracy of the information source when only one information source triggers the alarm, and determines a genuine emergency when the confidence exceeds 0.

85.

3. The active rescue guidance system for external defibrillators according to claim 1, characterized in that: The AED matching module is used to comprehensively evaluate and select the most suitable AED device for the current emergency dispatch among multiple available AED devices. This includes calculating the distance from each available AED device to the patient's location, querying the consumable status of each AED device, and querying the device health status of each AED device. Based on the distance, consumables, and health status, and according to the corresponding weighting coefficients, the AED device with the highest comprehensive score is selected as the optimal AED device.

4. The active rescue guidance system for external defibrillators according to claim 1, characterized in that: The rescuer screening module is used to comprehensively evaluate and select the most suitable rescuer from multiple potential rescuers, including: obtaining the location information of all registered users on campus, screening registered users within a preset range of the patient's location, prioritizing registered users with first aid training certificates, prioritizing registered users with short historical response times, excluding registered users who have participated in other emergency rescue missions, and selecting the registered user with the highest comprehensive score as the optimal rescuer.

5. The active rescue guidance system for external defibrillators according to claim 1, characterized in that: The path planning module plans a reasonable path based on the determined location of the rescuer and the matched AED location, according to map data. This includes: a first path from the rescuer's current location to the optimal AED device, taking into account the floor, elevator / staircase availability, and real-time pedestrian density; a second path from the optimal AED device to the patient's location, taking into account the portability of the AED device and the accessibility of the patient's location; and dynamic path updates, automatically replanning the path and pushing it to the rescuer when obstacles are detected on the path.

6. The active rescue guidance system for external defibrillators according to claim 1, characterized in that: Implementation of AR navigation for rescuers: When the rescuer turns on their phone camera, the system overlays directional arrows, distance indicators, and AED device location markers onto the camera screen. Precise indoor positioning via Bluetooth Beacon; Provide voice prompts at key locations such as corners, stairwells, and elevator entrances; It displays the estimated arrival time and remaining distance in real time.

7. The active rescue guidance system for an external defibrillator according to claim 1, characterized in that: The AED box operation guidance implementation layer is as follows: When the AED device is activated, the touch screen inside the box will automatically highlight and play a voice prompt saying "Please open the box and take out the AED" when the rescuer reaches within 2 meters of the device. The unboxing animation guides you through the process. The touchscreen displays an animated demonstration of opening the box and taking out the AED, while the voice prompts you: "Please press the button on the box to open the door and take out the AED device." Electrode placement guidance: After the AED device is powered on, the touch screen displays a human outline diagram, uses flashing arrows to indicate the electrode placement position, and provides voice guidance "Please place the electrode pads as shown in the diagram"; Defibrillation safety confirmation: When the AED analyzes the heart rhythm and recommends defibrillation, the camera in the enclosure detects people in the vicinity. If someone is detected approaching, a voice announcement will say "Everyone move away from the patient" until defibrillation instructions are given after it is confirmed to be safe.

8. The active rescue guidance system for external defibrillators according to claim 1, characterized in that: The campus surveillance camera AI pose analysis module implements: human detection, using the YOLOv8-nano model to detect human targets in the surveillance image; pose estimation, using the MediaPipe Pose model to extract 33 human key points; The fall detection is triggered when the torso tilt angle is greater than 60 degrees, the key point remains still for more than 5 seconds, and the body quickly changes from a standing position to a falling position, thus triggering an emergency signal. Multi-camera collaboration: when a single camera detects a fall, it automatically calls upon adjacent cameras to confirm the viewpoint, improving detection accuracy.

9. The active rescue guidance system for an external defibrillator according to claim 1, characterized in that: The AED box operation guidance implementation layer also includes a CPR auxiliary guidance module, which enables: when the AED analyzes the heart rhythm and recommends CPR, it plays a metronome sound through a speaker to guide the rescuer's compression rhythm; it provides voice guidance on compression depth and rebound requirements, and guides the ratio of compressions to artificial respiration; and it prompts the rescuer to take turns every 2 minutes.

10. The active rescue guidance system for an external defibrillator according to claim 1, characterized in that: The AED box operation guidance implementation layer also includes an emergency coordination module, which enables: automatically dialing 120 emergency number, sending a distress message containing the precise location, pushing emergency notifications to the school clinic and security office, and providing voice guidance for the division of labor when multiple rescuers arrive at the scene.

Images