Cooperative vital sign monitoring system based on fixed radar and mobile unmanned aerial vehicle

Abstract

According to the fixed radar and unmanned aerial vehicle cooperative vital sign monitoring system provided by the invention, large-range basic monitoring is realized through a millimeter wave radar array, and the specific unmanned aerial vehicle executes accurate re-checking after receiving an abnormal signal. The system adopts a multi-stage response mechanism: when the fixed radar detects abnormal signs, the unmanned aerial vehicle is dispatched to lock a target through millimeter wave echo feature matching, identity binding is performed by using a binocular camera, body temperature data is supplemented in combination with an infrared sensor, and finally an early warning decision is generated through a multi-modal fusion algorithm. The hardware level defines the lowest configuration requirements (including a millimeter wave radar module, a visual identification unit, an infrared temperature measurement assembly and an obstacle avoidance system) of the unmanned aerial vehicle, and the software level designs a cross-device data space-time alignment model and a dynamic task scheduling protocol. Application scenes cover night emergency monitoring of nursing homes, health monitoring of gymnasium groups and non-contact patrol of hospital isolation wards, and the problems that in a traditional scheme, tracking of moving targets is lost, and identity information is unhooked are effectively solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent monitoring, and specifically relates to a vital sign monitoring system in which a fixed radar array and a specific unmanned aerial vehicle cooperate to work, and is applicable to multimodal health monitoring and early warning in large-space fixed-population places such as nursing homes, gymnasiums, and hospitals. Background Art

[0002] Existing fixed radar systems have monitoring blind spots, while pure unmanned aerial vehicle solutions are limited by endurance and data continuity. Patent CN118633918A, a regional human health monitoring method based on 4D millimeter-wave radar, proposes the deployment of multiple radar arrays, but does not solve the problems of separately tracking multiple targets and identity binding; CN118749907A, a non-contact dynamic balance ability evaluation method and device based on dual radars, does not give the frequencies of the dual radars, and at the same time, the functions of its radars cannot monitor vital signs such as heart rate and respiration. Currently, there is an urgent need for a collaborative system that integrates the global perception of fixed nodes and the fine detection of dynamic vital signs of unmanned aerial vehicles. Summary of the Invention

[0003] The present invention proposes a vital sign monitoring system in which a fixed radar and an unmanned aerial vehicle cooperate. A large-range basic monitoring is realized through a millimeter-wave radar array, and a specific unmanned aerial vehicle performs precise review after receiving an abnormal signal. The system architecture is shown in Figure 1 , and mainly includes a fixed monitoring layer composed of a 24GHz horizontal radar array, a 60GHz vertical radar array, and an edge computing center, and a mobile unmanned aerial vehicle layer composed of a millimeter-wave radar module, a binocular camera, and an infrared sensor.

[0004] The fixed monitoring layer constructs a "foundation network" for global perception: The 24GHz horizontal radar array, whose function is to use millimeter waves with longer wavelengths (strong penetration) to continuously monitor the basic vital signs (breathing, heart rate) of people in a large range. Deployment method: Horizontally arranged on the wall or ceiling, covering more than 80% of the area (similar to the grid of "surveillance cameras"). Technical characteristics: Capturing micro-motion signals such as chest undulations through the Doppler effect, and can penetrate obstacles such as clothing and thin walls; The function of the 60GHz vertical radar array: High-precision three-dimensional positioning (the pitch angle resolution reaches 2°), accurately distinguishing overlapping targets. Deployment: Installed on the ceiling, tilted downward by 30° - 60°, forming a three-dimensional monitoring network. Technical characteristics: Short wavelength (5mm) provides higher spatial resolution, and can detect fine movements such as finger micro-motions; The role of the edge computing center: Real-time processing of radar data, running a preliminary anomaly detection algorithm, and coordinating the response of the unmanned aerial vehicle. Example: When it is detected that the breathing frequency in a certain area drops abnormally (such as <8 times / minute), it is immediately marked as a "suspected risk target".

[0005] The mobile drone layer serves as the "air special service" for precise verification. The drone fuselage is equipped with a millimeter-wave radar module. Its task is to quickly fly to the target area after receiving the abnormal coordinates sent by the fixed layer, and lock the specific individual through millimeter-wave echo feature matching. The key technology is to compare the micro-motion signal spectra (such as heartbeat waveforms) of the fixed radar and the drone radar to confirm whether it is the same target. The task of the binocular camera is to perform face recognition or gait feature extraction within a distance of 2-5 meters and associate with the health record. Its advantage is that it is bound to the radar data, and even if wearing a mask, the identity can be recognized through gait (similar to "face + gait" dual authentication). The task of the infrared sensor is to measure the surface temperature of the target (accuracy ±0.3°C) and supplement abnormal data such as fever.

[0006] The cooperation logic between the fixed radar and the drone is as follows: when the radar detects shortness of breath, the infrared data can distinguish whether it is a normal reaction after exercise or caused by fever.

[0007] The detailed working process of this system is as follows ( Figure 2 ). Phase 1: All-weather monitoring by the fixed layer ( Figure 3 ). 1) Each radar updates data every 10 seconds, and the edge computing center eliminates environmental noise through the Kalman filtering algorithm. Abnormal judgment rules (example): apnea > 30 seconds, sudden heart rate drop of more than 40%, detection of a fall (body position height drops suddenly from 1.5m to 0.5m); Phase 2: Drone dynamic response ( Figure 4 ), the technical details include: 1) Millimeter-wave fine scanning: The drone hovers 3-5 meters away from the target, scans with a narrow beam (3°), and confirms whether the target is a living body through the echo phase difference (excluding false alarms of furniture), 2) Identity binding: Bind the face features recognized visually and the radar physical sign data (such as heart rate curve) to the same ID to ensure data continuity, 3) Hierarchical early warning: Yellow early warning (text message notification): Slight body temperature abnormality (such as 37.5°C), Red early warning (sound and light alarm + emergency call): Detection of cardiac arrest or serious fall.

[0008] The advantages of the cooperation between the fixed radar and the drone in this system are reflected in three aspects: coverage complementarity, data fusion strategy, and emergency response mechanism.

[0009] Coverage complementarity: The fixed radar can cover high-frequency areas (such as beds, corridors) 7×24 hours, and the drone covers temporary blind spots as needed (such as bathrooms, corners). The battery life is designed to fly for 30 minutes after charging for 10 minutes.

[0010] Spatial-temporal alignment algorithm in the data fusion strategy: Solve the timestamp deviation between devices (error < 50 ms) and coordinate system differences (corrected by real-time SLAM mapping). Confidence weighting: The weight of radar vital sign data accounts for 70%, visual identity accounts for 20%, and infrared temperature accounts for 10% (priority scheduling is available: when multiple anomalies are detected simultaneously, calculate the urgency according to the formula: Priority = Degree of deviation of vital signs / Distance from the target to the dangerous area × Emergency coefficient. For example: The priority of a fallen patient near the window > A person with abnormal breathing in the center of the hall).

[0011] Through the collaboration of fixed radar global perception and precise drone verification, this system realizes the combination of "area monitoring" and "point diagnosis". Just like "electronic eyes + traffic police patrol" in the traffic system - the fixed radar continuously monitors the whole field like electronic eyes. After detecting an anomaly, the drone quickly arrives at the scene for processing like a traffic police officer. It not only ensures the continuity of monitoring but also significantly reduces the false alarm rate through cross-verification of multi-modal data, which is an innovative solution in the field of group health monitoring. Description of the Drawings

[0012] Figure 1 : System architecture diagram.

[0013] Figure 2 : Workflow diagram.

[0014] Figure 3 : All-weather monitoring of the fixed layer.

[0015] Figure 4 : Dynamic response of the drone.

[0016] Figure 5 : Night care in nursing homes.

[0017] Figure 6 : Monitoring of sudden myocardial infarction in gymnasiums.

[0018] Figure 7 : Monitoring of isolation wards in hospitals. Detailed Implementation Modes

[0019] Example 1: Night care in nursing homes ( Figure 5 ). Problem: An elderly person suddenly fainted when getting up to go to the toilet at night. System response: 1) The fixed radar detected a sudden stillness (no breathing signal) in the toilet; 2) The drone automatically took off from the charging pile, and the millimeter wave confirmed that the target had fallen to the ground; 3) The binocular camera identified the identity of the elderly person (comparing the pattern features of the pajamas); 4) Infrared temperature measurement ruled out the possibility of hypothermia; 5) Triggered a red alarm and opened the toilet door lock.

[0020] Example 2: Marathon event in the gymnasium ( Figure 6). Problem: A contestant suddenly has a myocardial infarction and collapses to the ground. System response: 1) The fixed radar detects an abnormal body position (a sudden drop in height) in the crowd; 2) The drone penetrates through the gaps in the crowd and uses millimeter waves to detect cardiac arrest; 3) The vision system identifies the contestant's bib number and retrieves the pre-competition health record; 4) Automatically navigates the nearest AED device to the scene.

[0021] Example 3: Hospital isolation ward monitoring ( Figure 7 ). Problem: Confirm whether the patient has a fever. 1) The vertical radar detects a gesture for help; 2) The millimeter wave monitors the chest undulation; 3) The infrared detects whether there is a fever; 4) The vision confirms the infusion completion status.

Claims

1. A fixed radar and drone coordinated vital signs monitoring system, characterized in that It includes a fixed millimeter-wave radar array, a drone equipped with millimeter-wave radar / binocular camera / infrared sensor, and a multimodal data fusion platform to achieve a three-level response mechanism of abnormal target locking-identity confirmation-depth detection.

2. The system according to claim 1, characterized in that The UAV receives the target space coordinates and preliminary vital sign data provided by the fixed radar, and realizes target relay tracking through millimeter wave echo feature comparison.

3. The system according to claim 1, characterized in that A spatiotemporal alignment algorithm is used to bind the visual identity information and radar vital sign data collected by the drone to the same health record.

4. The system according to claim 1, characterized in that A dynamic task priority model was established, and the calculation formula was: P = ΔV / Vbaseline × 1 / Dsafety (ΔV is the sign deviation value, and D is the distance to the dangerous area).

Images