WITH MULTI-SENSOR FUSION AND CREW SOUND SIGNATURE SUBSCRIPTION. DEVICE AND METHODS FOR DETECTING LIFE UNDER THE RUBBLE
Patent Information
- Authority / Receiving Office
- TR · TR
- Patent Type
- Applications
- Current Assignee / Owner
- İSA URHANOĞLU
- Filing Date
- 2026-05-22
- Publication Date
- 2026-06-22
Smart Images
Figure 00000012_0000 
Figure 00000013_0000 
Figure 00000014_0000
Abstract
Description
1 TARIFF WITH MULTI-SENSOR FUSION AND CREW SOUND SIGNATURE SUBSCRIPTION. DEVICE AND METHODS FOR DETECTING LIFE UNDER THE RUBBLE Technical Area 5 The invention allows for the recovery of debris from ruins caused by earthquakes, collapses, explosions, and similar disasters. portable search and rescue devices used in locating remaining survivors and this It relates to how the devices operate. The invention specifically includes a multi-channel microphone array, a seismic geophone, a thermal infrared camera, and a gas a sensor group that receives simultaneous data from at least four modalities, and processes that data from 10 by combining each with a Bayesian probability fusion model on the edge AI processor generating a map of the probability of a live person for the search point and the search team's own voice- A submerged structure that eliminates false positives by obtaining a vibration signature from the subscriber before the study. It is related to the life detection device and how this device works. State of the Art 15 In known technology, single-module search is used to locate survivors trapped under rubble. These devices are used. The majority of these devices are based solely on acoustic principles. It is sustained and transmitted through microphones at the ends of passive probes placed in the wreckage. It attempts to capture human-generated sounds; the operator listens through the headset. is making the decision. In this approach, environmental noise, wind, generator, in-team communication and 20 Factors such as building material noise and subsidence noise lead to a high rate of false positives; Operator fatigue reduces decision quality. The known technique also includes search configurations based solely on thermal infrared imaging. These structures are currently in use. Temperature anomalies on the debris surface are being monitored in these structures; However, in cases where the survivor is found deep within the rubble, a distance of 25 kilometers is created between the surface and the person. Thermal insulation prevents signal detection. Also, fire, hot water pipe. Leakage or the presence of animals leads to the misconception that the thermal signal is of human origin. Currently available technology also utilizes ground-penetrating radar (GPR) and seismic probe devices. These devices are capable of detecting survivors who have responded to mechanical impact. However, when these structures are used alone, they result in a situation where the survivor cannot deliver a blow, and his consciousness is impaired to 30. It becomes insufficient in cases where breathing is impaired or weakened. 2 In the known technique, multiple modalities are physically side-by-side in a single device. Although there are structures in which it is placed; in these structures, each modality They are viewed separately on separate screens; the operator has to mentally combine the data. remains. Among the modalities in question, the Bayesian probability fusion model with data 5 a system that combines and produces a single probability value for a living human for each scan point. An edge AI-powered architecture does not currently exist in known technology. The known technique also includes the search team's own voice, foot vibrations, and equipment. Its noise can be confused with surviving source signals by acoustic and seismic sensors; This situation increases false positives. The signals originating from this team are discussed in study 10. by recording beforehand and creating an audio and vibration signature for each team member, a system that eliminates false positives by extracting these signatures from the signal during real-time scanning The subscriber memory structure is not present in the known technology. Due to the aforementioned drawbacks, there is a need to make improvements in the relevant technical field. It has been heard. 15 Purpose of the Invention The invention is intended to utilize a multi-channel microphone array, seismic geophone, thermal infrared camera, and gas. receiving real-time data from a group of sensors and processing that data using Bayesian probability fusion. by combining it with the model, it produces a single probability value of a live human for each scan point. The goal is to develop a life-detecting device under the rubble. 20 Another purpose of the invention is to capture the speech, foot vibrations, and voice of each member of the search team. recording the sound-vibration signature resulting from equipment noise before starting work, and a system that eliminates false positives by extracting these signatures from the signal during real-time scanning The team's goal is to develop a rubble-under-life detection device with a subscriber memory containing voice signatures. Another aim of the invention is to analyze human-derived breath and heartbeat frequency bands (0.1-3 Hz 25 Adaptive frequencies that selectively amplify vibration and acoustic frequencies (50-500 Hz) and suppress other frequencies. The goal is to develop a debris-based life detection device that utilizes band-pass filtering. Another purpose of the invention is to transmit signals via beamforming through a microphone array. It estimates the orientation angle of the source in three dimensions and uses this orientation information with thermal infrared. A survivor under the rubble is marked on a scanning map by mapping it to the camera's field of view. 30 The goal is to present a detection device. 3 Another purpose of the invention is gas sensors at the tip of a telescopic sampling probe. through which carbon dioxide, oxygen and volatile organic compound levels from the interior of the wreckage measuring rising carbon dioxide and falling oxygen levels in a probability model of a living human The goal is to develop a rubble-under-life detection device that allows for input. 5 Another purpose of the invention is to detect depths that the operator cannot see with the naked eye. the points along with previous scan markings on the augmented reality glasses The goal is to develop a debris-buried life-detecting device with an interface module that displays this information. Another purpose of the invention is to enable the device to operate and to process the sensor data. By combining it with a Bayesian probability fusion model, the team voice signature is obtained by subscribing to 10. The goal is to present a method that includes the steps for filtering. Explanation of the Figures This document was prepared to better understand the invention, a life-detection device used in rubble. The figures are explained below. Figure 1: Perspective view of the device subject to the invention, including the carrying handle, screen and sensor head. 15 appearance. Figure 2: Exploded view showing the main components of the device separately. Figure 3: Cross-section of the microphone array and seismic geophone placement inside the sensor head. appearance. Figure 4: Detail of the gas sensor assembly at the tip of the telescopic sampling probe. 20 appearance. Figure 5: Flowchart of the method described in the invention. Explanation of References in Figures To better understand the elements of the device that is the subject of the invention, each element is described separately. They are numbered, and the meaning of each number is explained below. 25 1. Device housing 2. Sensor head 3. Microphone array 4 4. Seismic geophone 5. Thermal infrared camera 6. Telescopic sample collection probe 7. Sampling pump 5 8. Carbon dioxide sensor 9. Oxygen sensor 10. Volatile organic compound sensor 11. Edge AI processor 12. Bayesian fusion processing unit 10 13. Team voice signature subscriber memory 14. Display screen 15. Augmented reality glasses interface 16. Unit of inertia measurement 17. GNSS module 15 18. Telemetry and team communication module 19. Lithium-ion battery 20. Carrying handle Description of the Invention The subject of the invention is a life detection device under rubble (Figure 1, Figure 2); a device housing (1), word 20 The subject is a sensor head (2) mounted on the front of the device housing (1), sensor a microphone array (3) consisting of at least eight channels placed inside the header (2), sensor a low-frequency seismic geophone (4) mounted on the base of the sensor head (2), (2) a thermal infrared camera (5) mounted in front of the device housing (1) a telescopic sampling probe (6) and a sampling pump (7) attached to this probe, 25 A carbon dioxide sensor (8) located at the tip of the sampling probe (6) in question, oxygen sensor (9) and a volatile organic compound sensor (10), from the said sensors a Bayesian fusion compute unit containing an edge AI processor (11) that processes data (12), a team voice signature subscriber memory (13), a display screen (14) and optionally an augmented reality glasses interface (15), an inertial measurement unit (16), a GNSS module (17), a telemetry and team communication module (18), a lithium-ion battery (19) and a 5 Includes a carrying handle (20). Device housing (1); made of impact-resistant polymer or magnesium alloy material. manufactured, IP67 rated dust and temporary water immersion protection, carrying handle (20) It is an ergonomic case with attachment points for the shoulder strap. The sensor head (2) is mounted on the front of the device housing (1) via a pivot with a rotating joint. connected, directed in the direction targeted by the operator, containing a microphone array (3), It is a headset containing a seismic geophone (4) and a thermal infrared camera (5). Microphone array (3); distributed in circular or spherical topology within sensor head (2) It is an array consisting of at least eight MEMS microphones. Microphone array (3); beamforming The beamforming algorithm determines the direction angle of an acoustic signal source in three dimensions. It estimates that the output signals of the said sequence are sent to the edge artificial intelligence processor (11) It is transmitted and highlights acoustic components originating from breath and heartbeat in the 50-500 Hz range. It is passed through an adaptive band-pass filter that removes it. Seismic geophone (4); mounted on the base of the sensor head (2), in contact with the debris surface. in this case, low-frequency vibrations in the range of 0.1-3 Hz (physical movement, wall tapping, 20 It is a geophone element that detects (finger crackling). The geophone in question (4) is thermal insulation. due to the fact that the thermal infrared camera (5) cannot detect deep debris layers This allows the remaining part of the body to be detected using a gravity-vibration modulation. Thermal infrared camera (5); operating in the wavelength range of 8-14 micrometers, low A high-resolution but highly thermally sensitive long-wave infrared image 25 It is a sensor. Thermal infrared camera (5); 0.3 according to the average temperature on the surface of the debris. marking anomalies above degrees Celsius on the map and microphone array (3) It is placed in the same coordinate plane as the ray generation output. Telescopic sample collection probe (6); extending at least two meters from the device housing (1), It is a thin-diameter tube carrying gas sensors at its end. Sampling pump (7); the 30 in question By creating a continuous vacuum along the inner channel of the probe (6), a gas sample is taken from the interior of the debris. taking this example from carbon dioxide sensor (8), oxygen sensor (9) and volatile organic compound 6 It measures by passing the sensor (10) over the breath of the surviving person. As a result, carbon dioxide levels rise and oxygen levels decrease in the enclosed space where the debris is located. its level is decreasing; this trend is evidence of the Bayesian fusion processing unit (12). It is entered as a variable. 5 Edge AI processor (11); low-power, housed inside the device housing (1) However, it is a system-on-a-chip (SoC) component with high computational intensity. The subject processor (11); beamforming and adaptive band of the signal from the microphone array (3) It performs the filtering steps that pass the vibration from the seismic geophone (4); classifying with breath / heart rhythm pattern; image from thermal infrared camera (5) 10 converting gas sensor outputs into anomaly maps and trend variables. It translates. Bayesian fusion processing unit (12); modality from edge AI processor (11) based evidence variables, to a single live human probability value for each screening point It is a probabilistic graphical model that transforms. Bayesian fusion processing unit (12); every 15 Dynamically updated reliability based on modality and ambient noise level. by assigning coefficients; this reduces the weight of the acoustic modality in a windy environment. The emphasis on seismic and gas modalities is being increased. Team voice signature subscriber memory (13); at the start of the search operation, each member of the team thirty-second 20-minute segments of speech sounds, foot tremors, and the noise produced by the equipment he is carrying. by recording the data and generating a separate acoustic and seismic signature vector for each member, and these vectors It is a memory unit that stores signatures. These signatures are recorded during real-time scanning of the microphone. The surviving source was extracted from the signals from the array (3) and the seismic geophone (4). Signals are prioritized. This structure allows the team to communicate their own speech or gait effectively. This helps prevent the risk of misinterpretation as a residual signal. 25 Display screen (14); Bayesian, mounted on the top surface of the device housing (1). Color-coded probability map of living human produced by the fusion processing unit (12), A touch screen that presents the operator with the beam generation angle of the microphone array (3). It is a screen. Augmented reality glasses interface (15); a 30 that can be optionally connected to the device wirelessly transmits the output of the Bayesian fusion processing unit (12) and the previous to the eyeglass device. It is an interface that transmits the markings of the scanning points. The said eyeglass interface (15) 7 Thanks to this, the operator can overlay the detection points on the wreckage within the real-world field of view. They can view the data superimposed; they can compare past and current scans. Inertial unit (16); continuously measures the orientation and movement of the device, It is a sensor that includes a gyroscope, accelerometer and magnetometer. The unit in question (16) is a thermal 5 Scanning direction of the infrared camera (5) and microphone array (3) from GNSS module (17) Combining incoming global positioning information with geographic coordinate labels for detection points. is appointing. Telemetry and team communication module (18); a radio module operating in mesh topology and sharing the detection point data with other similar devices in the field and team 10 It provides internal voice communication. Lithium ion battery (19); capacity to operate the device at full load for at least four hours It has a removable battery that can be replaced in the field with a spare. The operating principle of the device described in the invention (Figure 5) includes the following steps: (i) Thirty seconds of speech and movement recording of each member of the search team will be recorded via microphone 15 Team voice signature is recorded in subscriber memory (13) via array (3) and seismic geophone (4) the saving step, (ii) Deployment of the device to the debris site and target scanning of the sensor head (2) the step of directing it in that direction, (iii) From the microphone array (3), seismic geophone (4), thermal infrared camera (5) and gas 20 Simultaneous data collection step from sensor group (8, 9, 10), (iv) From the collected acoustic and seismic signals, the team's voice signature is stored in the subscriber memory (13) Signatures are extracted by the edge AI processor (11) and team-sourced signals are extracted. filtering step, (v) The evidence variables of each modality are entered into the Bayesian fusion processing unit (12) 25 The step of generating a single probability value for a live human at the screening point, (vi) The said probability value is displayed on the display screen on a (14) color-coded map marking and optionally via the augmented reality glasses interface (15) The step of transmitting the information to the glasses. 8 Thanks to the device and method described in the invention, in single-modulus search configurations The high false positive rate encountered is significantly reduced, thanks to team-related factors. Errors from the signals are automatically eliminated via the subscriber memory, operator By eliminating the burden of mentally combining multiple screens, a single possibility of 5 is available. being able to focus on the map and thus identify survivors within the golden hours. The probability is being increased. How the invention can be applied to industry. The invention is a device for detecting life under rubble; it is used by fire departments, AFAD (Disaster and Emergency Management Presidency of Turkey), civil defense, and similar search and rescue organizations. They are suitable for direct use by rescue units in searching debris after a disaster. 10 All structural elements of the device in question are manufactured using standard production techniques, namely injection molding. molding, printed circuit board manufacturing, MEMS sensor placement, soldering, and final assembly. It is capable of being manufactured industrially using [method / technology]. The invention concerns a method / device. embedded software that is directly run by the edge AI processor on it It is implemented through. 15
Claims
9 REQUESTS 1. It is a device used to locate survivors trapped under rubble after a disaster. Its feature is a device housing (1), microphone array (3) attached to the said device housing (1), a sensor head (2) containing a seismic geophone (4) and a thermal infrared camera (5); device a telescopic sampling probe (6) extending from its housing (1) and located at the end of the probe 5 field carbon dioxide sensor (8), oxygen sensor (9) and volatile organic compound sensor (10); word The subject is an edge artificial intelligence processor (11) that processes data from sensors; edge artificial By combining the modality-based evidence variables from the intelligence processor (11), each scan A Bayesian fusion processing unit (12) that produces a single probability value of a living human for the point; It stores the speech and movement signatures of each member of the search team and provides real-time 10 a team that eliminates team-generated false positives by extracting these signatures from the signal during scanning the probability map from the voice signature subscriber memory (13) and the Bayesian fusion processing unit (12) It includes a display screen (14) that is presented to the operator.
2. It is a device according to claim 1, and its feature is that the microphone array (3) has a circular or spherical topology 15 It consists of at least eight MEMS microphones distributed with a beamforming algorithm. It is the ability to estimate the direction and angle of the signal source in three dimensions.
3. According to claim 1, it is a device whose characteristic is that of a seismic geophone (4), low in the range of 0.1-3 Hz. It is a geophone element that detects frequency vibrations. 20 4. According to claim 1, it is a device whose characteristic is a thermal infrared camera (5), 8-14 micrometers. operating in the wavelength range and at temperatures above 0.3 degrees Celsius on the debris surface. It is a long-wave infrared imaging sensor that maps anomalies.
5. It is a device according to claim 1, and its feature is that it has a telescopic sampling probe (6) at its tip. It includes a sampling pump (7) that takes a gas sample by creating a vacuum.
6. It is a device according to claim 1, and its feature is that it has a Bayesian fusion processing unit (12) for each modality. assigning confidence factors that are dynamically updated depending on the noise level of the environment It is a probabilistic graphical model.
7. According to claim 1, it is a device whose feature is; team voice signature subscriber memory (13), each team 5 Separate acoustics were produced from an initial thirty-second recording of a member's speech and movement. and it is a memory unit that stores seismic signature vectors.
8. It is a device according to claim 1, and its characteristic is that it connects wirelessly to a Bayesian fusion device. 10 showing the output of the processing unit (12) superimposed in the real world field of view It includes an augmented reality glasses interface (15).
9. According to claim 1, it is a device whose characteristic is an inertial unit that measures the orientation of the device. (16) and includes a GNSS module (17) that provides geographic location information and to the detection points It is configured to assign a geographic coordinate label. 15 10. A device according to Claim 1, characterized by its operation in a mesh topology, unlike other devices of the same type in the field. A telemetry and intra-team communication system that enables the sharing of point-of-detection data with devices. It includes module (18).
11. The operating method of the device in Claim 1 is characterized by including the following steps: (a) Speech and movement recording of each member of the search team microphone array (3) and seismic (b) the team voice signature is recorded in the subscriber memory (13) by taking it via geophone (4); by deploying to the wreck site, from the microphone array (3), seismic geophone (4), thermal Simultaneous data collection from the infrared camera (5) and the gas sensor group (8, 9, 10); (c) 25 The team sound signature from the collected acoustic and seismic signals is the edge of the signatures in the subscriber memory (13). Filtering of team-sourced signals by extracting them by the artificial intelligence processor (11); (d) each Scanning by entering the evidence variables of the modality into the Bayesian fusion processing unit (12). 11 (e) generating a single probability value for a living human for the point; the probability value in question Marking on a color-coded map (14) on the display screen.
12. It is a method according to claim 11, and its feature is; in step (c), the team voice signature in the subscriber memory (13) signatures are filtered from the signal with an adaptive filter coefficient during real-time scanning. It is removal.
13. It is a method according to claim 11, and its characteristic is; the Bayesian fusion processing unit in step (d). (12), the reliability of each modality is dynamically updated depending on the noise level of the environment. It is the assignment of coefficients. 10