WO2020139206A1

WO2020139206A1 - System and method of live human detection for disaster emergency search and rescue

Info

Publication number: WO2020139206A1
Application number: PCT/TR2018/050878
Authority: WO
Inventors: Dr. NAIL CADALLI; Reyhan ERGUN; Ugur TOPAY
Original assignee: Signalton Teknoloji Ltd. Sti.
Priority date: 2018-12-25
Filing date: 2018-12-25
Publication date: 2020-07-02

Abstract

This invention relates to a live human detection method and related sensor signal processing system for use in search and rescue operations at the time of a natural disaster including earthquake, landslide, avalanche, and so forth.

Description

SYSTEM AND METHOD OF LI VE HUMAN DETECTI ON FOR DI SASTER EMERGENCY

SEARCH AND RESCUE

The Technical Field Of The I nvention

Prior Art About The I nvention ( Previous Technique)

For certain disasters such as earthquakes, which happen so suddenly and without prior warning, it is critical to collect information about the human occupancy in a residence within a short period before the disaster hits. That way search and rescue teams can organize valueable resources at the right locations in a disaster region avoiding searching for nothing.

After the disaster, for instance in the case of an earthquake, current methods try to reach far from outside of the building to deep under the debris searching for live people. One primitive method is to call for live people and listen to replies. Some technical methods (Di Zhang et al, “Evaluation of a Sensor System for Detecting Humans Trapped under Rubble: A Pilot Study”, Sensors 18, no. 3, p. 852, 2018.) incorporate microphones and signal analysis for the listening phase. There are also various other methods and products that use doppler radar, thermal cameras, search robots and gas sensors. However, all of those techniques are utilized after the disaster and from outside a collapsed building.

Aims Of The I nvention and a Brief Explanation

This invention relates to a system and method of live human detection for disaster emergency search and rescue. This invention proposes a solution to the difficulty where all the search and rescue operations are focused to the time interval after a natural disaster occurs. With the current invention, a residential detector is placed at each residence so that the detector is close to the human subjects, which increases the likelihood of live human detection after the disaster. The device also collects prior occupancy information and feeds it to the information system that is part of the search and rescue operations.

This invention also introduces a specific combination of various signal processing techniques and algorithms to increase the rate of detection with efficient and smart processing of sensor data, signals and information.

Another aspect of the invention, wherein a residential detector or a network of such devices is positioned at a house; at an office or a residence within a building or an apartment; or at each floor of an apartment; outside on a street pole; on an outside wall of a building; or even at a workplace such as a coal mine.

Another aspect of the invention, wherein the sensor module of the residential detector device comprises at least one acoustic sensor (microphone) (102) , at least one vibration sensor (103) , at least one inertial sensor group (accelerometer, gyro, magnetometer) (104) , at least one global navigation satellite system (GNSS) based location sensor (GPS, GLONASS, etc.) (105) and at least one image/video sensor (camera) (106) .

Another aspect of the invention, wherein the microprocessor can be a digital signal processor (DSP) , a general purpose microprocessor, or a group of microprocessors from various sorts.

Another aspect of the invention, wherein a RF antenna(1 14) supports Bluetooth, Wi Fi, cellular (GSM, 3G, 4G) , Zigbee or similar wireless RF communication modalities.

Another aspect of the invention, wherein the RF subsystem (1 13) and (RF) antenna 1 14 can simultaneously support multiple of the RF channels such as Bluetooth, Wi-Fi, cellular (GSM, 3G, 4G) , Zigbee or similar wireless RF communication modalities.

Another aspect of the invention, wherein RSDN nodes communicate with each other and the gateway through one of the supported RF channels while the RDSN gateway might communicate to the mobile rescue team through the same or a different supported RF channel. The Descriptions Of The Figures Explaining The I nvention

The figures are used to better explain the developed system and the method of live human detection for disaster emergency search and rescue with this invention and their descriptions are as follows:

Figure 1 : System components of the residential detector.

Figure 2: Communication links between various components of the method.

Figure 3: Signal and sensor data processing of the detector device for different phases of operation.

Figure 4: Residential detector sensor network and its connection to an outer point of communication.

The Detailed Explanation Of The I nvention

To better explain the system and method of live human detection for disaster emergency search and rescue with this invention, the details are presented below.

The current invention introduces a method of emergency search and rescue information system as well as a sensor signal processing system . The information system tracks the residential occupancy information and focuses on the last minutes before the disaster instant to decide if there are people to search for in that residence or building.

The basic process flow of the method includes detection of live human signals and related signals (sound, vibration, image, etc.) through the use of signal processing and detection algorithms before and after the disaster instant and provide search and rescue teams with information indicating the possibility of people being alive under rubble of buildings hit by such disasters. The preferred embodiment of the system includes a residential detector that is positioned at a house or a residence within a building such as an apartment, or an office. More than one detector can be used in one residence to increase the detection rate. The residential detector is registered with the rescue center at the time of the installation so that the exact address, location and usual occupancy information is known before a disaster hits. I f the residence is mobile, the method is capable of keeping track of the location of the residence. The residence can, in general, be a workplace such as a coal mine as an extreme example. Residential detectors can be also placed at each floor of an apartment, outside on a street pole or on an outside wall of a building. The detection process of the method includes a short time frame (e.g. half an hour) before the disaster instant and continuing time frame after the disaster instant. The time frame before the disaster instant provides information about the occupancy, preferably the number of people, inside a residence. This information can provide valuable insight for the rescue teams to plan and organize the search and rescue efforts so that available resources can be used efficiently at the right locations throughout the disaster area. The time frame after the disaster instant starts from the hit instant of the disaster and continues until the search and rescue team’s call of no further search at a specific location (building) or until the power of the system runs off. This time frame provides information about the probability of live humans being in the rubble.

The system components of the residential detector are shown in Figure 1 . The detector system (101 ) makes use of analog or digital signals coming from on-board or connected sensors. The preferred embodiment of this invention uses acoustic sensors (microphones) (102) , vibration sensors (103) , inertial sensors (accelerometer, gyro, magnetometer) (104) , global navigation satellite system (GNSS) based location sensors (GPS, GLONASS, etc.) (105) and image/video sensors (cameras) (106) . The camera can be optical, thermal, infrared, or of another imaging modality. The signals coming from the sensors are pre-processed by the peripheral components (107) of the electronic hardware of the system, are digitized at proper sampling rates and input to a microprocessor (108) for algorithmic processing. The microprocessor can be a digital signal processor (DSP) , a general purpose microprocessor, or a group of microprocessors from various sorts. The microprocessor is capable of processing signals coming from the sensor in real-time or near real-time rates so that the proper information retrieval from the sensor data can be accommodated. The microprocessor can execute software tasks for system control (109) , signal processing and analysis algorithms (1 10) , data fusion (1 1 1 ) and data communication (1 12) .

The system is linked to the rescue center through bi-directional wireless radio frequency (RF) channels such as Wi-Fi, cellular (GSM, 3G, 4G, etc.) , Zigbee, etc. so that the rescue center can monitor and control the device and receive/send information from/to the device. The RF communication subsystem (1 13) connects to the RF antenna (1 14) . The rescue center can be at a stationary location in the area, city or region. I t can also be mobile with teams searching the area at the time of a disaster.

The residential detector can also connect through a wired connection to an outer point of connection by the wired communication subsystem (1 15) . The residential detector system includes a power subsystem (1 16) that manages the electrical power coming from mains electricity, battery or a solar recharge unit.

Algorithms that are executed on the microprocessor of the system for the signal processing of captured sensor data may include signal feature extraction, pre- or post processing of such features, sound event detection for human sounds, machine learning and pattern recognition for sound source and audio scene analysis, speech recognition and keyword recognition for human speech detection and understanding, vibration analysis for impact noise detection, inertial data analysis for seismic disaster (e.g. earthquake) detection and after-disaster impact detection, image and video capturing and communication, location extraction from navigational satellite systems such as GPS.

The microprocessor can also execute higher logic algorithms for further processing of data coming from various sensors and results coming from various algorithms being executed. Such higher logic algorithms can make decisions using such fused data and information retrieved from outside the system . Outside-retrieved data may include for instance the exact time of disaster, time information, commands from the rescue teams, etc.

Such higher logic decision making can take place in part or entirely in the server computers of the rescue center as well. Such server computers can be located physically in the rescue center, can be portable with the mobile rescue teams, or reside in the cloud (I nternet) . The entire processing thus can take place partly on the device that is the subject of this invention, partly on the mobile rescue teams portable server computer, and partly on the search and rescue command and control center computer (or the I nternet cloud computer) .

As depicted in Figure 2, as a preferred embodiment of this invention, one or more of the residential detectors (201 ) can communicate over RF channels to the server computer

(202) of a mobile rescue team which consists of subsystems for data fusion and processing

(203) , user interface (204) , RF communication (205) and RF antenna (206) . One or several mobile rescue teams may connect to a search and rescue command and control center

(207) . Such a command and control center keeps a database for residences and residents

(208) , executes software tasks such as data fusion and information processing (209) and coordinates search and rescue operations (210) .

The residential detector can also connect directly to the command and control center. All of the three can connect to the cloud (21 1 ) for information retrieval from the I nternet. A search and rescue server can also be set up to operate in the cloud for the residential detector and the mobile rescue team to connect. Note that the connections between the residential detector and the command and control center and the cloud (I nternet) can be wired or wireless.

The method also incorporates a mobile or a web application for the residents through which they can interact with the system . Using the app, residents can mark themselves safe (either as they have gotten out of the building before its collapse or they are out-of-town far from the disaster area) so that rescue teams do not search for them in their registered residence. That feature can save valuable time for the rescue authorities to direct their efforts to locations where rescue is actually needed.

The system can also provide information about the condition of other persons in the same residence, neighbors and relatives. That also eliminates the overload of the cellular communication system , which occurs due to people calling and checking on the relatives.

The system also introduces a specific combination of various signal processing techniques and algorithms to increase the rate of detection with efficient and smart processing of sensor data, signals and information.

Through an I nternet of things (l oT) approach, the processing that takes place on the processor of the residential detector device (1 1 0, 1 1 1 ) can be performed in the cloud as well. Note that the pre-processing (107) still needs to be performed at the detector device since its purpose is to apply certain temporal and spectral filters, preconditioning and data smoothing to incoming sensor signals. Employing the cloud based solution requires, though, a robust communication system to send high-data-rate signals such as audio and video to server computers (202, 207, or 21 1 ) . Since in a disaster situation this cannot be guaranteed, the preferred embodiment of this invention employs a high processing capability on the residential detector device for processing of the sensor signals locally (i.e. edge computing) .

The system and the method utilize various signal processing techniques and algorithms for various sensors for various objectives as shown in Figure 3. The objectives differ mainly depending on the time frame before and after the disaster.

I n a preferred embodiment, the system (301 ) uses the following signals and sensor data: Audio (302) from an acoustic microphone (102) , image and video (303) from a camera (106) , vibration (304) from a vibration sensor (103) , 3-axis I MU (inertial measurement unit) data (acceleration, rotation rate and magnetic direction) (305) from the inertial sensors (104) , and location data (306) from a GNSS location sensor (105) . The signals and sensor data are preprocessed at the front end of the system for temporal and spectral filtering, preconditioning and data smoothing.

Before the disaster instant (pre-disaster time frame) , the detector device aims to collect information about the occupancy in the residence by using certain algorithms. Those are:

1 . Audio activity detection (307) and speech recognition (308) to detect human presence in the residence (i.e. occupancy) (317) ,

2. Signal feature/ signature analysis (such as fundamental frequency or pitch) (310) to differentiate human sounds (i.e. person identification) (320) and estimate the number of people in the residence (318) ,

3. Audio scene analysis (309) (through machine learning and pattern recognition algorithms) to continuously monitor sound activity in the residence and learn to recognize specific sound patterns that might be specific to the residents in order to further decide on occupancy (317) ,

4. Video activity detection (312) for video based occupancy detection (319) ,

5. I mage/Video people counting (31 1 ) for the number of occupants (318) in the residence,

6. I mage/Video face-recognition (313) for person identification (320) .

All those information when put together in a data fusion logic can reveal the human presence and the number of people in the residence i.e. pre-disaster occupancy and number (325) .

The detector device continually collects data, processes it, writes the output information on a cyclic data buffer that corresponds to a time period before the disaster instant, for instance, half an hour, so that the last half an hour of the residence is recorded in terms of human activity. The device also communicates this information in regular intervals to the command and control center (connection between 201 and 207) . The command and control center collects such information from all the residential detector devices registered to its database and monitors the usual non-alarming situation in the pre disaster time frame.

The detector device detects the disaster instant using the I MU data (305) such that a change in the usual orientation of the device can be detected by a 3-D I MU data analysis (315) , which might indicate a seismic disaster event (323) . The residential detector device might try to confirm the detection with the information available online at the cloud (21 1 ) or at the command and control centers of the disaster relief authorities (202, 207) .

The detector device remains in the rubble continuing to collect data from its sensors and processing it after the disaster instant (post-disaster time frame) , and switches to the post-disaster mode. The detector device starts broadcasting the pre-disaster occupancy information (325) stored in its memory through its RF antenna to a possible nearby mobile search and rescue team (202) and at the same time through its connection to cloud servers (21 1 ) and to the disaster command and control center (207) . Note that the connection to the cloud and the disaster command and control might have been lost in the post-disaster time frame due to a damage to the communication infrastructure. Thus, the detector device employs an RF subsystem for broadcasting to nearby receivers, for instance, those of a mobile rescue team .

I n the post-disaster mode, the detector device aims to produce a probability (for presence and number) of live humans (321 and 322) remaining in the rubble by using certain algorithms. Those are:

1 . Audio activity detection (307) to detect live human activity in the vicinity

2. Speech recognition (308) to detect human voice / speech and certain keywords and phrases of calling for help,

3. Audio scene analysis (309) to detect certain human sound patterns such as screaming, calling for help, pounding something on concrete or metal, heavy breathing, crying, etc. ,

4. Audio feature/signature analysis (31 0) to detect live human activity in the vicinity and to differentiate sounds and estimate the number of people in the rubble

5. I mage/Video activity detection (312) to detect human movement in the rubble

6. People counting from image/video data (31 1 ) to estimate number of people in the rubble

7. I mage/Video face recognition (313) to detect and recognize human faces and identify live people

8. I mpact based sound detection (314) on the vibration signal such as pounding on concrete or metal under the rubble, which may also originate far from the location of the detector device.

All those information when put together in a data fusion logic can reveal the post disaster live human presence and the number of people in the rubble (326) . The GNSS location sensor provides global position of the detector device. This is particularly useful if the residence is, for instance, a mobile home so that it can be tracked on a map. Furthermore, the device can be tracked in case of relocation from its original position as a result of the disaster, for instance, due to a tsunami. This information can even be used as an indication of a tsunami if the residential detector unit has been registered with a non-mobile residence.

All the detection and signal processing algorithms cited above are well known in electrical and electronics engineering and signal processing literature, however, there are various different techniques and algorithm designs. The details of the algorithms are not the subject of this invention but the use of a unique combination of such algorithms within an information processing system is the core of the invention.

All the information produced by the residential detector device is communicated to the search and rescue command and control center (CCC) (207) in the pre- and post-disaster time frames and to a mobile rescue team (202) in the vicinity in the post-disaster time frame. The mobile rescue team can collect, fuse and process (203) all the data coming to its servers from all the devices in the area and see and assess the results on a user interface (204) . Data and information from all devices and mobile teams are collected at the CCC, where it can be fused and processed (209) and compared againts the residents’ database (208) . According to a search and rescue coordination logic that can be utilized in the CCC (210) , search and rescue operations can be planned optimally.

As described in the previous sections, the method and the system provides various advantages in terms of efficiency, accuracy, and time of response of emergency search and rescue operations. The time of response in emergency rescue and relief is life-critical in disaster situations. As the preferred embodiment, Figure 1 shows the residential detector as a single individual device to be placed in a residence. However, a number of such devices can be used in a residence depending on the size and RF reception characteristics of the residence. I n that case, the devices form a residential detection sensor network (RDSN) , where each device is a node. One of the nodes can function as a gateway between the RDSN and the outer point of communication, which might be a mobile rescue team, the command and control center or the cloud (I nternet) . This is depicted in Figure 4, where the RDSN (401 ) is shown as a N+ 1 node sensor network with nodes (402) and a gateway (403) communicating with an outer point of communication (404) . Note that the sensor nodes in RSDN can be configured to communicate with each other as well, forming a mesh network or any other network topology. The communication modality used within the sensor network can be wired or wireless. Furthermore the communication modalities within the sensor network and that between the gateway and the outer point of communication might as well be different. For instance the RSDN nodes can communicate with each other and the gateway through Bluetooth while the RDSN gateway communicates to the mobile rescue team through 3G connection.

Futhermore, in the RDSN, the sensor nodes and the gateway do not have to be identical devices. I n such an alternative embodiment, the sensors (102 to 106) reside on each of the node devices and each node device has computing capability to collect the sensor signals/data, perform simple pre-processing (107) and send sensor signals/data to the gateway device through wireless RF communication (1 13 and 1 14) or through wired communication (1 15) , whichever used within the RDSN. The gateway device has more computing capability so that it collects the sensor signals/data coming from the nodes and performs the signal and sensor data processing (307 through 316) to serve the detection functions, and finally sends the resulting information to the outer point of communication through a wired or wireless connection.

The residential detector device (either as an individual device or a node of RDSN) has preferably hard casing to protect the electronics from impacts at a time of disaster.

The residential detector device can use multiple RF channels to maximize the connection possibility in case of a disaster. For instance the RF subsystem (1 13 and 1 14) can support Bluetooth, 3G or Wi-Fi all at the same time.

The system of live human detection for disaster emergency search and rescue comprises:

• a residential detector device that, before and after the disaster instant, provides search and rescue teams with information indicating the possibility of people being alive under the rubble of buildings hit by disasters, and communicates this information to the search and rescue command and control center (CCC) (207) in the pre- and post-disaster time frames and to a mobile rescue team (202) in the vicinity in the post-disaster time frame. A residential detector comprises sensor modules, which provide sensor data as signals including images and video; at least one microprocessor, which is capable of processing sensor signals in real-time or near real-time rates and executing software tasks for system control (109) , signal processing and analysis algorithms (1 10) , data fusion (1 1 1 ) and data communication (1 12) ; peripheral components (107) which provide pre-processing of the sensor signals; bi-directional wireless radio frequency (RF) antenna (1 14) ; an RF communication subsystem (1 13) which connects to the RF antenna (1 14) ; a wired communication subsystem (1 15) ; and a power subsystem (1 16) that manages the electrical power coming from mains electricity, battery or a solar recharge unit,

• a search and rescue command and control center ( CCC ) (207) , which provides monitoring and control of the detector device and receives/ sends information from/to the detector device, and comprises a database for residences and residents (208) , data fusion and information processing (209) , search and rescue coordination logic (210) . The CCC collects information from all the residential detector devices registered to its database and monitors the usual non-alarming situation in the pre disaster time frame, and coordinates the search and rescue operations in the post disaster time frame,

• a server computer of a mobile rescue team (202) , which comprises data fusion and processing (203) , user interface (204) , RF communication (205) and RF antenna (206) , and collects, fuses and processes (203) the data coming from all the residential detector devices in the area and see and assess the results on a user interface (204) ,

• a mobile/web application for the residents’ use, which provides information between residents, rescue teams and control centers,

• an I nternet cloud platform (21 1 ) , which provides transaction, storage, communication and processing of data

• a residential detection sensor network (RDSN) (401 ) , where each detector device is a node and one of the nodes acts as a gateway between the RDSN and the outer point of communication, which might be a mobile rescue team , the command and control center or the cloud (I nternet) .

The method (the inner workings of the detector device and the overall information system) comprises the following steps:

• Collection of signals and data from sensors in the pre-disaster time frame to produce the information of pre-disaster occupancy and number (325) ,

• Detection of the occurrence of seismic disaster instant (323) ,

• Ongoing collection of signals and data from sensors in the post-disaster time frame to produce the information of post-disaster live human presence and number (326) ,

• Preprocessing of signals and data collected from the sensors • Signal and sensor data processing of the pre-processed signals and data, which is detailed further in the following for the pre-disaster time frame:

• Audio activity detection (307) and speech recognition (308) to detect human

presence in the residence (i.e. occupancy) (317) ,

• Signal feature/signature analysis (such as fundamental frequency or pitch) (310) to differentiate human sounds (i.e. person identification) (320) and estimate the number of people in the residence (318) ,

• Audio scene analysis (309) (through machine learning and pattern recognition algorithms) to continuously monitor sound activity in the residence and learn to recognize specific sound patterns that might be specific to the residents in order to further decide on occupancy (317) ,

• Video activity detection (312) for video based occupancy detection (319) ,

• I mage/Video people counting (31 1 ) for the number of occupants in the residence (318) ,

• I mage/ Video face-recognition (313) for person identification (320) ,

• The information output (317) through (320) compose the information of pre-disaster occupancy and number (325) after being fused by the data fusion algorithms of the detector device,

• The residential detector device also communicates the pre-disaster occupancy and number information in regular intervals to the command and control center (connection between 201 and 207) . The command and control center collects such information from all the residential detector devices registered to its database and monitors the usual non-alarming situation in the pre-disaster time frame,

• The residential detector device might try to confirm the detection with the information available online at the cloud (21 1 ) or at the command and control centers of the disaster relief authorities (202, 207) ,

• The detector device remains in the rubble continuing to collect data from its sensors and processing it after the disaster instant (post-disaster time frame) , and switches to the post-disaster mode,

• Signal and sensor data processing of the pre-processed signals and data for the post disaster time frame produces the information of sound based (321 ) and image/video based (322) live human presence and number, which compose the information output of post-disaster live human presence and number (326) . The processing steps are further detailed in the following:

• Audio activity detection (307) to detect live human activity in the vicinity,

• Speech recognition (308) to detect human voice / speech and certain keywords and phrases of calling for help,

• Audio scene analysis (309) to detect certain human sound patterns such as screaming, calling for help, pounding something on concrete or metal, heavy breathing, crying, or so forth,

• Audio feature/ signature analysis (31 0) to detect live human activity in the vicinity and to differentiate sounds and estimate the number of people in the rubble,

• Detection of image/video activity (312) to detect human movement in the rubble,

• People counting from image/video data (31 1 ) to estimate number of people in the rubble,

• Face recognition of people from image/video (313) to detect and recognize human faces and identify live people,

• I mpact based sound detection (314) on the vibration signal which may also originate far from the location of the detector device,

• I n the post disaster mode, the residential detector device starts broadcasting the pre disaster occupancy information (325) stored in its memory and any detection about post-disaster live human presence and number (326) through its wired or wireless RF communication links to a possible nearby mobile search and rescue team (202) and at the same time through its connection to cloud servers (21 1 ) and to the disaster command and control center (207) .

• The residential detector device also broadcasts location information (324) through its wired or wireless RF communication links so that the current location of the device is known.

Claims

CLAI MS

1. The system of live human detection for disaster emergency search and rescue, which comprises:

• a residential detector device that, before and after the disaster instant, provides search and rescue teams with information indicating the possibility of people being alive under the rubble of buildings hit by disasters, and communicates this information to the search and rescue command and control center (CCC) (207) in the pre- and post-disaster time frames and to a mobile rescue team (202) in the vicinity in the post-disaster time frame. A residential detector comprises sensor modules, which provide sensor data as signals including images and video; at least one microprocessor, which is capable of processing sensor signals in real-time or near real-time rates and executing software tasks for system control (1 09) , signal processing and analysis algorithms (1 10) , data fusion (1 1 1 ) and data communication (1 12) ; peripheral components (107) which provide pre-processing of the sensor signals; bi-directional wireless radio frequency (RF) antenna (1 14) ; an RF communication subsystem (1 13) which connects to the RF antenna (1 14) ; a wired communication subsystem (1 15) ; and a power subsystem (1 16) that manages the electrical power coming from mains electricity, battery or a solar recharge unit,

• a search and rescue command and control center (CCC) (207) , which provides monitoring and control of the detector device and receives/sends information from/to the detector device, and comprises a database for residences and residents (208) , data fusion and information processing (209) , search and rescue coordination logic (210) . The CCC collects information from all the residential detector devices registered to its database and monitors the usual non-alarming situation in the pre-disaster time frame, and coordinates the search and rescue operations in the post-disaster time frame,

• a server computer of a mobile rescue team (202) , which comprises data fusion and processing (203) , user interface (204) , RF communication (205) and RF antenna (206) , and collects, fuses and processes (203) the data coming from all the residential detector devices in the area and see and assess the results on a user interface (204) , • a mobile/web application for the residents’ use, which provides information between residents, rescue teams and control centers,

• a residential detection sensor network (RDSN) (401 ) , where each detector device is a node and one of the nodes acts as a gateway between the RDSN and the outer point of communication, which might be a mobile rescue team, the command and control center or the cloud (I nternet) .

2. The system according to claim 1 ; wherein a residential detector or a network of them (RDSN) is positioned at a house; at an office or a residence within a building or an apartment ; or at each floor of an apartment ; outside on a street pole; on an outside wall of a building; or even at a workplace such as a coal mine.

3. The system according to claim 1 ; wherein the sensor module of the residential detector device comprises at least one acoustic sensor (microphone) (102) , at least one vibration sensor (103) , at least one inertial sensor group (accelerometer, gyro, magnetometer) (104) , at least one global navigation satellite system (GNSS) based location sensors (GPS, GLONASS, etc.) (105) and at least one image/video sensor (camera) (106) .

4. The system according to claim 1 ; wherein the microprocessor can be a digital signal processor (DSP) , a general purpose microprocessor, or a group of microprocessors from various sorts.

5. The system according to claim 1 ; wherein a radio frequency (RF) antenna (1 14) supports Bluetooth, Wi-Fi, cellular (GSM, 3G, 4G) , Zigbee or similar wireless RF communication modalities.

6. The system according to claim 1 ; wherein a the RF subsystem (1 13) and (RF) antenna 1 14 can simultaneously support multiple of the RF channels such as Bluetooth, Wi-Fi, cellular (GSM, 3G, 4G) , Zigbee or similar wireless RF communication modalities.

7. The system according to claim 1 ; wherein a RSDN nodes communicates with each other and the gateway through one of the supported RF channels while the RDSN gateway might communicate to the mobile rescue team through the same or a different supported RF channel.

8. The working method of the system according to claim 1 , comprises steps of ,

o Collection of signals and data from sensors in the pre-disaster time frame to produce the information of pre-disaster occupancy and number (325) ,

o Detection of the occurrence of seismic disaster instant (323) , Ongoing collection of signals and data from sensors in the post-disaster time frame to produce the information of post-disaster live human presence and number (326) ,

Preprocessing of signals and data collected from the sensors

Signal and sensor data processing of the pre-processed signals and data, which is detailed further in the following for the pre-disaster time frame:

Audio activity detection (307) and speech recognition (308) to detect human presence in the residence (i.e. occupancy) (317) ,

Signal feature/signature analysis (such as fundamental frequency or pitch) (310) to differentiate human sounds (i.e. person identification) (320) and estimate the number of people in the residence (318) ,

Audio scene analysis (309) (through machine learning and pattern recognition algorithms) to continuously monitor sound activity in the residence and learn to recognize specific sound patterns that might be specific to the residents in order to further decide on occupancy (317) ,

Video activity detection (312) for video based occupancy detection (319) ,

I mage/Video people counting (31 1 ) for the number of occupants in the residence (318) ,

I mage/ Video face-recognition (313) for person identification (320) ,

The information output (317) through (320) compose the information of pre disaster occupancy and number (325) after being fused by the data fusion algorithms of the detector device,

The residential detector device also communicates the pre-disaster occupancy and number information in regular intervals to the command and control center (connection between 201 and 207) . The command and control center collects such information from all the residential detector devices registered to its database and monitors the usual non-alarming situation in the pre-disaster time frame,

The residential detector device might try to confirm the detection with the information available online at the cloud (21 1 ) or at the command and control centers of the disaster relief authorities (202, 207) ,

The detector device remains in the rubble continuing to collect data from its sensors and processing it after the disaster instant (post-disaster time frame) , and switches to the post-disaster mode, Signal and sensor data processing of the pre-processed signals and data for the post-disaster time frame produces the information of sound based (321 ) and image/video based (322) live human presence and number, which compose the information output of post-disaster live human presence and number (326) . The processing steps are further detailed in the following :

Audio activity detection (307) to detect live human activity in the vicinity,

Speech recognition (308) to detect human voice / speech and certain keywords and phrases of calling for help,

Audio scene analysis (309) to detect certain human sound patterns such as screaming, calling for help, pounding something on concrete or metal, heavy breathing, crying, or so forth, Audio feature/ signature analysis (31 0) to detect live human activity in the vicinity and to differentiate sounds and estimate the number of people in the rubble, Detection of image/video activity (312) to detect human movement in the rubble, People counting from image/video data (31 1 ) to estimate number of people in the rubble,

Face recognition of people from image/video (313) to detect and recognize human faces and identify live people,

I mpact based sound detection (314) on the vibration signal which may also originate far from the location of the detector device,

I n the post disaster mode, the residential detector device starts broadcasting the pre-disaster occupancy information (325) stored in its memory and any detection about post-disaster live human presence and number (326) through its wired or wireless RF communication links to a possible nearby mobile search and rescue team (202) and at the same time through its connection to cloud servers (21 1 ) and to the disaster command and control center (207) .

The residential detector device also broadcasts location information (324) through its wired or wireless RF communication links so that the current location of the device is known.