Recognition system, recognition device, recognition method, and program

Abstract

To provide a recognition device, etc. that can suppress power consumption and computing resources and also can suppress human costs for confirmation by operating sensors installed when and where needed in order to recognize a wide range of environments.SOLUTION: A recognition device includes: a video acquisition unit using a camera to capture images of light-emitting elements corresponding to a plurality of sound-light conversion devices and acquiring video data; a specification unit for specifying sound-light conversion devices emitting a predetermined light pattern based on optical information of light-emitting elements corresponding to the plurality of sound-light conversion devices included in the video data and sending wake-up signals to sensors in an area corresponding to the specified sound-light conversion devices; and a recognition unit for receiving data acquired by the sensors from the sensors that have received the wake-up signals, and executing processing to recognize the environment based on the data acquired by the sensors.SELECTED DRAWING: Figure 5

Description

[Technical field]

[0001] The present invention relates to a wide-range environment recognition technology. [Background technology]

[0002] Wide-area environmental recognition is the problem of automatically detecting the activities of people and surrounding machines over a very wide area, such as an entire city. If this could be realized, various applications could be conceived, such as crime prevention by capturing the behavior of suspicious people in advance, speeding up life-saving efforts by automatically detecting people who suddenly fall ill, and metaverse applications that allow users to feel information about a distant city as if they were there.

[0003] So far, such wide-area environmental recognition technology has been put to practical use for security purposes, where visual monitoring of numerous surveillance cameras installed throughout the city is useful for detecting signs of crime in advance and preserving evidence of crime. In addition to human surveillance, technology that automatically detects suspicious behavior using information obtained from cameras is also beginning to be put to practical use. Furthermore, studies are being conducted on the possibility of more accurately and comprehensively capturing human activities by combining multiple distributed sensors (see Patent Document 1). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] International Publication No. 2023013043 Summary of the Invention [Problem to be solved by the invention]

[0005] Conventional wide-area environmental recognition devices are basically designed to operate all sensors at all times. However, with the emergence of technologies such as environmental recognition that combine distributed sensors, if we consider increasing the number of sensors to be installed in the future, such constantly operating systems will require a large amount of power consumption. In particular, systems that automatically detect target actions from sensor information are beginning to use deep learning models and other models that require large computational resources, and are expected to consume more power. Furthermore, even when performing visual surveillance, it requires a lot of manpower to check many sensors, so it is difficult to fully utilize them even if many sensors are installed.

[0006] The present invention aims to provide a recognition system, recognition device, recognition method, and program that can reduce power consumption and computational resources and reduce human costs for confirmation by operating sensors installed in necessary places when necessary in order to recognize a wide range of environments. [Means for solving the problem]

[0007] In order to solve the above problem, according to one aspect of the present invention, a recognition system includes a plurality of sound-to-light conversion devices, a recognition device, and a plurality of sensors. Each sound-to-light conversion device includes an acoustic signal acquisition unit that uses a microphone to collect surrounding sounds and acquire an acoustic signal, a detection unit that uses the acoustic signal and detects a predetermined event by a detection model that inputs information based on the acoustic signal and outputs a detection result indicating whether a predetermined event has occurred, and a light emission unit that causes a light-emitting element to emit light in a predetermined light emission pattern when the detection result indicates that the predetermined event has occurred. The recognition device includes a video acquisition unit that uses a camera to capture light-emitting elements corresponding to the plurality of sound-to-light conversion devices and acquires video data, an identification unit that identifies a sound-to-light conversion device that has emitted a predetermined light emission pattern based on optical information of the light-emitting elements corresponding to the plurality of sound-to-light conversion devices included in the video data, and transmits a wake-up signal to a sensor in an area corresponding to the identified sound-to-light conversion device, and a recognition unit that receives data acquired by the sensor from the sensor that has received the wake-up signal, and performs processing to recognize the environment based on the data acquired by the sensor.

[0008] In order to solve the above problem, according to another aspect of the present invention, a recognition device recognizes a wide range of environments. When a predetermined event is detected using an acoustic signal that captures surrounding sounds, the sound-to-light conversion device causes a light-emitting element to emit light in a predetermined light-emitting pattern, and the recognition device includes: a video acquisition unit that uses a camera to capture light-emitting elements corresponding to a plurality of sound-to-light conversion devices and acquires video data; a determination unit that identifies a sound-to-light conversion device that has emitted the predetermined light-emitting pattern based on optical information of the light-emitting elements corresponding to the plurality of sound-to-light conversion devices included in the video data and transmits a wake-up signal to a sensor in an area corresponding to the identified sound-to-light conversion device; and a recognition unit that receives data acquired by the sensor from the sensor that has received the wake-up signal and performs processing to recognize the environment based on the data acquired by the sensor.

[0009] In order to solve the above problem, according to another aspect of the present invention, a recognition method uses a plurality of sound-to-light conversion devices, a recognition device, and a plurality of sensors. The recognition method includes an acoustic signal acquisition step in which each sound-to-light conversion device uses a microphone to collect surrounding sounds and acquires an acoustic signal, a detection step in which each sound-to-light conversion device detects a predetermined event using an acoustic signal and a detection model that outputs a detection result indicating whether a predetermined event has occurred using information based on the acoustic signal as an input, a light emission step in which each sound-to-light conversion device causes a light-emitting element to emit light in a predetermined light emission pattern when the detection result indicates that a predetermined event has occurred, a video acquisition step in which the recognition device uses a camera to capture a light-emitting element corresponding to the plurality of sound-to-light conversion devices and acquires video data, a specification step in which the recognition device specifies a sound-to-light conversion device that has emitted a predetermined light emission pattern based on the light information of the light-emitting element corresponding to the plurality of sound-to-light conversion devices included in the video data, and transmits a wake-up signal to a sensor in an area corresponding to the specified sound-to-light conversion device, and a recognition step in which the recognition device receives data acquired by the sensor from the sensor that received the wake-up signal, and performs a process to recognize the environment based on the data acquired by the sensor. Effect of the Invention

[0010] According to the present invention, it is possible to reduce power consumption and calculation resources, and to reduce the personnel costs required for confirmation. [Brief description of the drawings]

[0011] [Figure 1] 1 is a diagram showing an example of the configuration of a wide-range environment recognition system according to a first embodiment; [Diagram 2] 1 is a diagram showing an example of the arrangement of a sound-to-light converter, a camera, and a sensor according to a first embodiment; [Diagram 3] FIG. 2 is a diagram showing an example of a processing flow of the wide-area environment recognition system according to the first embodiment. [Figure 4] FIG. 2 is a functional block diagram of a sound-to-light converter. [Diagram 5] FIG. 2 is a functional block diagram of a recognition device. [Figure 6]FIG. 13 is a diagram showing an example of the configuration of a computer to which the present technique is applied. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] Hereinafter, an embodiment of the present invention will be described. In the drawings used in the following description, components having the same functions and steps performing the same processes are denoted by the same reference numerals, and duplicated explanations will be omitted.

[0013] <Key Points of the First Embodiment> In the wide-area environment recognition system of this embodiment, a sound-to-light converter is used for system wake-up. The sound-to-light converter is a device that emits light corresponding to an observed sound signal. In this embodiment, it is considered that sounds different from normal sounds should be generated around the object to be recognized. Therefore, in this system, sound-to-light converters are sparsely arranged in the space to be recognized and are set to emit light when an acoustic signal different from normal sounds by a certain amount or more is obtained. This light is captured by a small number of cameras installed at a distance, thereby limiting the area in which the object to be recognized exists. In other words, by using the light emitted by the sound-to-light converter as a trigger to wake up the surrounding sensors, it becomes possible to efficiently perform environmental recognition using a large number of sensor systems only at the place and time where the object to be recognized exists.

[0014] By using the sound-to-light converter for system wake-up, it becomes possible to drive only the sensors at the times and places necessary for wide-area environment recognition, which makes it possible to reduce power consumption. Furthermore, it is expected that the power consumption saved by this mechanism can be allocated to the computational resources, enabling high-precision environment recognition using a larger number of sensors.

[0015] First Embodiment FIG. 1 shows an example of the configuration of a wide-range environment recognition system according to the first embodiment.

[0016] The wide-range environment recognition system includes a recognition device 200, N sound-light conversion devices 100-(k,m), and Q sensors 300-(k,m,p). The recognition device 200 acquires video data using K cameras 90-k. The video data acquired using the k-th camera 90-k includes n k The camera 90-k is installed so that the light emitting element corresponding to the sound-light conversion device 100-(k,m) is captured. k,m Sensors 300-(k, m, p) are arranged such that k=1, 2, ..., K, m=1, 2, ..., n k , p=1,2,…,q k,m year,

number

number

[0017] FIG. 2 shows an example of the arrangement of sound-to-light converters 100-(k,m), cameras 90-k, and sensors 300-(k,m,p). The sound-to-light converters 100-(k,m) turn on or off the light-emitting elements and use the camera 90-k to obtain video data of the light-emitting elements, whereby the light converters 100-(k,m) can transmit light information to the recognition device 200. The light information is, for example, information indicating whether the light is on or off. For example, a signal in which the on state is 1 and the off state is 0 can be expressed by the light information. On the other hand, when n signals are received from the recognition device 200, k2 indicates the position of the sound-light converter 100-(k,m), and the white circle in FIG. 2 indicates the position of the sensor 300-(k,m,p). In the example of FIG. 2, K=4, n1=3, and q 1,1 = 4. The N sound-light converters 100-(k, m) are arranged so that the microphones can pick up sounds generated in the area that is the target of wide-range environment recognition evenly. At the same time, it is desirable to arrange the N sound-light converters 100-(k, m) sparsely so that, at the minimum volume of the sound generated when a specific event occurs, the sound picked up by the microphone corresponding to one sound-light converter and the sound picked up by the microphone corresponding to another sound-light converter do not come from the same sound source.

[0018] The sound-light converter 100-(k,m) collects surrounding sounds using a microphone, and when it collects a sound generated when a predetermined event occurs, causes a light-emitting element to emit light in a predetermined emission pattern.

[0019] When the recognition device 200 receives the predetermined light emission pattern, the recognition device 200 recognizes q k,m A wake-up signal is sent to each sensor 300-(k,m,p) and sensor information x(k,m,p) acquired by the activated sensor 300-(k,m,p) is received. For example, if the sensor is a microphone, the "sensor information" refers to a sound signal acquired by the microphone. Also, if the sensor is a camera, the "sensor information" refers to a video signal acquired by the camera.

[0020] The recognition device 200 k,m Based on the sensor information x(k,m,p), processing for recognizing the environment, such as event detection, is executed and the recognition result is output.

[0021] The sound-light conversion device 100-(k,m) and the recognition device 200 are special devices configured by loading a special program into a publicly known or dedicated computer having, for example, a central processing unit (CPU), a main storage device (RAM), etc. The sound-light conversion device 100-(k,m) and the recognition device 200 execute each process under the control of, for example, the central processing unit. Data input to the sound-light conversion device 100-(k,m) and the recognition device 200 and data obtained in each process are stored in, for example, the main storage device, and the data stored in the main storage device is read out to the central processing unit as necessary and used for other processes. At least a part of each processing unit of the sound-light conversion device 100-(k,m) and the recognition device 200 may be configured by hardware such as an integrated circuit. Each storage unit of the sound-light conversion device 100-(k,m) and the recognition device 200 may be configured by, for example, a main storage device such as a RAM (Random Access Memory), or middleware such as a relational database or a key-value store. However, each memory unit does not necessarily need to be provided inside the sound-light conversion device 100-(k,m) and the recognition device 200, but may be configured as an auxiliary memory device consisting of a semiconductor memory element such as a hard disk, optical disk, or flash memory, and be provided outside the sound-light conversion device 100-(k,m) and the recognition device 200.

[0022] Fig. 3 shows the process flow of the wide-area environment recognition system. The process of each device will be explained with reference to Fig. 3.

[0023] <Sound-light conversion device 100-(k,m)> 4 is a functional block diagram of the sound-to-light converter 100-(k,m). The sound-to-light converter 100-(k,m) includes an acoustic signal acquisition unit 110, a detection unit 120, and a light emission unit .

[0024] (Acoustic signal acquisition unit 110) The acoustic signal acquisition unit 110 collects surrounding sounds using a single-channel or multi-channel microphone, acquires an acoustic signal (S110), and outputs the signal. The acoustic signal acquisition unit 110 may include a single-channel or multi-channel microphone, or may be configured to receive an acoustic signal output by an external single-channel or multi-channel microphone.

[0025] (Detection unit 120) The detection unit 120 receives an acoustic signal as input, detects a predetermined event using a detection model (S120), and outputs a detection result. The detection model receives information based on an acoustic signal (for example, the acoustic signal itself or an acoustic feature (for example, MFCC, etc.) extracted from the acoustic signal) as input, and outputs a detection result indicating whether or not a predetermined event has occurred. For example, the predetermined event may be the occurrence of an abnormal sound, and for example, an existing abnormal sound detection technology may be used (see, for example, Reference 1). (Reference 1) JP 2019-036112 A

[0026] (Light emitting unit 130) The light emitting unit 130 receives the detection result as an input, and when the detection result indicates that a predetermined event has occurred, causes one or more light emitting elements to emit light in a predetermined light emitting pattern (S130). The light emitting unit 130 may include one or more light emitting elements, or may be configured to output an electrical signal corresponding to the predetermined light emitting pattern to one or more external light emitting elements. For example, when the detection result indicates that a predetermined event has occurred, the light emitting element is turned on, and when the detection result indicates that the predetermined event has not occurred, the light emitting element is turned off.

[0027] The processes performed by the detection unit 120 and the light emission unit 130 are collectively referred to as a sound-to-light conversion algorithm. The sound-to-light conversion algorithm can also be said to be an algorithm that causes light to be emitted when an event to be observed (a predetermined event) occurs.

[0028] For example, a street light can be used as a light emitting element, and a microphone can be embedded in the pole of the street light. For example, by blinking the light emitting element at a speed that humans cannot distinguish the blinking (setting the light emitting element to be in an off state for a time shorter than 50 ms), light information can be transmitted to the recognition device 200 without being noticed by people around the street light. When the purpose is to detect signs of a crime in advance or to preserve evidence of a crime, such a configuration is effective because it is not desired for people who are about to commit a crime or criminals (hereinafter also referred to as "criminals, etc.") in the vicinity of the street light to realize that their voice will be collected by a microphone or that their image will be captured by a camera. Conversely, by blinking the street light in a light emission pattern that is clearly different from that of a normal street light so that people around the street light will unintentionally look at the street light, light information can be transmitted to the recognition device 200 and the attention of the surroundings can be attracted. When the purpose is to prevent or stop a crime, such a configuration is effective because criminals, etc., want to avoid being noticed. However, actual environment recognition needs to be performed using sensor information. <Recognition device 200> 5 shows a functional block diagram of the recognition device 200. The recognition device 200 includes an image acquisition unit 210, an identification unit 220, and a recognition unit 230.

[0029] (Video acquisition unit 210) The image acquisition unit 210 k Using a single or multiple cameras 90-k installed at a distance so that the sound-to-light conversion devices 100-(k,m) can be captured at once, kThe light-emitting elements corresponding to the sound-light conversion devices 100-(k,m) are photographed, and the image data is acquired (S210) and output. As described above, k=1, 2, ..., K, and the image acquisition unit 210 acquires image data photographing the light-emitting elements corresponding to a total of N sound-light conversion devices 100-(k,m) from the K cameras 90-k. The image acquisition unit 210 may include a single or multiple cameras 90-k, or may be configured to receive image data output from a single or multiple external cameras 90-k. For example, a single or multiple cameras 90-k may be installed on an aircraft such as a drone, on the rooftop of a building, etc.

[0030] (Specific section 220) The identification unit 220 receives the video data as input, and identifies the sound-light conversion device 100-(k,m) that has emitted a predetermined light emission pattern based on the light information of the light-emitting elements corresponding to a total of N sound-light conversion devices 100-(k,m) contained in the video data (S220), and identifies q of the areas corresponding to the identified sound-light conversion devices 100-(k',m'). k',m' A wake-up signal is transmitted to the sensors 300-(k',m',p) (S221). Note that k' is a number indicating a camera corresponding to the video data capturing the light-emitting element emitting the predetermined light-emitting pattern, and m' is a number indicating a sound-to-light conversion device corresponding to the light-emitting element emitting the predetermined light-emitting pattern captured in the video data of the k'th camera 90-k'. The area corresponding to the identified sound-to-light conversion device 100-(k',m') is the periphery of the microphone corresponding to the identified sound-to-light conversion device 100-(k',m'). Note that the sound-to-light conversion device 100-(k,m) corresponding to the light-emitting element emitting the predetermined light-emitting pattern is not limited to one, and there may be no sound-to-light conversion device corresponding to the light-emitting element emitting the predetermined light-emitting pattern, or there may be two or more sound-to-light conversion devices. For example, when the volume of the sound generated when a predetermined event occurs is large, two or more sound-to-light conversion devices 100-(k,m) may be identified.

[0031] For example, when the camera 90-k is installed on the rooftop of a building, the area in the video data where the light-emitting element corresponding to the sound-light conversion device 100-(k,m) is shown does not change, so the identification unit 220 acquires light information from the video data of that area. Also, when the camera 90-k is installed on an aircraft such as a drone, the area in the video data where the light-emitting element corresponding to the sound-light conversion device 100-(k,m) is shown changes, so the identification unit 220 uses the position information of the light-emitting element corresponding to the sound-light conversion device 100-(k,m) and information indicating the position, altitude, orientation, viewing angle, etc. of the camera 90-k at each time to calculate the area in the video data where the light-emitting element corresponding to the sound-light conversion device 100-(k,m) is shown at each time, and acquires light information from the video data of that area. For information indicating the position, altitude, orientation, viewing angle, etc. of the camera 90-k, for example, sensor information of an aircraft such as a drone is used.

[0032] The identification unit 220 determines whether the acquired light information matches a predetermined light emission pattern, and if there is a match, identifies the sound-to-light conversion device 100-(k,m) corresponding to the light-emitting element that emitted the matching light information.

[0033] For example, a sound-light conversion device 100-(k,m) and q microphones corresponding to the sound-light conversion device 100-(k,m) are stored in a storage unit (not shown). k,m The sensor 300-(k, m, p) is stored in association with the sensor 300-(k, m, p).

[0034] The identification unit 220 refers to a storage unit (not shown) and identifies q microphones in the vicinity of the microphone corresponding to the identified sound-light conversion device 100-(k',m'). k',m' The sensors 300-(k', m', p) are identified and a wake-up signal is transmitted to them.

[0035] (Sensor 300-(k,m,p)) q k',m'When the sensor 300-(k', m', p) receives the wake-up signal, it starts up, starts collecting target information, obtains sensor information x(k', m', p) (S300), and transmits it to the recognition device 200. Here, p=1, 2, ..., q m',k' The sensor 300-(k,m,p) may be an element or device that collects desired information and converts it into a signal that can be processed by a machine, such as a single or multiple microphones, a single or multiple cameras, or a device including a microphone and a camera. The sensor 300-(k,m,p) before receiving the wake-up signal is in a standby state, and can reduce power consumption compared to when it is in an active state. The sensor 300-(k,m,p) in the standby state does not transmit sensor information to the recognition device 200, and can reduce the computational resources in the recognition unit 230.

[0036] (Recognition unit 230) The recognition unit 230 is k',m' Receive sensor information x(k',m',p) and k',m' Based on the pieces of sensor information x(k', m', p), a process for recognizing the environment, such as event detection, is executed (S230), and the recognition result is output. For example, using the technology of Patent Document 1, the recognition unit 230 executes an environment recognition task, such as event detection.

[0037] <Effects> With the above configuration, it is possible to reduce power consumption and computational resources in wide-area environment recognition, and to reduce the human costs required for confirmation.

[0038] <Modification> In the first embodiment shown in Fig. 2, a certain light-emitting element is photographed by one camera, but a configuration in which a certain light-emitting element is photographed by two or more cameras may be used. The point is that the position of the light-emitting element captured in the video data should be known. In this embodiment, the camera 90-k, sound-to-light conversion device 100-(k,m), and sensor 300-(k,m,p) are used, and the camera, sound-to-light conversion device, and sensor are expressed as being associated with each other; however, it is sufficient that the sound-to-light conversion device and the sensor located in the vicinity of the microphone corresponding to that sound-to-light conversion device are associated with each other, and the camera and sensor do not necessarily have to be associated with each other.

[0039] In Figure 2 of the first embodiment, one sensor 300-(k,m,p) is associated with one sound-light conversion device 100-(k,m), but if part of the areas corresponding to two or more sound-light conversion devices 100-(k,m) overlap and one sensor 300-(k,m,p) is placed in that overlapping area, one sensor 300-(k,m,p) may be configured to be associated with two or more sound-light conversion devices 100-(k,m).

[0040] In the first embodiment, the predetermined light emission pattern is described only as a light emission pattern indicating the occurrence of a predetermined event, but a predetermined light emission pattern indicating other information may be used. For example, a light emission pattern indicating the type of a predetermined event or the direction or position where the predetermined event occurred is shared between the sound-to-light conversion device 100-(k,m) and the recognition device 200, and the sound-to-light conversion device 100-(k,m) turns on or off the light-emitting element and obtains video data of the light-emitting element using the camera 90-k, so that the light conversion device 100-(k,m) can transmit the above-mentioned information to the recognition device 200. Examples of the type of the predetermined event include the sound of a car horn, the sound of braking, and a scream. When transmitting information indicating the direction or position where the predetermined event occurred, it is possible to use a directional microphone array or the like as a microphone for collecting surrounding sounds. In addition, when transmitting a wake-up signal, a configuration may be used in which the wake-up signal is transmitted only to the sensor 300-(k,m,p) corresponding to the direction or position where the predetermined event occurred. In the first embodiment, an example of information indicating whether the light is on or off is shown as the light information, but the light information is not limited to this. For example, the light information may be information indicating light intensity.

[0041] <Other Modifications> The present invention is not limited to the above-mentioned embodiment and modified examples. For example, the above-mentioned various processes may be executed not only in chronological order as described, but also in parallel or individually depending on the processing capacity of the device executing the processes or as necessary. In addition, appropriate modifications are possible within the scope of the present invention.

[0042] <Programs and recording media> The various processes described above can be implemented by loading a program for executing each step of the above method into the recording unit 2020 of the computer 2000 shown in FIG. 6, and operating the control unit 2010, input unit 2030, output unit 2040, display unit 2050, etc.

[0043] The program describing the processing contents can be recorded on a computer-readable recording medium, which may be, for example, a magnetic recording device, an optical disk, a magneto-optical recording medium, a semiconductor memory, or any other type of recording medium.

[0044] The program may be distributed, for example, by selling, transferring, lending, etc. portable recording media such as DVDs and CD-ROMs on which the program is recorded. Furthermore, the program may be distributed by storing the program in a storage device of a server computer and transferring the program from the server computer to other computers via a network.

[0045] A computer that executes such a program, for example, first stores the program recorded on a portable recording medium or the program transferred from a server computer in its own storage device. Then, when executing a process, the computer reads the program stored in its own recording medium and executes a process according to the read program. As another execution form of the program, the computer may directly read the program from the portable recording medium and execute a process according to the program, or may execute a process according to the received program each time a program is transferred from the server computer to the computer. The above-mentioned process may be executed by a so-called ASP (Application Service Provider) type service that realizes a processing function only by issuing an execution instruction and obtaining a result, without transferring a program from the server computer to the computer. Note that the program in this embodiment includes information used for processing by an electronic computer and equivalent to a program (data that is not a direct command to the computer but has a nature that specifies the processing of the computer, etc.).

[0046] In addition, in this embodiment, the present device is configured by executing a specific program on a computer, but at least a part of the processing contents may be realized by hardware.

Claims

1. A recognition system including a plurality of sound-to-light conversion devices, a recognition device, and a plurality of sensors, Each of the sound-to-light converting devices is an acoustic signal acquisition unit that collects surrounding sounds using a microphone and acquires acoustic signals; a detection unit that detects a predetermined event by a detection model that uses the acoustic signal, receives information based on the acoustic signal as an input, and outputs a detection result indicating whether a predetermined event has occurred; a light emitting unit that causes a light emitting element to emit light in a predetermined light emitting pattern when the detection result indicates that a predetermined event has occurred; The recognition device includes: an image acquiring unit that acquires image data by capturing images of light-emitting elements corresponding to the plurality of sound-to-light converters using a camera; an identification unit that identifies a sound-to-light conversion device that has emitted a predetermined light emission pattern based on light information of the light-emitting elements corresponding to the plurality of sound-to-light conversion devices included in the video data, and transmits a wake-up signal to a sensor in an area corresponding to the identified sound-to-light conversion device; a recognition unit that receives data acquired by the sensor from the sensor that has received the wake-up signal, and executes a process for recognizing an environment based on the data acquired by the sensor; Recognition system.

2. A recognition device that performs wide-range environment recognition, the sound-to-light converter causes the light-emitting element to emit light in a predetermined light-emitting pattern when a predetermined event is detected using an acoustic signal obtained by collecting surrounding sounds; an image acquiring unit that acquires image data by capturing images of light-emitting elements corresponding to a plurality of sound-to-light converters using a camera; an identification unit that identifies a sound-to-light conversion device that has emitted a predetermined light emission pattern based on light information of the light-emitting elements corresponding to the plurality of sound-to-light conversion devices included in the video data, and transmits a wake-up signal to a sensor in an area corresponding to the identified sound-to-light conversion device; a recognition unit that receives data acquired by the sensor from the sensor that has received the wake-up signal, and executes a process for recognizing an environment based on the data acquired by the sensor; recognition device.

3. A recognition method using a plurality of sound-to-light conversion devices, a recognition device, and a plurality of sensors, comprising: an acoustic signal acquisition step in which each of the sound-to-light conversion devices picks up surrounding sounds using a microphone and acquires an acoustic signal; a detection step in which each of the sound-to-light conversion devices detects a predetermined event using the sound signal by a detection model that receives information based on the sound signal as an input and outputs a detection result indicating whether or not a predetermined event has occurred; a light emitting step of causing a light emitting element of each of the sound-to-light converters to emit light in a predetermined light emitting pattern when a detection result indicates that a predetermined event has occurred; an image acquiring step in which the recognition device captures images of light-emitting elements corresponding to the plurality of sound-to-light conversion devices using a camera to acquire image data; a step of the recognition device identifying a sound-to-light conversion device that has emitted a predetermined light emission pattern based on light information of the light-emitting element corresponding to the plurality of sound-to-light conversion devices included in the video data, and transmitting a wake-up signal to a sensor in an area corresponding to the identified sound-to-light conversion device; a recognition step in which the recognition device receives data acquired by the sensor from the sensor that has received the wake-up signal, and executes a process for recognizing an environment based on the data acquired by the sensor. Recognition method.

4. A program for causing a computer to function as the recognition device of claim 2.

Images