Event recognition device, event recognition method, and program

Abstract

To recognize an event without performing model learning using an observation signal acquired by optical fiber sensing.SOLUTION: An event recognition device (20) comprises: an acquisition section (21) that acquires an observation signal indicating sound or vibration that occurs at a point along an optical fiber and is detected by optical fiber sensing and also acquires a noise signal indicating optical noise; a derivation section (22) that uses statistical information (23) on a sound signal or a vibration signal acquired by a sound sensor or a vibration sensor to derive a posterior probability that an event occurs at a point along the optical fiber from the observation signal and derive a posterior probability that the event occurs at a point along the optical fiber from the noise signal; and a comparison section (24) that compares the posterior probability derived from the observation signal with the posterior probability derived from the noise signal, derives, from a result of the comparison, the posterior probability that the event occurs at a point along the optical fiber, and outputs a result of the derivation.SELECTED DRAWING: Figure 8

Description

[Technical field]

[0001] The present disclosure relates to an event recognition device, an event recognition method, and a program. [Background technology]

[0002] Optical fiber sensing, typified by distributed acoustic sensing (DAS), is capable of detecting sounds and vibrations generated at points along an optical fiber.

[0003] In recent years, technologies have been proposed that acquire observation signals indicating sound or vibration that occurs at points along an optical fiber and is detected by optical fiber sensing, and recognize events such as abnormalities that have occurred at points along the optical fiber based on the acquired observation signals.

[0004] Furthermore, a technology for recognizing an event based on an observation signal acquired by optical fiber sensing includes a technology for performing model learning on the observation signal to generate a machine learning model and recognizing an event using the generated machine learning model. For example, Patent Document 1 discloses a technology for recognizing an event using a machine learning model generated by supervised learning. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] International Publication No. 2020 / 044648 Summary of the Invention [Problem to be solved by the invention]

[0006] Incidentally, machine learning techniques include the above-mentioned supervised learning, as well as transfer learning, fine-tuning, domain adaptation, self-supervised learning, etc. When using these techniques in optical fiber sensing, all of them require observed signals acquired by optical fiber sensing for model learning.

[0007] However, the observation signals acquired by optical fiber sensing are low quality signals with low SNR (Signal-to-Noise Ratio) and non-uniform sensitivity due to optical noise (dominated by shot noise with high whiteness). Therefore, there are few observation signals that are effective for model learning, and it is difficult to perform model learning using the observation signals and generate a machine learning model.

[0008] Therefore, there is a demand for a technology that can recognize events using observed signals acquired by optical fiber sensing without performing model learning.

[0009] In view of the above-mentioned problems, an object of the present disclosure is to provide an event recognition device, an event recognition method, and a program capable of recognizing an event without performing model learning using observation signals acquired by optical fiber sensing. [Means for solving the problem]

[0010] An event recognition apparatus according to one aspect includes: an acquisition unit that acquires an observation signal indicative of sound or vibration generated at a point along an optical fiber and detected by optical fiber sensing, and also acquires a noise signal indicative of optical noise present at the point along the optical fiber; a derivation unit that utilizes statistical information of an acoustic signal or a vibration signal acquired by an acoustic sensor or a vibration sensor to derive a posterior probability that at least one event has occurred at a point along the optical fiber from the observed signal, and that derives the posterior probability that the at least one event has occurred at a point along the optical fiber from the noise signal; a comparison unit that compares the posterior probability derived from the observed signal with the posterior probability derived from the noise signal, derives a posterior probability that the at least one event has occurred at a point along the optical fiber from a comparison result, and outputs the derived result; Equipped with.

[0011] An event recognition method according to one aspect includes: An event recognition method by an event recognition device, acquiring an observation signal indicative of sound or vibration occurring at a point along the optical fiber and detected by optical fiber sensing, and acquiring a noise signal indicative of optical noise present at the point along the optical fiber; a derivation step of deriving, from the observed signals, a posterior probability that at least one event has occurred at a point along the optical fiber, using statistical information of an acoustic signal or a vibration signal acquired by an acoustic sensor or a vibration sensor, and deriving, from the noise signal, a posterior probability that the at least one event has occurred at a point along the optical fiber; a comparing step of comparing the posterior probability derived from the observed signal with the posterior probability derived from the noise signal, deriving from the comparison result a posterior probability that the at least one event has occurred at a point along the optical fiber, and outputting the derived result; Includes.

[0012] A program according to one aspect includes: On the computer, an acquisition step of acquiring an observation signal indicative of sound or vibration occurring at a point along an optical fiber and detected by optical fiber sensing, and acquiring a noise signal indicative of optical noise present at the point along the optical fiber; a derivation step of deriving, from the observed signals, a posterior probability that at least one event has occurred at a point along the optical fiber, using statistical information of an acoustic signal or a vibration signal acquired by an acoustic sensor or a vibration sensor, and deriving, from the noise signal, a posterior probability that the at least one event has occurred at a point along the optical fiber; a comparison step of comparing the posterior probability derived from the observed signal with the posterior probability derived from the noise signal, deriving from the comparison a posterior probability of the at least one event occurring at a point along the optical fiber, and outputting the derived result; Execute the command. Effect of the Invention

[0013] According to the above-described aspects, it is possible to provide an event recognition device, an event recognition method, and a program that are capable of recognizing an event without performing model learning using an observation signal acquired by optical fiber sensing. [Brief description of the drawings]

[0014] [Figure 1] FIG. 11 is a diagram showing an example of a comparison between the distribution of an acoustic signal acquired by a microphone and the distribution of an observation signal indicating an acoustic signal acquired by a DAS. [Diagram 2] 1 is a block diagram showing an example of a configuration of an event recognition device according to a first embodiment. [Diagram 3] FIG. 2 is a diagram illustrating a schematic operation example of the event recognition device according to the first embodiment. [Figure 4] 4 is a diagram illustrating a specific operation example of the event recognition device according to the first embodiment. FIG. [Diagram 5] 4 is a diagram illustrating a specific operation example of the event recognition device according to the first embodiment. FIG. [Figure 6] 2 is a diagram illustrating an example of an event recognition result by the event recognition device according to the first embodiment, in comparison with an event recognition result by the related art. FIG. [Figure 7]2 is a diagram illustrating an example of an event recognition result by the event recognition device according to the first embodiment, in comparison with an event recognition result by the related art. FIG. [Figure 8] FIG. 11 is a block diagram showing an example of a configuration of an event recognition device according to a second embodiment. [Figure 9] FIG. 2 is a block diagram showing an example of a hardware configuration of a computer that realizes the event recognition device according to each embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0015] Hereinafter, an embodiment of the present disclosure will be described with reference to the drawings. Note that the following description and drawings are omitted and simplified as appropriate for clarity of explanation. In addition, in each of the following drawings, the same elements are given the same reference numerals, and duplicated explanations are omitted as necessary. In addition, the specific numerical values ​​etc. shown below are merely examples for facilitating understanding of the present disclosure, and are not limited thereto.

[0016] <Outline of each embodiment> Before describing the details of each embodiment of the present disclosure, an overview of each embodiment will be described. First, an example of comparing the distribution of an acoustic signal acquired by a microphone with the distribution of an observation signal indicating an acoustic signal acquired by a DAS will be described with reference to Fig. 1. Note that in Fig. 1, the distribution of the observation signal is a distribution after removing the influence of optical noise.

[0017] As shown in FIG. 1, it can be seen that the distribution of the observed signal indicating the sound acquired by the DAS is very similar to the distribution of the sound signal acquired by the microphone, when the influence of optical noise is eliminated.

[0018] In addition, statistical information such as open source data and machine learning models regarding acoustic signals or vibration signals acquired by acoustic sensors such as microphones or vibration sensors is available on the market.

[0019] Therefore, in each embodiment described below, statistical information of acoustic signals or vibration signals acquired by acoustic sensors or vibration sensors available on the market is used, which makes model learning using observed signals acquired by optical fiber sensing unnecessary.

[0020] In each embodiment described below, the above-mentioned statistical information is used to derive a posterior probability of an event occurrence from each of an observation signal indicating sound or vibration detected by optical fiber sensing and a noise signal indicating optical noise. Then, the posterior probability derived from the observation signal is compared with the posterior probability derived from the noise signal, and finally, a posterior probability of an event occurrence is derived from the comparison result. This eliminates the influence of optical noise and improves event recognition performance. Each embodiment of the present disclosure will be described below.

[0021] <Embodiment 1> First, a configuration example of the event recognition device 10 according to the first embodiment will be described with reference to Fig. 2. The first embodiment is an example in which a machine learning model trained from an acoustic signal or a vibration signal acquired by an acoustic sensor or a vibration sensor is used as statistical information of the acoustic signal or the vibration signal. As shown in FIG. 2, the event recognition device 10 according to the first embodiment includes an acquisition unit 11, a derivation unit 12, and a comparison unit .

[0022] The acquisition unit 11 acquires an observation signal indicating sound or vibration generated at each point along the optical fiber and detected by optical fiber sensing. For example, the acquisition unit 11 may acquire the observation signal from a sensing device such as a DAS device. In addition, for example, the observation signal may be a time domain signal R that indicates a time change in the intensity of sound or vibration generated at each point along the optical fiber. T Alternatively, the observed signal may be a time domain signal R T The frequency domain signal C after Fourier transform of T×F It may be.

[0023] The acquiring unit 11 also acquires a noise signal indicating optical noise present at an arbitrary point along the optical fiber. The optical noise is dominated by shot noise with high whiteness. When the type of optical noise or the time period in which the optical noise exists is unknown, the acquiring unit 11 may acquire an arbitrary noise signal simulating an arbitrary type of optical noise from an arbitrary device. Alternatively, when the time period in which an event occurs is known, the acquiring unit 11 may acquire an observation signal in a time period in which no event occurs as a noise signal from a sensing device such as a DAS device.

[0024] The derivation unit 12 holds a machine learning model 13 that is available on the market and is an example of statistical information of an acoustic signal or a vibration signal acquired by an acoustic sensor or a vibration sensor. The machine learning model 13 is a model that outputs a posterior probability that at least one event has occurred for an input signal. Furthermore, the machine learning model 13 is a model that is trained using an acoustic signal or a vibration signal acquired by an acoustic sensor or a vibration sensor as an input signal. The acoustic sensor is, for example, a microphone that acquires an acoustic signal, and the vibration sensor is, for example, an acceleration sensor or a hydrophone that acquires a vibration signal.

[0025] The derivation unit 12 utilizes the machine learning model 13 to derive, for each point along the optical fiber, a posterior probability that at least one event has occurred at that point from the observation signal at that point acquired by the acquisition unit 11. In detail, for each point, the derivation unit 12 inputs the observation signal at that point as an input signal to the machine learning model 13, and obtains, as an output signal of the machine learning model 13, the posterior probability that at least one event has occurred at that point.

[0026] Furthermore, the derivation unit 12 utilizes the machine learning model 13 to derive a posterior probability that at least one event has occurred at any point along the optical fiber, from a noise signal present at the arbitrary point acquired by the acquisition unit 11. In detail, the derivation unit 12 inputs the noise signal at the arbitrary point as an input signal to the machine learning model 13, and obtains, as an output signal of the machine learning model 13, the posterior probability that at least one event has occurred at the arbitrary point.

[0027] The comparison unit 14 compares, for each point along the optical fiber, the posterior probability derived from the observed signal at that point with the posterior probability derived from the noise signal by the derivation unit 12. Then, from the comparison result, the comparison unit 14 finally derives the posterior probability that at least one event has occurred at that point. Furthermore, the comparison unit 14 outputs the posterior probability that at least one event has occurred derived for each point to the outside of the event recognition device 10 as an event recognition result.

[0028] Next, a schematic operation example of the event recognition device 10 according to the first embodiment will be described with reference to Fig. 3. In Fig. 3, it is assumed that the acquisition unit 11 has already acquired the observed signals at each point along the optical fiber and the noise information at an arbitrary point, and a subsequent operation example will be described (the same applies to Figs. 4 and 5 described later).

[0029] As shown in Figure 3, for each point, the derivation unit 12 inputs the observation signal at that point as an input signal to the machine learning model 13 (step S11), and obtains the posterior probability that at least one event has occurred at that point as an output signal of the machine learning model 13 (step S13).

[0030] In addition, the derivation unit 12 inputs a noise signal at an arbitrary point as an input signal to the machine learning model 13 (step S12), and obtains, as an output signal of the machine learning model 13, the posterior probability that at least one event has occurred at the arbitrary point (step S14).

[0031] The comparison unit 14 compares, for each point, the posterior probability derived from the observed signal at that point with the posterior probability derived from the noise signal, and finally, from the comparison result, derives the posterior probability that at least one event has occurred at that point (step S15).

[0032] After that, the comparison unit 14 outputs the posterior probability that at least one event has occurred, which is derived for each point, as an event recognition result to the outside of the event recognition device 10 (step S16).

[0033] Next, a specific operation example of the event recognition device 10 according to the first embodiment will be described. (1) Specific operation example 1 First, a first specific operation example of the event recognition device 10 according to the first embodiment will be described with reference to FIG. The present operation example 1 is an example in which the type of optical noise present at any point along the optical fiber and the time period in which the optical noise exists are unknown.

[0034] In the case of this operation example 1, the acquisition unit 11 acquires an arbitrary noise signal that simulates an arbitrary type of optical noise from an arbitrary device. For example, the noise signal in this operation example 1 is a signal (N(μ,√μ)) that follows the central limit theorem when the number of detected photons is large, assuming a Poisson distribution. The acquisition unit 11 also receives the above-mentioned time domain signal R as an observation signal at each point along the optical fiber from a sensing device such as a DAS device. T or frequency domain signal C T×F Get the.

[0035] As shown in FIG. 4, the derivation unit 12 calculates the observation signal R T Or C T×Fis input as an input signal to the machine learning model 13 (step S21). As a result, the derivation unit 12 obtains, for each point, the posterior probability that each of the five events occurs at that point as an output signal of the machine learning model 13 (step S23). Of the posterior probabilities obtained here, for example, the posterior probability that an event e out of the five events occurs at a certain point at time t is represented by the following formula (1).

number

[0036] Furthermore, the derivation unit 12 inputs the noise signal N(μ,√μ) at an arbitrary point as an input signal to the machine learning model 13 (step S22). As a result, the derivation unit 12 obtains the posterior probability that each of the five events occurs at the arbitrary point as an output signal of the machine learning model 13 (step S24). Of the posterior probabilities obtained here, for example, the posterior probability that an event e of the five events occurs at a time t at an arbitrary point is represented by the following formula (2).

number

[0037] The comparator 14 compares the posterior probability derived from the observed signal at each point with the posterior probability derived from the noise signal, and finally derives the posterior probability that each of the five events occurs at that point from the comparison result (step S25). Of the posterior probabilities obtained here, for example, the posterior probability that an event e out of the five events occurs at a certain point at time t is expressed as shown in the following formula (3).

number

[0038] Here, it is assumed that the comparison unit 14 derives the posterior probability of the above-mentioned formula (3) by using, for example, the following formula (4) or (5).

number

number

[0039] Thereafter, the comparison unit 14 outputs the posterior probability that each of the five events occurs, which is derived for each point, as an event recognition result to the outside of the event recognition device 10 (step S26).

[0040] In addition, in this operation example 1, the derivation unit 12 uses the machine learning model 13 without re-learning. Therefore, basically, the derivation unit 12 does not perform model learning using observed signals. However, this is not limited to this, and the derivation unit 12 may perform few-shot learning using a small number of observed signals and re-learn the machine learning model 13.

[0041] In addition, when a specific optical noise cannot be uniquely determined, the derivation unit 12 may use noise signals of multiple types of optical noise acquired in advance by the acquisition unit 11, and the comparison unit 14 may compare each of the posterior probabilities of the multiple types of noise signals with the posterior probability of the observed signal to recognize an event.

[0042] Furthermore, since the derivation of the posterior probability is a calculation at the probability level, the posterior probability is always a value between 0 and 1. Therefore, the derivation unit 12 may use each of a plurality of machine learning models 13, and the comparison unit 14 may compare the posterior probabilities derived using each of the plurality of machine learning models 13 to recognize an event.

[0043] Furthermore, the derivation unit 12 and the comparison unit 14 may not derive the posterior probability for all of the five events, and may perform denoising to mask events of any event class without deriving the posterior probability.

[0044] (2) Specific operation example 2 Next, a second specific operation example of the event recognition device 10 according to the first embodiment will be described with reference to FIG. This operation example 2 is an example of a case where a time period during which an event occurs is known at an arbitrary point along an optical fiber. In other words, this operation example 2 is an example of a case where a time period during which optical noise exists at an arbitrary point is known.

[0045] In the case of this operation example 2, the acquisition unit 11 acquires a time domain signal R , which is an observation signal in a time section in which no event occurs, as a noise signal at an arbitrary point from a sensing device such as a DAS device. T or frequency domain signal C T×F Get the. The acquisition unit 11 also receives the above-mentioned time domain signal R as an observation signal at each point from a sensing device such as a DAS device. T or frequency domain signal C T×F Get the.

[0046] As shown in FIG. 5, the derivation unit 12 calculates the observation signal R T Or C T×F is input as an input signal to the machine learning model 13 (step S31). As a result, the derivation unit 12 obtains, for each point, the posterior probability that each of the five events occurs at that point as an output signal of the machine learning model 13 (step S33). Of the posterior probabilities obtained here, for example, the posterior probability that an event e out of the five events occurs at a certain point at time t is represented as shown in the above-mentioned formula (1).

[0047] Furthermore, the derivation unit 12 inputs the noise signal N(μ,√μ) at an arbitrary point as an input signal to the machine learning model 13 (step S32). As a result, the derivation unit 12 obtains the posterior probability that each of the five events occurs at the arbitrary point as an output signal of the machine learning model 13 (step S34). Of the posterior probabilities obtained here, for example, the posterior probability that an event e of the five events occurs at a time t at an arbitrary point is represented as shown in the above-mentioned formula (2).

[0048] The comparator 14 compares the posterior probability derived from the observed signal at each point with the posterior probability derived from the noise signal, and finally derives the posterior probability that each of the five events occurs at that point from the comparison result (step S35). Of the posterior probabilities obtained here, for example, the posterior probability that an event e out of the five events occurs at a certain point at time t is represented as shown in the above-mentioned formula (3). Also, here, it is assumed that the comparison unit 14 derives the posterior probability of the above-mentioned Equation (3) by using, for example, the above-mentioned Equation (4) or Equation (5).

[0049] Thereafter, the comparison unit 14 outputs the posterior probability of occurrence of each of the five events derived for each point as an event recognition result to the outside of the event recognition device 10 (step S36).

[0050] In addition, in this operation example 2, the derivation unit 12 uses the machine learning model 13 without re-learning. Therefore, basically, the derivation unit 12 does not perform model learning using observed signals. However, this is not limited to this, and the derivation unit 12 may perform few-shot learning using a small number of observed signals and re-learn the machine learning model 13.

[0051] Furthermore, the derivation unit 12 and the comparison unit 14 may not derive the posterior probability for all of the five events, and may perform denoising to mask events of any event class without deriving the posterior probability.

[0052] Next, with reference to FIG. 6 and FIG. 7, an example of an event recognition result by the event recognition device 10 according to the first embodiment will be described in comparison with an event recognition result by the related art. 6 and 7 show, from top to bottom, an event recognition result at a certain point by the related art, an event recognition result at the same point by the first embodiment, and a spectrogram at the same point input as an observed signal to the machine learning model 13 when the event recognition is performed. The spectrogram is a signal obtained by dividing the frequency domain signal C T×F is equivalent to.

[0053] 6 and 7, the related technology is assumed to be a technology that outputs the posterior probability derived from the noise signal as it is as an event recognition result without performing a process of deriving the posterior probability from a noise signal indicating optical noise or a process of comparing the posterior probability of the observed signal with that of the noise signal. In other words, in the example of Fig. 3, the related technology is assumed to be a technology that outputs the posterior probability obtained from the observed signal in step S13 as the event recognition result in step S16.

[0054] Also, the spectrograms in FIG. 6 and FIG. 7 are different from each other, and FIG. 6 is referred to as case #1, and FIG. 7 is referred to as case #2. In addition, in Figs. 6 and 7, the horizontal axis of the three figures indicates time, the vertical axis of the upper two figures indicates the event class of each event, and the vertical axis of the lower figure indicates frequency. In addition, in the upper two figures of FIG. 6 and FIG. 7, the whiter the image, the higher the posterior probability, and the darker the image, the lower the posterior probability.

[0055] As shown in Fig. 6, in case #1, the related technology appears darker overall than the first embodiment. This shows that the related technology is affected by optical noise and has clearly poor event recognition performance. In other words, the first embodiment eliminates the effects of optical noise and clearly improves event recognition performance. Moreover, as shown in FIG. 7, case #2 also shows the same tendency as case #1.

[0056] As described above, according to the first embodiment, the acquisition unit 11 acquires an observation signal indicating sound or vibration occurring at a point along the optical fiber and detected by optical fiber sensing, and also acquires a noise signal indicating optical noise present at the point along the optical fiber. The derivation unit 12 derives a posterior probability that at least one event has occurred at a point along the optical fiber from each of the observation signal and the noise signal, using a machine learning model 13 trained using the sound signal or vibration signal acquired by the acoustic sensor or vibration sensor. The comparison unit 14 compares the posterior probability derived from the observation signal with the posterior probability derived from the noise signal, derives a posterior probability that at least one event has occurred at a point along the optical fiber from the comparison result, and outputs the derived posterior probability as an event recognition result.

[0057] In this way, according to the first embodiment, the machine learning model 13 trained using the acoustic signal or vibration signal acquired by the acoustic sensor or vibration sensor is used. This makes it unnecessary to train the model using the observation signal acquired by optical fiber sensing.

[0058] Moreover, according to the first embodiment, the machine learning model 13 is used to derive the posterior probability that an event has occurred at a point along the optical fiber from each of the observed signal and the noise signal. Then, the posterior probability derived from the observed signal is compared with the posterior probability derived from the noise signal, and finally, the posterior probability that an event has occurred is derived from the comparison result. This eliminates the influence of optical noise, thereby improving the event recognition performance.

[0059] <Embodiment 2> The second embodiment corresponds to an embodiment that is a higher-level concept than the first embodiment described above. First, with reference to FIG. 8, a configuration example of the event recognition device 20 according to the second embodiment will be described. As shown in FIG. 8, the event recognition device 20 according to the second embodiment includes an acquisition unit 21, a derivation unit 22, and a comparison unit .

[0060] The acquisition unit 21 acquires observation signals indicative of sound or vibration occurring at points along the optical fiber and detected by optical fiber sensing, and also acquires noise signals indicative of optical noise present at the points along the optical fiber.

[0061] The derivation unit 22 holds statistical information 23 of the acoustic signal or vibration signal acquired by the acoustic sensor or vibration sensor. Then, the derivation unit 22 uses the statistical information 23 to derive a posterior probability that at least one event has occurred at a point along the optical fiber from the observed signal, and to derive a posterior probability that at least one event has occurred at a point along the optical fiber from the noise signal.

[0062] The comparison unit 24 compares the posterior probability derived from the observed signal with the posterior probability derived from the noise signal, and from the comparison result derives the posterior probability that at least one event has occurred at a point along the optical fiber, and outputs the derived result.

[0063] In this way, according to the second embodiment, the statistical information 23 of the acoustic signal or vibration signal acquired by the acoustic sensor or vibration sensor is utilized. This makes it unnecessary to perform model learning using the observation signal acquired by optical fiber sensing.

[0064] Moreover, according to the second embodiment, the statistical information 23 is used to derive the posterior probability that an event has occurred at a point along the optical fiber from each of the observed signal and the noise signal. Then, the posterior probability derived from the observed signal is compared with the posterior probability derived from the noise signal, and finally, the posterior probability that an event has occurred is derived from the comparison result. This eliminates the influence of optical noise, thereby improving the event recognition performance.

[0065] The statistical information 23 may be a machine learning model that outputs a posterior probability that at least one event occurs for an input signal. The machine learning model may be a model trained by an acoustic signal or a vibration signal acquired by an acoustic sensor or a vibration sensor. The derivation unit 22 may input the observation signal as an input signal to the machine learning model to derive the posterior probability from the observation signal, and may input a noise signal as an input signal to the machine learning model to derive the posterior probability from the noise signal. Furthermore, the derivation unit 22 may use the machine learning model without re-learning.

[0066] In addition, in one example, the observed signal may be a signal indicative of sound, and the machine learning model may be a model trained using a sound signal acquired by a microphone as an acoustic sensor.

[0067] In another example, the observed signal may be a signal indicating vibration, and the machine learning model may be a model trained using a vibration signal acquired by an acceleration sensor or a hydrophone as a vibration sensor.

[0068] <Hardware configuration of the event recognition device according to the embodiment> Next, with reference to FIG. 9, an example of a hardware configuration of a computer that realizes the event recognition devices 10 and 20 according to the first and second embodiments will be described.

[0069] 9, a computer 90 includes a processor 91, a memory 92, a storage 93, an input / output interface (input / output I / F) 94, and a communication interface (communication I / F) 95. The processor 91, the memory 92, the storage 93, the input / output interface 94, and the communication interface 95 are connected by a data transmission path for transmitting and receiving data to and from each other.

[0070] The processor 91 is, for example, an arithmetic processing device such as a central processing unit (CPU) or a graphics processing unit (GPU). The memory 92 is, for example, a random access memory (RAM) or a read only memory (ROM). The storage 93 is, for example, a storage device such as a hard disk drive (HDD), a solid state drive (SSD), or a memory card. The storage 93 may also be a memory such as a RAM or a ROM.

[0071] A program is stored in the storage 93. When the program is loaded into a computer, it includes a set of instructions (or software code) for causing the computer 90 to perform one or more functions of the event recognition devices 10 and 20 described above. The components of the event recognition devices 10 and 20 described above may be realized by the processor 91 loading and executing a program stored in the storage 93. Furthermore, the storage function of the event recognition devices 10 and 20 described above may be realized by the memory 92 or the storage 93.

[0072] The above-mentioned programs may also be stored on non-transitory computer-readable media or tangible storage media. By way of example and not limitation, computer-readable media or tangible storage media include RAM, ROM, flash memory, SSD or other memory technology, CD (Compact Disc)-ROM, DVD (Digital Versatile Disc), Blu-ray (registered trademark) disk or other optical disk storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage device. The programs may also be transmitted on a transitory computer-readable medium or communication medium. By way of example and not limitation, transitory computer-readable media or communication media include electrical, optical, acoustic, or other forms of propagated signals.

[0073] The input / output interface 94 is connected to a display device 941, an input device 942, a sound output device 943, and the like. The display device 941 is a device that displays a screen corresponding to drawing data processed by the processor 91, such as an LCD (Liquid Crystal Display), a CRT (Cathode Ray Tube) display, or a monitor. The input device 942 is a device that accepts an operation input by an operator, and is, for example, a keyboard, a mouse, or a touch sensor. The display device 941 and the input device 942 may be integrated and realized as a touch panel. The sound output device 943 is a device that acoustically outputs a sound corresponding to the sound data processed by the processor 91, such as a speaker.

[0074] The communication interface 95 transmits and receives data to and from an external device. For example, the communication interface 95 communicates with the external device via a wired communication path or a wireless communication path.

[0075] Although the present disclosure has been described above with reference to the embodiments, the present disclosure is not limited to the above-described embodiments. Various modifications that can be understood by a person skilled in the art can be made to the configuration and details of the present disclosure within the scope of the present disclosure. Furthermore, each embodiment can be appropriately combined with other embodiments.

[0076] Furthermore, some or all of the above-described embodiments can be described as, but are not limited to, the following supplementary notes. (Appendix 1) an acquisition unit that acquires an observation signal indicative of sound or vibration generated at a point along an optical fiber and detected by optical fiber sensing, and also acquires a noise signal indicative of optical noise present at the point along the optical fiber; a derivation unit that utilizes statistical information of an acoustic signal or a vibration signal acquired by an acoustic sensor or a vibration sensor to derive a posterior probability that at least one event has occurred at a point along the optical fiber from the observed signal, and that derives the posterior probability that the at least one event has occurred at a point along the optical fiber from the noise signal; a comparison unit that compares the posterior probability derived from the observed signal with the posterior probability derived from the noise signal, derives a posterior probability that the at least one event has occurred at a point along the optical fiber from a comparison result, and outputs the derived result; An event recognition device comprising: (Appendix 2) The statistical information is a machine learning model that outputs a posterior probability that the at least one event occurs for an input signal; the machine learning model is a model trained using the acoustic signal or the vibration signal acquired by the acoustic sensor or the vibration sensor, The derivation unit derives a posterior probability from the observed signal by inputting the observed signal to the machine learning model as the input signal, and derives a posterior probability from the noise signal by inputting the noise signal to the machine learning model as the input signal. 2. An event recognition apparatus as claimed in claim 1. (Appendix 3) The derivation unit uses the machine learning model without relearning. 3. An event recognition apparatus as claimed in claim 2. (Appendix 4) the observation signal is a signal indicative of sound; The machine learning model is a model trained by an acoustic signal acquired by a microphone as the acoustic sensor. 3. An event recognition apparatus as claimed in claim 2. (Appendix 5) the observation signal is a signal indicative of vibration; The machine learning model is a model trained using a vibration signal acquired by an acceleration sensor or a hydrophone as the vibration sensor. 3. An event recognition apparatus as claimed in claim 2. (Appendix 6) An event recognition method by an event recognition device, acquiring an observation signal indicative of sound or vibration occurring at a point along the optical fiber and detected by optical fiber sensing, and acquiring a noise signal indicative of optical noise present at the point along the optical fiber; a derivation step of deriving, from the observed signals, a posterior probability that at least one event has occurred at a point along the optical fiber, using statistical information of an acoustic signal or a vibration signal acquired by an acoustic sensor or a vibration sensor, and deriving, from the noise signal, a posterior probability that the at least one event has occurred at a point along the optical fiber; a comparing step of comparing the posterior probability derived from the observed signal with the posterior probability derived from the noise signal, deriving from the comparison result a posterior probability that the at least one event has occurred at a point along the optical fiber, and outputting the derived result; The event recognition method includes: (Appendix 7) The statistical information is a machine learning model that outputs a posterior probability that the at least one event occurs for an input signal; the machine learning model is a model trained using the acoustic signal or the vibration signal acquired by the acoustic sensor or the vibration sensor, In the derivation step, the observation signal is input to the machine learning model as the input signal to derive a posterior probability from the observation signal, and the noise signal is input to the machine learning model as the input signal to derive a posterior probability from the noise signal. 7. The event recognition method according to claim 6. (Appendix 8) In the derivation step, the machine learning model is used without re-learning. 8. The event recognition method according to claim 7. (Appendix 9) the observation signal is a signal indicative of sound; The machine learning model is a model trained by an acoustic signal acquired by a microphone as the acoustic sensor. 8. The event recognition method according to claim 7. (Appendix 10) the observation signal is a signal indicative of vibration; The machine learning model is a model trained using a vibration signal acquired by an acceleration sensor or a hydrophone as the vibration sensor. 8. The event recognition method according to claim 7. (Appendix 11) On the computer, an acquisition step of acquiring an observation signal indicative of sound or vibration occurring at a point along an optical fiber and detected by optical fiber sensing, and acquiring a noise signal indicative of optical noise present at the point along the optical fiber; a derivation step of deriving, from the observed signals, a posterior probability that at least one event has occurred at a point along the optical fiber, using statistical information of an acoustic signal or a vibration signal acquired by an acoustic sensor or a vibration sensor, and deriving, from the noise signal, a posterior probability that the at least one event has occurred at a point along the optical fiber; a comparison step of comparing the posterior probability derived from the observed signal with the posterior probability derived from the noise signal, deriving from the comparison a posterior probability of the at least one event occurring at a point along the optical fiber, and outputting the derived result; A program that executes the following. (Appendix 12) The statistical information is a machine learning model that outputs a posterior probability that the at least one event occurs for an input signal; the machine learning model is a model trained using the acoustic signal or the vibration signal acquired by the acoustic sensor or the vibration sensor, In the derivation step, the observation signal is input to the machine learning model as the input signal to derive a posterior probability from the observation signal, and the noise signal is input to the machine learning model as the input signal to derive a posterior probability from the noise signal. 12. The program according to claim 11. (Appendix 13) In the derivation step, the machine learning model is used without re-learning. 13. The program according to claim 12. (Appendix 14) the observation signal is a signal indicative of sound; The machine learning model is a model trained by an acoustic signal acquired by a microphone as the acoustic sensor. 13. The program according to claim 12. (Appendix 15) the observation signal is a signal indicative of vibration; The machine learning model is a model trained using a vibration signal acquired by an acceleration sensor or a hydrophone as the vibration sensor. 13. The program according to claim 12. [Explanation of symbols]

[0077] 10,20 Event Recognition Device 11,21 Acquisition Department 12,22 Derivation part 13 Machine Learning Models 14,24 Comparison section 23 Statistical information 90 Computer 91 Processor 92 Memory 93 Storage 94 Input / Output Interface 941 Display device 942 Input Device 943 Sound output device 95 Communication Interface

Claims

1. an acquisition unit that acquires an observation signal indicative of sound or vibration generated at a point along an optical fiber and detected by optical fiber sensing, and also acquires a noise signal indicative of optical noise present at the point along the optical fiber; a derivation unit that utilizes statistical information of an acoustic signal or a vibration signal acquired by an acoustic sensor or a vibration sensor to derive a posterior probability that at least one event has occurred at a point along the optical fiber from the observed signal, and that derives the posterior probability that the at least one event has occurred at a point along the optical fiber from the noise signal; a comparison unit that compares the posterior probability derived from the observed signal with the posterior probability derived from the noise signal, derives a posterior probability that the at least one event has occurred at a point along the optical fiber from a comparison result, and outputs the derived result; An event recognition device comprising:

2. the statistical information is a machine learning model that outputs a posterior probability that the at least one event occurs for an input signal; the machine learning model is a model trained using the acoustic signal or the vibration signal acquired by the acoustic sensor or the vibration sensor, The derivation unit derives a posterior probability from the observed signal by inputting the observed signal to the machine learning model as the input signal, and derives a posterior probability from the noise signal by inputting the noise signal to the machine learning model as the input signal. The event recognition device according to claim 1 .

3. The derivation unit uses the machine learning model without relearning. The event recognition device according to claim 2 .

4. the observation signal is a signal indicative of sound; The machine learning model is a model trained by an acoustic signal acquired by a microphone as the acoustic sensor. The event recognition device according to claim 2 .

5. the observation signal is a signal indicative of vibration; The machine learning model is a model trained using a vibration signal acquired by an acceleration sensor or a hydrophone as the vibration sensor. The event recognition device according to claim 2 .

6. An event recognition method by an event recognition device, acquiring an observation signal indicative of sound or vibration occurring at a point along the optical fiber and detected by optical fiber sensing, and acquiring a noise signal indicative of optical noise present at the point along the optical fiber; a derivation step of deriving, from the observed signals, a posterior probability that at least one event has occurred at a point along the optical fiber, using statistical information of an acoustic signal or a vibration signal acquired by an acoustic sensor or a vibration sensor, and deriving, from the noise signal, a posterior probability that the at least one event has occurred at a point along the optical fiber; a comparing step of comparing the posterior probability derived from the observed signal with the posterior probability derived from the noise signal, deriving a posterior probability of the at least one event occurring at a point along the optical fiber from the comparison result, and outputting the derived result; The event recognition method includes:

7. the statistical information is a machine learning model that outputs a posterior probability that the at least one event occurs for an input signal; the machine learning model is a model trained using the acoustic signal or the vibration signal acquired by the acoustic sensor or the vibration sensor, In the derivation step, the observation signal is input to the machine learning model as the input signal to derive a posterior probability from the observation signal, and the noise signal is input to the machine learning model as the input signal to derive a posterior probability from the noise signal. The event recognition method according to claim 6.

8. In the derivation step, the machine learning model is used without re-learning. The event recognition method according to claim 7.

9. the observation signal is a signal indicative of sound; The machine learning model is a model trained by an acoustic signal acquired by a microphone as the acoustic sensor. The event recognition method according to claim 7.

10. the observation signal is a signal indicative of vibration; The machine learning model is a model trained using a vibration signal acquired by an acceleration sensor or a hydrophone as the vibration sensor. The event recognition method according to claim 7.

11. On the computer, an acquisition step of acquiring an observation signal indicative of sound or vibration occurring at a point along an optical fiber and detected by optical fiber sensing, and acquiring a noise signal indicative of optical noise present at the point along the optical fiber; a derivation step of deriving, from the observed signals, a posterior probability that at least one event has occurred at a point along the optical fiber, using statistical information of an acoustic signal or a vibration signal acquired by an acoustic sensor or a vibration sensor, and deriving, from the noise signal, a posterior probability that the at least one event has occurred at a point along the optical fiber; a comparison step of comparing the posterior probability derived from the observed signal with the posterior probability derived from the noise signal, deriving from the comparison a posterior probability of the at least one event occurring at a point along the optical fiber, and outputting the deriving result; A program that executes the following.

12. the statistical information is a machine learning model that outputs a posterior probability that the at least one event occurs for an input signal; the machine learning model is a model trained using the acoustic signal or the vibration signal acquired by the acoustic sensor or the vibration sensor, In the derivation step, the observation signal is input to the machine learning model as the input signal to derive a posterior probability from the observation signal, and the noise signal is input to the machine learning model as the input signal to derive a posterior probability from the noise signal. The program according to claim 11.

13. In the derivation step, the machine learning model is used without re-learning. The program according to claim 12.

14. the observation signal is a signal indicative of sound; The machine learning model is a model trained by an acoustic signal acquired by a microphone as the acoustic sensor. The program according to claim 12.

15. the observation signal is a signal indicative of vibration; The machine learning model is a model trained using a vibration signal acquired by an acceleration sensor or a hydrophone as the vibration sensor. The program according to claim 12.

Images