Determination device

Abstract

This determination device comprises: a signal determination unit that determines the presence or absence of a signal for each prescribed frequency domain and time domain by using a feature amount extracted from a received signal; a signal bandwidth comparison unit that, in accordance with determination results from the signal determination unit, estimates a signal bandwidth, which is feature information not dependent on the absolute value of received electric power, and compares the estimated signal bandwidth with a pretrained signal bandwidth at normal time; and an abnormality determination unit that determines the presence or absence of an abnormality in accordance with the result of comparison by the signal bandwidth comparison unit.

Description

Judgment device

[0001] The present disclosure relates to a determination device, a determination method, a recording medium, and a learning device.

[0002] 2. Description of the Related Art In wireless communication, techniques for determining abnormalities such as radio wave interference are known.

[0003] For example, Patent Document 1 discloses a radio wave anomaly detection system that detects radio wave anomalies contained in received data. According to Patent Document 1, the radio wave anomaly detection system includes a first feature extraction unit that extracts a feature of the received data for each first predetermined period as a first feature, and a second feature extraction unit that extracts a feature of the received data for each second predetermined period that is longer than the first predetermined period as a second feature. The radio wave anomaly detection system detects radio wave anomalies contained in the received data using the first feature extracted by the first feature extraction unit and the second feature extracted by the second feature extraction unit.

[0004] Furthermore, as a technique for detecting abnormalities such as radio wave interference, for example, a technique called spectrum masking is known, which sets a threshold value for the reception level of each frequency of the received radio wave.

[0005] JP 2022-182844 A

[0006] The technology described in Patent Document 1 and the technology using a spectrum mask perform anomaly detection based on information on the absolute value of received power. Therefore, when a mobile sensor is used, for example, and the radio wave environment fluctuates, the detection accuracy may decrease. As a result, there is a problem that it may be difficult to properly detect anomalies such as radio wave interference.

[0007] Therefore, one object of the present disclosure is to provide a determination device, a determination method, a recording medium, and a learning device that can solve the above-mentioned problems.

[0008] In order to achieve this object, the determination device of the present disclosure has a configuration including: a signal determination unit that determines the presence or absence of a signal for each predetermined frequency domain and time domain using features extracted from a received signal; a signal bandwidth comparison unit that estimates a signal bandwidth, which is feature information that does not depend on the absolute value of the received power, according to the determination result by the signal determination unit, and compares the estimated signal bandwidth with a normal signal bandwidth that has been learned in advance; and an abnormality determination unit that determines the presence or absence of an abnormality according to the comparison result by the signal bandwidth comparison unit.

[0009] Furthermore, the determination method in the present disclosure is configured such that an information processing device determines the presence or absence of a signal for each predetermined frequency domain and time domain using features extracted from the received signal, estimates the signal bandwidth, which is feature information that does not depend on the absolute value of the received power, according to the determination result, compares the estimated signal bandwidth with a normal signal bandwidth learned in advance, and determines the presence or absence of an abnormality according to the comparison result.

[0010] Furthermore, the recording medium in the present disclosure is a computer-readable recording medium having recorded thereon a program for causing an information processing device to perform the following processes: determine the presence or absence of a signal for each predetermined frequency domain and time domain using features extracted from the received signal; estimate the signal bandwidth, which is feature information that does not depend on the absolute value of the received power, based on the determination result; compare the estimated signal bandwidth with a normal signal bandwidth learned in advance; and determine the presence or absence of an abnormality based on the comparison result.

[0011] Furthermore, the learning device in the present disclosure has a configuration including: a signal determination unit that determines the presence or absence of a signal for each predetermined frequency domain and time domain using features extracted from a received signal; and a signal bandwidth learning unit that estimates and learns a signal bandwidth, which is feature information that does not depend on the absolute value of the received power, according to the determination result by the signal determination unit.

[0012] According to the above-described configurations, abnormalities such as radio wave interference can be appropriately determined.

[0013] 1 is a diagram illustrating an example of a configuration of an abnormality determination system in the present disclosure. FIG. 2 is a block diagram illustrating an example of a configuration of a learning device. FIG. 3 is a diagram for explaining an example of a signal determination process. FIG. 4 is a diagram for explaining an example of a signal determination process. FIG. 5 is a diagram for explaining an example of a continuous transmission time learning process. FIG. 6 is a diagram for explaining an example of a continuous transmission time learning process. FIG. 7 is a diagram for explaining an example of a signal bandwidth learning process. FIG. 8 is a diagram for explaining an example of a continuous bandwidth learning process. FIG. 9 is a block diagram illustrating an example of a configuration of a determination device. FIG. 10 is a flowchart illustrating an example of an operation of the learning device. FIG. 11 is a flowchart illustrating an example of an operation of the determination device. FIG. 12 is a block diagram illustrating another example of a configuration of the learning device. FIG. 13 is a block diagram illustrating another example of a configuration of the determination device. FIG. 14 is a diagram illustrating an example of a hardware configuration of a second determination device in the present disclosure. FIG. 15 is a block diagram illustrating an example of a configuration of the determination device. FIG. 16 is a flowchart illustrating an example of an operation of the determination device. FIG. 17 is a block diagram illustrating an example of a configuration of the second learning device.

[0014] [First Embodiment] An example configuration of an abnormality determination system 100 according to the present disclosure will be described with reference to FIGS. 1 to 14. FIG. 1 is a diagram illustrating an example configuration of the abnormality determination system 100. FIG. 2 is a block diagram illustrating an example configuration of a learning device 200. FIGS. 3 and 4 are diagrams illustrating an example of a signal determination process. FIGS. 5 and 6 are diagrams illustrating an example of a continuous transmission time learning process. FIG. 7 is a diagram illustrating an example of a signal bandwidth learning process. FIGS. 8 and 9 are diagrams illustrating an example of a continuous bandwidth learning process. FIG. 10 is a block diagram illustrating an example configuration of a determination device 300. FIG. 11 is a flowchart illustrating an example operation of the learning device 200. FIG. 12 is a flowchart illustrating an example operation of the determination device 300. FIG. 13 is a block diagram illustrating another example configuration of the learning device 200. FIG. 14 is a block diagram illustrating another example configuration of the determination device 300. Note that in the present disclosure, the drawings may be associated with one or more embodiments.

[0015] In a first embodiment of the present disclosure, an abnormality determination system 100 that determines an abnormality such as radio wave interference occurring in wireless communication is described. As described below, the abnormality determination system 100 generates a signal presence / absence map of a wireless signal that is independent of the absolute value of the received power (reception level) in two domains, time and frequency, according to the received signal to be determined. For example, the abnormality determination system 100 generates the signal presence / absence map by determining the presence or absence of a signal at each time and frequency using statistical features such as a PDF (Probability Density Function) or a CDF (Cumulative Distribution Function) extracted from the received signal. Furthermore, the abnormality determination system 100 uses the generated signal presence / absence map to acquire radio wave characteristic information that is independent of the absolute value of the received power, such as the continuous transmission time and signal bandwidth. The abnormality determination system 100 then determines an abnormality such as radio wave interference by comparing the acquired radio wave characteristic information with pre-stored radio wave characteristic information during normal operation. For example, the abnormality determination system 100 can determine an abnormality when the acquired radio wave characteristic information differs from pre-stored radio wave characteristic information during normal times.

[0016] The frequency domain and time domain resolutions of the signal presence / absence map generated by the anomaly determination system 100 may be adjusted as desired. That is, the resolution of each grid in the two-dimensional signal presence / absence map may be adjusted as desired. For example, if the time domain is set to a short time, such as on the order of ns / us, there is a risk that environmental fluctuations will increase due to the effects of fading. Therefore, the time domain resolution may be set based on environmental information, such as the installation interval of the determination device 300, the distance between the transmitter and receiver, and the frequency domain resolution, so as to suppress (average) the effects of fading. As an example, the time domain resolution may be on the order of ms. The frequency domain resolution may also be adjusted as desired based on environmental information, etc.

[0017] Furthermore, in the present disclosure, radio wave characteristic information refers to information indicating characteristics of radio waves that can be determined based on the determination result of the presence or absence of a signal for each frequency and time. For example, the radio wave characteristic information includes at least some of the following: a continuous transmission time indicating the time during which a signal continues to be present at each frequency, a signal bandwidth indicating the bandwidth of the signal, and a continuous bandwidth indicating the bandwidth that continues for a predetermined period of time, such as during the continuous transmission time. The radio wave characteristic information may also include information indicating characteristics other than those exemplified above that can be determined based on the determination result of the presence or absence of a signal. Of the radio wave characteristic information exemplified above, the continuous transmission time and signal bandwidth are one-dimensional information, and the continuous bandwidth is two-dimensional information.

[0018] Fig. 1 shows an example of the configuration of an abnormality determination system 100. Referring to Fig. 1, the abnormality determination system 100 includes, for example, a learning device 200 and a determination device 300. As shown in Fig. 1, the learning device 200 and the determination device 300 can be connected to each other via a wired or wireless connection so that they can communicate with each other. Note that the learning device 200 and the determination device 300 do not necessarily need to be directly connected, and data such as learning results may be exchanged via a recording medium or the like.

[0019] The learning device 200 is a device that learns radio wave characteristic information under normal circumstances. For example, the learning device 200 is a portable mobile sensor. Therefore, the radio wave environment may vary depending on the location where the learning device 200 is installed. Note that the learning device 200 may be a fixed sensor, an edge processing device, a computing device, a cloud, or the like. In addition, in the present disclosure, "normal circumstances" refers to communication conditions expected under normal circumstances.

[0020] FIG. 2 shows an example configuration of a learning device 200. Referring to FIG. 2, the learning device 200 includes an antenna 210, a receiving unit 220, a feature extraction unit 230, and a learning unit 240. For example, the learning device 200 includes a computation device such as a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), or a central processing unit (CPU), and a storage device. The learning device 200 can realize each of the above-described processing units by having the computation device execute a program stored in the storage device. Note that the computation device may include a graphic processing unit (GPU), a digital signal processor (DSP), a micro processing unit (MPU), a floating point number processing unit (FPU), a physics processing unit (PPU), a tensor processing unit (TPU), a quantum processor, a microcontroller, or a combination thereof, instead of the CPU.

[0021] The antenna 210 receives the received signal. The antenna 210 may be a general antenna that receives radio waves and converts the received radio waves into electrical signals. The receiving unit 220 receives the received signal from the antenna 210. The receiving unit 220 may have a function of amplifying the signal, a filter function, etc.

[0022] The feature extraction unit 230 extracts statistical feature quantities, such as PDF or CDF relating to the reception level for each frequency, from the received signal. For example, as shown in Fig. 2, the feature extraction unit 230 is composed of a spectrogram generation unit 231, a signal extraction unit 232, and a feature generation unit 233, and extracts statistical feature quantities from the received signal using each processing unit.

[0023] The spectrogram generating unit 231 generates a spectrogram representing time, frequency, signal level, etc. from the received signal by performing spectrogram generating processing on the received signal received by the receiving unit 220. The spectrogram generating unit 231 may perform the spectrogram generating processing using a general method such as a short-time Fourier transform.

[0024] The signal extraction unit 232 extracts, from the spectrogram generated by the spectrogram generation unit, a signal region, a reception level, etc. of a transmission signal that is included in the received signal and that has been transmitted from an arbitrary wireless device, a radio wave interference source, etc. The signal extraction unit 232 may perform the extraction using a noise level threshold, etc.

[0025] The feature quantification unit 233 generates statistical features using the results extracted by the signal extraction unit 232. For example, the feature quantification unit 233 generates at least one of statistical features such as PDF and CDF. Note that the feature quantification unit 233 may generate, as the statistical feature, statistical information over a certain period of time, such as a frequency distribution (histogram) relating to the amount of signal information, such as amplitude, phase, IQ information, or power.

[0026] For example, the feature quantity generating unit 233 generates a histogram for each reception level extracted for each predetermined period, and generates a PDF by converting the frequency unit into a function as a probability density instead of an absolute value. The feature quantity generating unit 233 can also generate a CDF by converting the generated histogram into a cumulative probability. The feature quantity generating unit 233 may generate statistical features using a method other than those exemplified above.

[0027] The learning unit 240 estimates and learns radio wave characteristic information that is independent of the absolute value of the received power, such as the continuous transmission time and signal bandwidth, based on the result of determining whether or not a signal is present using the statistical characteristic extracted by the characteristic extraction unit 230. For example, the learning unit 240 generates a signal presence / absence map indicating the presence or absence of a signal for each time and frequency, by using the statistical characteristic extracted by the characteristic extraction unit 230. Then, the learning unit 240 uses the generated signal presence / absence map to learn radio wave characteristic information, such as the continuous transmission time and signal bandwidth, during normal operation. Referring to FIG. 2 , the learning unit 240 includes a signal determination unit 241, an output time learning unit 242, a signal bandwidth learning unit 243, a continuous bandwidth learning unit 244, and an output unit 245.

[0028] The signal determination unit 241 generates a signal presence / absence map indicating the presence / absence of a signal for each time and frequency by determining the presence / absence of a signal using the statistical feature extracted by the feature extraction unit 230. For example, the signal determination unit 241 can generate the signal presence / absence map by determining the presence / absence of a signal for each time and frequency using a boundary estimated using the statistical feature extracted by the feature extraction unit 230.

[0029] Fig. 3 shows an example of a signal presence / absence map generated by the signal determination unit 241. As shown in Fig. 3, the signal determination unit 241 determines the presence or absence of a signal at each time and each frequency, thereby generating a signal presence / absence map indicating the presence or absence of a signal at each time and frequency. For example, in Fig. 3, black squares indicate that a signal has been determined to be present, and white squares indicate that a signal has been determined to be absent. As described above, the resolution of each grid in the two-dimensional signal presence / absence map may be adjusted as desired.

[0030] For example, the signal determination unit 241 performs the following process to determine the presence or absence of a signal using statistical features and generate a signal presence / absence map. For example, the signal determination unit 241 generates a distribution of reception levels for each frequency using statistical features. Furthermore, the signal determination unit 241 uses the generated results to estimate the boundary between the noise level and the signal level in the distribution of reception levels. For example, as shown in FIG. 4 , the signal determination unit 241 determines that the peak on the distribution side with the lowest reception level corresponds to the noise level. Then, the signal determination unit 241 estimates the right foot of the peak corresponding to the noise level, such as the first minimum value following the peak corresponding to the noise level, as the boundary. The signal determination unit 241 may also cluster the entire distribution of reception levels and estimate the boundary between the distribution cluster with the lowest reception level and the distribution cluster with the next lowest reception level. Then, the signal determination unit 241 determines the presence or absence of a signal depending on whether the reception level of a corresponding frequency at a certain time is equal to or greater than the estimated boundary. For example, the signal determination unit 241 can generate a signal presence / absence map by performing the above-described determination at each time and each frequency.

[0031] The output time learning unit 242 learns the continuous transmission time, which is radio wave characteristic information, according to the signal presence / absence determination result by the signal determination unit 241. The output time learning unit 242 can learn the continuous transmission time for each frequency by referring to the signal presence / absence map and checking the signal presence / absence information in the continuous time domain for each frequency.

[0032] For example, as shown in FIG. 5 , the output time learning unit 242 extracts a duration, which is the time a signal lasts, for each frequency bin. Then, the output time learning unit 242 estimates and learns a continuous transmission time to be used as radio wave characteristic information based on the extracted results. For example, as shown in FIG. 6 , the output time learning unit 242 generates a frequency map indicating the number of times each duration occurs for each frequency bin based on the extracted results. Then, the output time learning unit 242 estimates a continuous transmission time to be used as radio wave characteristic information using the generated frequency map. As an example, the output time learning unit 242 extracts the maximum value from the generated frequency map or performs probabilistic estimation from a distribution, thereby estimating the extracted or estimated value as the continuous transmission time. In addition to the above example, the output time learning unit 242 may extract any statistical value, such as a mode or a median. Furthermore, the output time learning unit 242 may estimate a single value or multiple values ​​as the continuous transmission time. The output time learning unit 242 may estimate, as the continuous transmission time, a likelihood value for each continuous transmission time corresponding to the frequency distribution based on the frequency map.

[0033] The output time learning unit 242 performs the above-described learning for each frequency bin, thereby estimating and learning the continuous transmission time used as radio wave characteristic information for each frequency.

[0034] The signal bandwidth learning unit 243 learns the signal bandwidth, which is radio wave characteristic information, in accordance with the signal presence / absence determination result by the signal determination unit 241. The signal bandwidth learning unit 243 can learn the signal bandwidth by referring to a signal presence / absence map and checking information on the presence / absence of a signal in adjacent frequency bins.

[0035] For example, as shown in FIG. 7 , the signal bandwidth learning unit 243 checks information on the presence or absence of signals in adjacent frequency bins and extracts a bandwidth. Then, the signal bandwidth learning unit 243 estimates and learns a signal bandwidth to be used as radio wave characteristic information based on the extracted results. For example, similar to the output time learning unit 242, the signal bandwidth learning unit 243 generates a frequency map indicating the number of times each bandwidth occurs for each frequency bin. Then, the signal bandwidth learning unit 243 uses the generated frequency map to estimate a signal bandwidth to be used as radio wave characteristic information. Similar to the output time learning unit 242, the signal bandwidth learning unit 243 may estimate the extracted or estimated value as the signal bandwidth by extracting the maximum value from the generated frequency map or by performing probabilistic estimation from a distribution. Note that, in addition to the above-mentioned example, the signal bandwidth learning unit 243 may extract any statistical value, such as a mode or a median. Furthermore, the signal bandwidth learning unit 243 may estimate one value or multiple values ​​as the signal bandwidth. The signal bandwidth learning unit 243 may estimate, as the signal bandwidth, a likelihood value for each signal bandwidth corresponding to the frequency distribution based on the frequency map.

[0036] The continuous bandwidth learning unit 244 learns the continuous bandwidth, which is radio wave characteristic information, in accordance with the signal presence / absence determination result by the signal determination unit 241. The continuous bandwidth learning unit 244 can learn the continuous bandwidth by referring to the signal presence / absence map and checking the bandwidth that continues for a predetermined period of time. For example, the continuous bandwidth learning unit 244 may learn the continuous bandwidth by checking the bandwidth that continues for the continuous transmission time learned by the output time learning unit 242.

[0037] For example, as shown in FIG. 8 , the continuous bandwidth learning unit 244 extracts candidate continuous bandwidths by checking the bandwidths that continue during the continuous transmission time learned by the output time learning unit 242. Then, the continuous bandwidth learning unit 244 estimates and learns the continuous bandwidths to be used as radio wave characteristic information based on the extracted results. For example, similar to the output time learning unit 242, the continuous bandwidth learning unit 244 generates a frequency map indicating the number of times each candidate continuous bandwidth occurs. Then, the continuous bandwidth learning unit 244 uses the generated frequency map to estimate the continuous bandwidth to be used as radio wave characteristic information. Similar to the output time learning unit 242, the continuous bandwidth learning unit 244 may extract the maximum value from the generated frequency map or perform probabilistic estimation from a distribution to estimate the extracted or estimated value as the continuous bandwidth. Note that, in addition to the above example, the continuous bandwidth learning unit 244 may extract any statistical value, such as a mode or a median. Furthermore, the continuous bandwidth learning unit 244 may estimate one value or multiple values ​​as the continuous bandwidth. The continuous bandwidth learning unit 244 may estimate, as the continuous bandwidth, a likelihood value for each continuous bandwidth candidate corresponding to the frequency distribution based on the frequency map.

[0038] As shown in FIG. 9 , some frequency bins may have peaks or valleys in the signal presence / absence map. In such cases, the continuous bandwidth learning unit 244 may use any method to extract candidate continuous bandwidths. For example, the continuous bandwidth learning unit 244 may extract only portions of the signal presence / absence map that do not form peaks or valleys, such as by extracting only bandwidths that are constantly continuous for a predetermined period of time. The continuous bandwidth learning unit 244 may extract the continuous bandwidth so as to include portions that form peaks or valleys in the signal presence / absence map, such as by extracting bandwidths that are included at least a predetermined percentage of the predetermined period of time.

[0039] The output unit 245 outputs to an external device the learning results of the learning unit 240. For example, the output unit 245 can output the continuous transmission time learned by the output time learning unit 242, the signal bandwidth learned by the signal bandwidth learning unit 243, the continuous bandwidth learned by the continuous bandwidth learning unit 244, and the like to an external device such as the determination device 300.

[0040] The above is an example of the configuration of the learning device 200. Note that the configuration of the learning device 200 is not limited to the example shown in FIG. 2 . For example, the learning device 200 may not have an antenna 210 or the like. In this case, the learning device 200 may be configured to acquire a received signal from an external device or the like and perform learning using the acquired received signal. Furthermore, the learning unit 240 may be configured with a part of the above-described configuration, such as including only the signal bandwidth learning unit 243 without including the output time learning unit 242 or the continuous bandwidth learning unit 244.

[0041] The determination device 300 is a device that determines whether or not there is an abnormality such as radio wave interference, using the results of learning by the learning device 200. For example, the determination device 300 is a mobile sensor, similar to the learning device 200. Note that the determination device 300 may also be a fixed sensor.

[0042] FIG. 10 shows an example configuration of a determination device 300. Referring to FIG. 10, the determination device 300 includes an antenna 310, a receiving unit 320, a feature extraction unit 330, and a determination unit 340. For example, the determination device 300 includes a calculation device such as an FPGA, an ASIC, or a CPU, and a storage device. The determination device 300 can realize each of the above-described processing units by having the calculation device execute a program stored in the storage device. Note that, as in the case of the learning device 200, the calculation device may include a GPU or the like instead of the above-described CPU.

[0043] The configurations of the antenna 310 and the receiving unit 320 may be similar to those of the antenna 210 and the receiving unit 220 of the learning device 200. The configuration of the feature extraction unit 330 may also be similar to that of the feature extraction unit 230. As shown in Fig. 10 , the feature extraction unit 330 is composed of a spectrogram generation unit 331, a signal extraction unit 332, and a feature generation unit 333, and can extract statistical features from the received signal using these processing units.

[0044] The determination unit 340 determines whether or not there is an abnormality such as radio wave interference using the results of learning by the learning device 200. For example, the determination unit 340 acquires and stores radio wave characteristic information under normal conditions, which is the result of learning by the learning device 200, from the learning device 200 in advance. Furthermore, the determination unit 340 acquires radio wave characteristic information corresponding to the extracted feature amounts in response to the extraction of statistical feature amounts by the feature amount extraction unit 330. The determination unit 340 then compares the acquired radio wave characteristic information with the pre-stored radio wave characteristic information under normal conditions, and determines whether or not there is an abnormality such as radio wave interference based on the comparison result. Referring to FIG. 10 , the determination unit 340 includes a signal determination unit 341, an output time comparison unit 342, a signal bandwidth comparison unit 343, a continuous bandwidth comparison unit 344, an abnormality determination unit 345, and an output unit 346.

[0045] The signal determination unit 341 generates a signal presence / absence map indicating the presence / absence of a signal for each time and frequency by determining the presence / absence of a signal using the statistical feature extracted by the feature extraction unit 330. The signal determination unit 341 may generate the signal presence / absence map using a method similar to that of the signal determination unit 241 included in the learning device 200.

[0046] The output time comparison unit 342 estimates the continuous transmission time, which is radio wave characteristic information, according to the result of the signal presence / absence determination by the signal determination unit 341. Then, the output time comparison unit 342 compares the estimated continuous transmission time with the normal continuous transmission time acquired from the learning device 200 and stored.

[0047] For example, the output time comparison unit 342 estimates the continuous transmission time for each frequency bin by using a method similar to that of the output time learning unit 242 of the learning device 200. As an example, the output time comparison unit 342 references a signal presence / absence map and extracts, for each frequency bin, a duration during which a signal is present. Then, the output time comparison unit 342 estimates the continuous transmission time to be used as radio wave characteristic information based on the extracted results. For example, the output time comparison unit 342 generates a frequency map indicating the number of times each duration occurs based on the extracted results. Then, the output time comparison unit 342 uses the generated frequency map to estimate the continuous transmission time to be used as radio wave characteristic information.

[0048] Furthermore, the output time comparison unit 342 compares the estimated continuous transmission time with the normal continuous transmission time. For example, the output time comparison unit 342 performs the comparison to check whether the estimated continuous transmission time matches the normal continuous transmission time. The output time comparison unit 342 may also check whether the estimated continuous transmission time is equivalent to or does not deviate from the normal continuous transmission time. Note that the range of values ​​that are considered equivalent may be set arbitrarily.

[0049] The signal bandwidth comparison unit 343 estimates the signal bandwidth, which is radio wave characteristic information, in accordance with the result of the signal presence / absence determination by the signal determination unit 341. Then, the signal bandwidth comparison unit 343 compares the estimated signal bandwidth with the normal signal bandwidth acquired from the learning device 200 and stored.

[0050] For example, the signal bandwidth comparison unit 343 estimates the signal bandwidth using a method similar to that of the signal bandwidth learning unit 243 of the learning device 200. As an example, the signal bandwidth comparison unit 343 extracts the bandwidth by, for example, referring to a signal presence / absence map and checking information on the presence / absence of a signal in adjacent frequency bins. Then, the signal bandwidth comparison unit 343 estimates the signal bandwidth to be used as the radio wave characteristic information based on the extracted result. For example, the signal bandwidth comparison unit 343 generates a frequency map indicating the number of times each bandwidth occurs. Then, the signal bandwidth comparison unit 343 uses the generated frequency map to estimate the signal bandwidth to be used as the radio wave characteristic information.

[0051] Furthermore, the signal bandwidth comparison unit 343 compares the estimated signal bandwidth with the normal signal bandwidth. For example, by performing the comparison, the signal bandwidth comparison unit 343 checks whether the estimated signal bandwidth matches the normal signal bandwidth. The signal bandwidth comparison unit 343 may also check whether the estimated signal bandwidth is equivalent to or does not deviate from the normal signal bandwidth. Note that the range of values ​​that are considered equivalent may be set arbitrarily.

[0052] The continuous bandwidth comparison unit 344 estimates the continuous bandwidth, which is radio wave characteristic information, in accordance with the result of the signal presence / absence determination by the signal determination unit 341. Then, the continuous bandwidth comparison unit 344 compares the estimated continuous bandwidth with the normal continuous bandwidth acquired from the learning device 200 and stored.

[0053] For example, the continuous bandwidth comparison unit 344 estimates the continuous bandwidth using a method similar to that used by the continuous bandwidth learning unit 244 of the learning device 200. As an example, the continuous bandwidth comparison unit 344 extracts candidate continuous bandwidths by, for example, referring to a signal presence / absence map and confirming the bandwidth that continues during the continuous transmission time estimated by the output time comparison unit 342. The continuous bandwidth comparison unit 344 may also extract candidate continuous bandwidths other than those exemplified above, such as by confirming the bandwidth that continues during the continuous transmission time estimated by the output time learning unit 242. Then, the continuous bandwidth comparison unit 344 estimates the continuous bandwidth to be used as radio wave characteristic information based on the extracted results. For example, the continuous bandwidth comparison unit 344 generates a frequency map indicating the number of times each candidate continuous bandwidth occurs. Then, the continuous bandwidth comparison unit 344 uses the generated frequency map to estimate the continuous bandwidth to be used as radio wave characteristic information.

[0054] Furthermore, the continuous bandwidth comparison unit 344 compares the estimated continuous bandwidth with the continuous bandwidth in normal operation. For example, the continuous bandwidth comparison unit 344 performs the comparison to check whether the estimated continuous bandwidth and the continuous bandwidth in normal operation match. The continuous bandwidth comparison unit 344 may also check whether the estimated continuous bandwidth is equivalent to or does not deviate from the continuous bandwidth in normal operation. Note that the range to be considered equivalent may be set arbitrarily.

[0055] The abnormality determination unit 345 determines whether or not there is an abnormality such as radio wave interference, depending on the comparison results by each comparison unit. The abnormality determination unit 345 can determine whether or not there is an abnormality, depending on whether or not the comparison results by each comparison unit satisfy a predetermined condition.

[0056] For example, the abnormality determination unit 345 acquires comparison results from each of the output time comparison unit 342, the signal bandwidth comparison unit 343, and the continuous bandwidth comparison unit 344. Then, the abnormality determination unit 345 determines whether or not there is an abnormality depending on whether at least one of the comparison units among the output time comparison unit 342, the signal bandwidth comparison unit 343, and the continuous bandwidth comparison unit 344 confirms that there is no match. For example, the abnormality determination unit 345 determines that there is an abnormality when at least one of the comparison units among the output time comparison unit 342, the signal bandwidth comparison unit 343, and the continuous bandwidth comparison unit 344 confirms that there is no match. On the other hand, when each of the output time comparison unit 342, the signal bandwidth comparison unit 343, and the continuous bandwidth comparison unit 344 confirms that there is a match, the abnormality determination unit 345 determines that there is no abnormality.

[0057] For example, as described above, the abnormality determination unit 345 can determine whether or not an abnormality exists based on the individual comparison results of each comparison unit. Note that the abnormality determination unit 345 may determine whether or not an abnormality exists by combining the comparison results of each comparison unit. For example, the abnormality determination unit 345 may be configured to determine whether or not an abnormality exists when the signal bandwidth comparison unit 343 confirms that there is no match and the continuous bandwidth comparison unit 344 also confirms that there is no match. The abnormality determination unit 345 may also determine whether or not an abnormality exists based on combinations other than those exemplified above.

[0058] The output unit 346 outputs the result of the determination by the abnormality determination unit 345 to an external device, etc. The output unit 346 may output the result of the determination as well as the results of the comparison by each comparison unit, feature information used in the comparison, frequency bin information, a signal presence / absence map, etc. to an external device, etc. In other words, the output unit 346 can output information indicating the cause of the abnormality in addition to information indicating the presence or absence of an abnormality.

[0059] The above is an example of the configuration of the determination device 300. Note that the determination device 300 may have the same modified examples as the learning device 200.

[0060] For example, the abnormality determination system 100 includes the learning device 200 and the determination device 300 having the above-described configuration. Note that the abnormality determination system 100 may be configured from a single device that has both the functions of the learning device 200 and the determination device 300. Furthermore, the learning device 200 and the determination device 300 may each be configured from a plurality of devices.

[0061] 11 and 12, the operations of the learning device 200 and the determination device 300 that constitute the abnormality determination system 100 will be described. First, an example of the operation of the learning device 200 will be described with reference to FIG.

[0062] 11 is a flowchart showing an example of the operation of the learning device 200. Referring to FIG. 11, the feature extraction unit 230 extracts statistical features, such as PDF or CDF relating to the reception level for each frequency, from the received signal (step S101). For example, the feature extraction unit 230 is composed of a spectrogram generation unit 231, a signal extraction unit 232, and a feature generation unit 233, and extracts statistical features from the received signal using these processing units.

[0063] The signal determination unit 241 determines the presence or absence of a signal using the statistical feature extracted by the feature extraction unit 230 (step S102). The signal determination unit 241 can generate a signal presence / absence map indicating the presence or absence of a signal by determining the presence or absence of a signal for each time and each frequency.

[0064] The output time learning unit 242 learns the continuous transmission time, which is radio wave characteristic information, in accordance with the signal presence / absence determination result by the signal determination unit 241 (step S103). The output time learning unit 242 can learn the continuous transmission time for each frequency by referring to the signal presence / absence map and checking the signal presence / absence information in the continuous time domain for each frequency.

[0065] The signal bandwidth learning unit 243 learns the signal bandwidth, which is radio wave characteristic information, in accordance with the signal presence / absence determination result by the signal determination unit 241 (step S104). The signal bandwidth learning unit 243 can learn the signal bandwidth for each frequency by referring to the signal presence / absence map and checking the signal presence / absence information in frequency bins adjacent to each frequency.

[0066] The continuous bandwidth learning unit 244 learns the continuous bandwidth, which is radio wave characteristic information, in accordance with the signal presence / absence determination result by the signal determination unit 241 (step S105). The continuous bandwidth learning unit 244 can learn the continuous bandwidth by referring to the signal presence / absence map and checking the bandwidth that continues for a predetermined period of time. For example, the continuous bandwidth learning unit 244 may learn the continuous bandwidth by checking the bandwidth that continues for the continuous transmission time learned by the output time learning unit 242.

[0067] The output unit 245 outputs the learning results of the learning unit 240 to an external device (step S106). For example, the output unit 245 outputs the continuous transmission time learned by the output time learning unit 242, the signal bandwidth learned by the signal bandwidth learning unit 243, the continuous bandwidth learned by the continuous bandwidth learning unit 244, and the like to an external device such as the determination device 300.

[0068] The above is an example of the operation of the learning device 200. Note that the order of the processes of steps S103, S104, and S105 may be other than that illustrated in Fig. 11. Next, an example of the operation of the determination device 300 will be described with reference to Fig. 12.

[0069] 12 is a flowchart showing an example of the operation of the determination device 300. Referring to FIG. 12, the feature extraction unit 330 extracts statistical feature quantities, such as PDF or CDF relating to the reception level for each frequency, from the received signal (step S201). For example, the feature extraction unit 330 is composed of a spectrogram generation unit 331, a signal extraction unit 332, and a feature generation unit 333, and extracts statistical feature quantities from the received signal using these processing units.

[0070] The signal determination unit 341 determines the presence or absence of a signal using the statistical feature extracted by the feature extraction unit 330 (step S202). The signal determination unit 341 can generate a signal presence / absence map indicating the presence or absence of a signal by determining the presence or absence of a signal for each time and each frequency.

[0071] The output time comparison unit 342 estimates a continuous transmission time, which is radio wave characteristic information, in accordance with the signal presence / absence determination result by the signal determination unit 341. Then, the output time comparison unit 342 compares the estimated continuous transmission time with the continuous transmission time acquired from the learning device 200. Furthermore, the signal bandwidth comparison unit 343 estimates a signal bandwidth, which is radio wave characteristic information, in accordance with the signal presence / absence determination result by the signal determination unit 341. Then, the signal bandwidth comparison unit 343 compares the estimated signal bandwidth with the signal bandwidth acquired from the learning device 200. Furthermore, the continuous bandwidth comparison unit 344 estimates a continuous bandwidth, which is radio wave characteristic information, in accordance with the signal presence / absence determination result by the signal determination unit 341. Then, the continuous bandwidth comparison unit 344 compares the estimated continuous bandwidth with the continuous bandwidth acquired from the learning device 200. For example, as described above, each comparison unit estimates and compares radio wave characteristic information (step S203).

[0072] The abnormality determination unit 345 determines whether or not an abnormality exists depending on whether the comparison results of each comparison unit satisfy predetermined conditions (step S204). For example, if the conditions are satisfied (step S204, Yes), the abnormality determination unit 345 determines that an abnormality exists (step S205). On the other hand, if the conditions are not satisfied (step S204, No), the abnormality determination unit 345 determines that no abnormality exists (step S206).

[0073] The output unit 346 outputs the result of the determination by the abnormality determination unit 345 to an external device or the like (step S207). The output unit 346 may output the result of the determination, the results of the comparison by each comparison unit, the feature information used in the comparison, the frequency bin information, the signal presence / absence map, and the like to the external device or the like.

[0074] The above is an example of the operation of the determination device 300.

[0075] As described above, the determination device 300 includes a feature extraction unit 330 and a determination unit 340. With this configuration, the determination unit 340 can determine the presence or absence of an abnormality by comparing radio wave feature information, which is acquired according to the statistical feature extracted by the feature extraction unit 330 and does not depend on the absolute value of the received power, with the results of prior learning by the learning device 200. In other words, the determination unit 340 can determine the presence or absence of an abnormality based on feature information, such as the continuous transmission time and signal bandwidth, that does not depend on the absolute value of the received power. This allows the determination unit 340 to appropriately determine the presence or absence of an abnormality even when the received power changes. As a result, the determination unit 340 can appropriately determine the presence or absence of an abnormality even in situations where the radio wave environment fluctuates, for example.

[0076] For example, an example of a source of interference in wireless communication is an interference signal such as a noise with a weak signal power. Such an interference signal is characterized by being continuously output in a relatively wide band. As described above, the method described in the present disclosure determines whether or not an abnormality exists based on characteristic information such as the continuous transmission time. Therefore, the method described in the present disclosure is also suitable for determining an abnormality caused by an interference signal having the above-described characteristics.

[0077] Furthermore, the above-described method is not a black-box type determination method but a white-box type determination method that focuses on the continuous transmission time and signal bandwidth for each frequency. Therefore, the determination device 300 can output the result of the determination, as well as the results of the comparisons by each comparison unit, characteristic information used in the comparison, frequency bin information, a signal presence / absence map, etc. This makes it possible to determine the cause of interference, etc.

[0078] Furthermore, the learning device 200 includes a feature extraction unit 230 and a learning unit 240. With this configuration, the learning unit 240 can estimate and learn radio wave feature information that is not dependent on the absolute value of the received power by determining the presence or absence of a signal according to the statistical feature extracted by the feature extraction unit 230. This makes it possible to make a determination using the learned results.

[0079] As described above, the learning device 200 and the determination device 300 may be fixed sensors rather than mobile sensors. Furthermore, when the learning device 200 and the determination device 300 are fixed sensors and the radio wave environment is expected to fluctuate relatively little, the learning unit 240 of the learning device 200 may include a maximum power learning unit 246, as shown in FIG. 13 . The maximum power learning unit 246 can learn the maximum power value for each frequency bin. For example, the maximum power learning unit 246 may extract and learn a power step value at which CDF = 1.0 as the maximum power value. Furthermore, the determination device 300 may include a maximum power comparator 347, as shown in FIG. 14 . The maximum power comparator 347 estimates the maximum power value using a method similar to that of the maximum power learning unit 246. The maximum power comparator 347 can then compare the estimated maximum power value with the maximum power value acquired from the learning device 200. Furthermore, in this configuration, the abnormality determination unit 345 may be configured to determine the presence or absence of an abnormality by also referring to the comparison results by the maximum power comparison unit 347. The abnormality determination unit 345 may determine the presence or absence of an abnormality based on the individual comparison results by each comparison unit, or may determine the presence or absence of an abnormality by combining the comparison results by each comparison unit. For example, the abnormality determination unit 345 may determine that there is an abnormality when the maximum power comparison unit 347 confirms that there is no match and at least one of the output time comparison unit 342, the signal bandwidth comparison unit 343, and the continuous bandwidth comparison unit 344 confirms that there is no match. The abnormality determination unit 345 may also determine the presence or absence of an abnormality based on conditions other than those exemplified above.

[0080] The learning unit 240 of the learning device 200 may also include a model learning unit 247, as shown in FIG. 13 . The model learning unit 247 may perform machine learning of a model using statistical features such as PDF and CDF. For example, the model learning unit 247 may perform machine learning, such as unsupervised learning, using a feature vector that represents statistical features in multiple dimensions. For example, the model learning unit 247 may perform machine learning using a method similar to that described in Patent Document 1. The determination device 300 may also include an anomaly level estimation unit 348, as shown in FIG. 14 . The anomaly level estimation unit 348 can obtain outputs such as anomaly levels by inputting a feature vector that represents statistical features in multiple dimensions into a trained model. In this configuration, the anomaly determination unit 345 may also be configured to determine the presence or absence of an anomaly by referring to the estimation results by the anomaly level estimation unit 348. As in the above-described case, the anomaly determination unit 345 may determine the presence or absence of an anomaly based on the comparison results and estimation results by each comparison unit, or may determine the presence or absence of an anomaly by combining these results.

[0081] For example, the learning device 200 and the determination device 300 can have any one of the above-described modified examples or a combination thereof. By configuring the learning device 200 and the determination device 300, which are fixed sensors, as described above, it becomes possible to determine an abnormality using the maximum power value and the results of machine learning, and to output information indicating the cause of interference, etc.

[0082] Second Embodiment Next, configuration examples of a determination device 400 and a learning device 500 will be described with reference to Fig. 15 to Fig. 18. Fig. 15 is a diagram showing an example of the hardware configuration of the determination device 400. Fig. 16 is a block diagram showing an example of the configuration of the determination device 400. Fig. 17 is a flowchart showing an example of the operation of the determination device 400. Fig. 18 is a block diagram showing an example of the configuration of the learning device 500.

[0083] In a second embodiment of the present disclosure, a determination device 400 is described that estimates feature information such as a signal bandwidth according to a determination result of the presence or absence of a signal, and determines the presence or absence of an abnormality such as interference by comparing the estimated feature information with previously learned information. Also described is a learning device 500 that learns information used when the determination device 400 makes a determination. Fig. 15 shows an example of the hardware configuration of the determination device 400. Referring to Fig. 15, the determination device 400 has, as an example, the following hardware configuration. CPU (Central Processing Unit) 401 (arithmetic unit) ROM (Read Only Memory) 402 (storage device) RAM (Random Access Memory) 403 (storage device) Programs 404 loaded into RAM 403 Storage device 405 for storing the programs 404 Drive device 406 for reading and writing data from and to a storage medium 410 external to the information processing device Communication interface 407 for connecting to a communication network 411 external to the information processing device Input / output interface 408 for inputting and outputting data Bus 409 for connecting the various components

[0084] 16 . The program group 404 is stored in advance in the storage device 405 or the ROM 402, for example, and is loaded into the RAM 403 or the like by the CPU 401 as needed for execution. The program group 404 may be supplied to the CPU 401 via the communication network 411, or may be stored in advance in the recording medium 410, and the drive device 406 may read out the program and supply it to the CPU 401.

[0085] 15 shows an example of the hardware configuration of the determination device 400. The hardware configuration of the determination device 400 is not limited to the above-described case. For example, the determination device 400 may be configured with only a part of the above-described configuration, such as excluding the drive device 406. Furthermore, the CPU 401 may be a GPU or FPGA, as exemplified in the first embodiment.

[0086] The signal determination unit 421 determines the presence or absence of a signal for each predetermined frequency domain and time domain using the feature amount extracted from the received signal. The signal determination unit 421 may generate a signal presence / absence map indicating the presence or absence of a signal for each frequency and time according to the determination result.

[0087] The signal bandwidth comparison unit 422 estimates the signal bandwidth, which is characteristic information that does not depend on the absolute value of the received power, according to the determination result by the signal determination unit 421. Then, the signal determination unit 421 compares the estimated signal bandwidth with the normal signal bandwidth learned in advance.

[0088] The abnormality determination unit 423 determines whether or not there is an abnormality based on the result of the comparison by the signal bandwidth comparison unit 422. For example, the abnormality determination unit 423 can determine that there is an abnormality when the estimated signal bandwidth does not match the signal bandwidth under normal conditions that has been learned in advance.

[0089] The above is an example of the configuration of the determination device 400. Next, an example of the operation of the determination device 400 will be described with reference to FIG.

[0090] Fig. 17 shows an example of the operation of the determination device 400. Referring to Fig. 17, the signal determination unit 421 uses features extracted from a received signal to determine the presence or absence of a signal for each predetermined frequency domain and time domain (step S301).

[0091] The signal bandwidth comparison unit 422 estimates the signal bandwidth, which is characteristic information that does not depend on the absolute value of the received power, in accordance with the determination result by the signal determination unit 421. Then, the signal determination unit 421 compares the estimated signal bandwidth with the normal signal bandwidth learned in advance (step S302).

[0092] The abnormality determination unit 423 determines whether or not an abnormality exists based on the result of the comparison by the signal bandwidth comparison unit 422 (step S303).

[0093] As described above, the determination device 400 includes the signal bandwidth comparison unit 422 and the abnormality determination unit 423. With this configuration, the signal bandwidth comparison unit 422 can estimate the signal bandwidth, which is characteristic information that does not depend on the absolute value of the received power, and compare the estimated signal bandwidth with the normal signal bandwidth that has been learned in advance. As a result, the abnormality determination unit 423 can determine whether or not an abnormality exists based on the comparison result by the signal bandwidth comparison unit 422. This allows determination to be made using characteristic information that does not depend on the absolute value of the received power, making it possible to appropriately determine whether or not an abnormality exists, even in situations where the radio wave environment fluctuates, for example.

[0094] The above-described determination device 400 can be realized by incorporating a predetermined program into an information processing device such as the determination device 400. Specifically, a program according to another embodiment of the present disclosure is a program for causing an information processing device to realize processing of determining the presence or absence of a signal for each predetermined frequency domain and time domain using feature amounts extracted from a received signal, estimating a signal bandwidth that is feature information that does not depend on the absolute value of the received power according to the determination result, comparing the estimated signal bandwidth with a signal bandwidth under normal conditions that has been learned in advance, and determining the presence or absence of an abnormality according to the comparison result.

[0095] Furthermore, a determination method executed by an information processing device such as the determination device 400 described above is a method in which the information processing device determines the presence or absence of a signal for each predetermined frequency domain and time domain using features extracted from the received signal, estimates the signal bandwidth, which is feature information that does not depend on the absolute value of the received power, based on the determination result, compares the estimated signal bandwidth with a normal signal bandwidth that has been learned in advance, and determines the presence or absence of an abnormality based on the comparison result.

[0096] Any program having the above-described configuration, or a computer-readable recording medium having the program recorded thereon, or a determination method, etc., can achieve the same functions and effects as the above-described determination device 400, and therefore can achieve the above-described objective of the present disclosure.

[0097] Furthermore, information such as the signal bandwidth under normal conditions used in the determination device 400 described above can be learned using a learning device 500. Fig. 18 shows an example configuration of the learning device 500. The learning device 500 can realize the functions of the signal determination unit 521 and the signal bandwidth learning unit 522 shown in Fig. 18 by having a CPU acquire and execute a group of programs. The hardware configuration of the learning device 500 may be the same as the hardware configuration of the determination device 400 described with reference to Fig. 15.

[0098] The signal determination unit 521 uses the feature amount extracted from the received signal to determine the presence or absence of a signal for each predetermined frequency domain and time domain.

[0099] The signal bandwidth learning unit 522 estimates and learns the signal bandwidth, which is characteristic information that does not depend on the absolute value of the received power, according to the determination result by the signal determination unit 521. The learning result by the signal bandwidth learning unit 522 may be transmitted to an external device such as the determination device 400.

[0100] With this configuration, the determination device 400 can make a determination using the learning results from the learning device 500. Therefore, even with the learning device 500 having the above configuration, the object of the present invention can be achieved in the same way as the determination device 400.

[0101] <Supplementary Notes> Part or all of the above-described embodiments can be described as follows: The following provides an overview of the determination device and other components of the present disclosure. However, the present disclosure is not limited to the following configuration.

[0102] (Supplementary Note 1) A determination device comprising: a signal determination unit that determines the presence or absence of a signal for each predetermined frequency domain and time domain using a feature amount extracted from a received signal, a signal bandwidth comparison unit that estimates a signal bandwidth, which is feature information independent of an absolute value of received power, according to a determination result by the signal determination unit, and compares the estimated signal bandwidth with a pre-learned signal bandwidth under normal conditions, and an abnormality determination unit that determines the presence or absence of an abnormality according to a result of the comparison by the signal bandwidth comparison unit. (Supplementary Note 2) The determination device according to Supplementary Note 1, further comprising: an output time comparison unit that estimates a continuous transmission time indicating a time for which a signal continues as the feature information for each frequency domain according to a determination result by the signal determination unit, and compares the estimated continuous transmission time with a pre-learned continuous transmission time under normal conditions, (Supplementary Note 3) The determination device according to Supplementary Note 1 or Supplementary Note 2, further comprising: a continuous bandwidth comparison unit that estimates a continuous bandwidth indicating a bandwidth that will continue for a predetermined time period in accordance with a determination result by the signal determination unit, and compares the estimated continuous bandwidth with a continuous bandwidth in normal operation that has been learned in advance, wherein the abnormality determination unit determines the presence or absence of an abnormality in accordance with a result of the comparison by the signal bandwidth comparison unit and a result of the comparison by the continuous bandwidth comparison unit. (Supplementary Note 4) The determination device according to Supplementary Note 3, further comprising: an output time comparison unit that estimates a continuous transmission time indicating a time period for which a signal will continue as the feature information for each frequency domain in accordance with a determination result by the signal determination unit, and compares the estimated continuous transmission time with a continuous transmission time in normal operation that has been learned in advance, wherein the continuous bandwidth comparison unit estimates a continuous bandwidth indicating a bandwidth that will continue for the estimated continuous transmission time in accordance with a determination result by the signal determination unit. (Supplementary Note 5) The determination device according to any one of Supplementary Notes 1 to 4, wherein the signal determination unit generates a signal presence / absence map by determining the presence or absence of a signal for each time and frequency using a boundary estimated using a feature extracted from a received signal, and the signal bandwidth comparison unit estimates the signal bandwidth using the signal presence / absence map generated by the signal determination unit.(Supplementary Note 6) The determination device according to any one of Supplementary Notes 1 to 5, comprising a maximum power comparison unit that estimates a maximum power value for each frequency domain and compares the estimated maximum power value with a maximum power value in normal operation that has been learned in advance, wherein the abnormality determination unit determines the presence or absence of an abnormality based on a result of the comparison by the signal bandwidth comparison unit and a result of the comparison by the maximum power comparison unit. (Supplementary Note 7) The determination device according to any one of Supplementary Notes 1 to 6, comprising an abnormality degree estimation unit that obtains an output of an abnormal value by inputting feature amounts extracted from a received signal into a model that has been trained by performing machine learning using feature amounts, wherein the abnormality determination unit determines the presence or absence of an abnormality based on a result of the comparison by the signal bandwidth comparison unit and an output from the abnormality degree estimation unit. (Supplementary Note 7-1) The determination device according to any one of Supplementary Notes 1 to 7, comprising a feature extraction unit that performs spectrogram processing on a received signal, extracts the signal, and characterizes it to extract features from the received signal, wherein the signal determination unit uses the features extracted by the feature extraction unit to determine the presence or absence of a signal for each of a predetermined frequency domain and a time domain. (Supplementary Note 8) A determination method in which an information processing device uses the features extracted from the received signal to determine the presence or absence of a signal for each of a predetermined frequency domain and a time domain, estimates a signal bandwidth that is feature information that does not depend on the absolute value of the received power according to the determination result, compares the estimated signal bandwidth with a normal signal bandwidth that has been learned in advance, and determines the presence or absence of an abnormality according to the comparison result. (Supplementary Note 9) A program for implementing in an information processing device the following process: using features extracted from a received signal, determining the presence or absence of a signal for each predetermined frequency domain and time domain; estimating the signal bandwidth, which is feature information that does not depend on the absolute value of the received power, according to the determination result; comparing the estimated signal bandwidth with a pre-learned normal signal bandwidth; and determining the presence or absence of an abnormality according to the comparison result.(Supplementary Note 10) A learning device comprising: a signal determination unit that determines the presence or absence of a signal for each predetermined frequency domain and time domain using features extracted from a received signal; and a signal bandwidth learning unit that estimates and learns a signal bandwidth, which is feature information that does not depend on the absolute value of the received power, according to a determination result by the signal determination unit.

[0103] Note that some or all of the configurations described in Supplementary Notes 2 to 7-1 that are dependent on the determination device described in Supplementary Note 1 may also be dependent on the determination method described in Supplementary Note 8, the program described in Process 9, and the like in a similar dependent relationship. Furthermore, not limited to Supplementary Notes 8 and 9, but within the scope of each of the above-mentioned embodiments, some or all of the configurations described as Supplements may also be dependent on various hardware, software, various recording means for recording software, or systems. Furthermore, the learning device described in Supplementary Note 10 may also have a configuration corresponding to Supplementary Notes 2 to 7-1.

[0104] The programs described in the above embodiments and appendices may be stored in a storage device or a computer-readable recording medium, such as a portable medium such as a flexible disk, an optical disk, a magneto-optical disk, or a semiconductor memory.

[0105] Although the present disclosure has been described above with reference to the above-described embodiments, the present disclosure is not limited to the above-described embodiments. Various modifications that can be understood by those 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 combined with other embodiments as appropriate.

[0106] The present invention claims the benefit of priority based on the patent application of Japanese Patent Application No. 2023-209180 filed on December 12, 2023 in Japan, and all contents described in the patent application are herein.

[0107] 100 Abnormality determination system 200 Learning device 210 Antenna 220 Receiving unit 230 Feature extraction unit 231 Spectrogramming unit 232 Signal extraction unit 233 Feature quantity generation unit 240 Learning unit 241 Signal determination unit 242 Output time learning unit 243 Signal bandwidth learning unit 244 Continuous bandwidth learning unit 245 Output unit 246 Maximum power learning unit 247 Model learning unit 300 Determination device 310 Antenna 320 Receiving unit 330 Feature extraction unit 331 Spectrogramming unit 332 Signal extraction unit 333 Feature quantity generation unit 340 Determination unit 341 Signal determination unit 342 Output time comparison unit 343 Signal bandwidth comparison unit 344 Continuous bandwidth comparison unit 345 Abnormality determination unit 346 Output unit 347 Maximum power comparison unit 348 Abnormality degree estimation unit 400 Determination device 401 CPU 402 ROM 403 RAM 404 Program group 405 Storage device 406 Drive device 407 Communication interface 408 Input / output interface 409 Bus 410 Recording medium 411 Communication network 421 Signal determination unit 422 Signal bandwidth comparison unit 423 Abnormality determination unit 500 Learning device 521 Signal determination unit 522 Signal bandwidth learning unit

Claims

1. A determination device having: a signal determination unit that determines the presence or absence of a signal for each predetermined frequency domain and time domain using features extracted from a received signal; a signal bandwidth comparison unit that estimates a signal bandwidth, which is feature information independent of the absolute value of the received power, according to a determination result by the signal determination unit, and compares the estimated signal bandwidth with a normal signal bandwidth that has been learned in advance; and an abnormality determination unit that determines the presence or absence of an abnormality according to the comparison result by the signal bandwidth comparison unit.

2. A determination device as claimed in claim 1, comprising an output time comparison unit which estimates a continuous transmission time indicating the time a signal continues as the characteristic information for each frequency domain according to the determination result by the signal determination unit, and compares the estimated continuous transmission time with a normal continuous transmission time which has been learned in advance, and the abnormality determination unit determines whether or not there is an abnormality according to the result of the comparison by the signal bandwidth comparison unit and the result of the comparison by the output time comparison unit.

3. A determination device as claimed in claim 1, comprising a continuous bandwidth comparison unit which estimates a continuous bandwidth indicating a bandwidth continuing for a predetermined period of time according to the determination result by said signal determination unit, and compares the estimated continuous bandwidth with a normal continuous bandwidth which has been learned in advance, and said abnormality determination unit determines the presence or absence of an abnormality according to the result of the comparison by said signal bandwidth comparison unit and the result of the comparison by said continuous bandwidth comparison unit.

4. A determination device as claimed in claim 3, comprising an output time comparison unit which estimates a continuous transmission time indicating the time a signal continues as the characteristic information for each frequency domain according to the determination result by the signal determination unit, and compares the estimated continuous transmission time with a normal continuous transmission time which has been learned in advance, and the continuous bandwidth comparison unit estimates a continuous bandwidth indicating the bandwidth which continues during the estimated continuous transmission time according to the determination result by the signal determination unit.

5. A determination device as claimed in claim 1, wherein the signal determination unit generates a signal presence / absence map by determining the presence or absence of a signal for each time and frequency using boundaries estimated using features extracted from the received signal, and the signal bandwidth comparison unit estimates the signal bandwidth using the signal presence / absence map generated by the signal determination unit.

6. A determination device as claimed in claim 1, further comprising a maximum power comparison section which estimates a maximum power value for each frequency domain and compares the estimated maximum power value with a previously learned maximum power value under normal conditions, and wherein said abnormality determination section determines the presence or absence of an abnormality according to a result of comparison made by said signal bandwidth comparison section and a result of comparison made by said maximum power comparison section.

7. A determination device as claimed in claim 1, comprising an anomaly degree estimation unit which obtains an output of an abnormal value by inputting features extracted from a received signal into a model trained by performing machine learning using features, and the anomaly determination unit determines the presence or absence of an abnormality according to a result of comparison by the signal bandwidth comparison unit and an output from the anomaly degree estimation unit.

8. A determination device as claimed in claim 1, further comprising a feature extraction unit which performs spectrogram processing on the received signal, extracts the signal and characterizes it to extract features from the received signal, and the signal determination unit uses the features extracted by the feature extraction unit to determine the presence or absence of a signal for each of a predetermined frequency domain and a predetermined time domain.

9. A method of determination in which an information processing device uses features extracted from a received signal to determine the presence or absence of a signal for each specified frequency domain and time domain, estimates the signal bandwidth, which is feature information independent of the absolute value of the received power, based on the determination result, compares the estimated signal bandwidth with a normal signal bandwidth that has been learned in advance, and determines the presence or absence of an abnormality based on the comparison result.

10. A computer-readable recording medium having recorded thereon a program for implementing the process of: determining the presence or absence of a signal for each specified frequency domain and time domain using features extracted from a received signal; estimating a signal bandwidth, which is feature information independent of the absolute value of the received power, based on the determination result; comparing the estimated signal bandwidth with a normal signal bandwidth that has been learned in advance; and determining the presence or absence of an abnormality based on the comparison result.

11. A learning device having: a signal determination unit that determines the presence or absence of a signal for each specified frequency domain and time domain using features extracted from a received signal; and a signal bandwidth learning unit that estimates and learns a signal bandwidth, which is feature information that does not depend on the absolute value of the received power, based on a determination result by the signal determination unit.

Images