Signal processing device, signal processing method, and program

Abstract

Provided is a signal processing device capable of suitably detecting the occurrence of an event. A signal processing device according to one embodiment of the present disclosure is equipped with: a conversion means for converting an input signal into a prescribed signal, which is a signal in a time-frequency domain; a target signal estimation means for estimating the peak of the time-frequency intensity of the prescribed signal as the intensity of a target signal, which is a signal pertaining to the occurrence of an event; a band estimation means for estimating a band, which at least contains the bandwidth from the frequency pertaining to said peak to the prescribed frequency and does not contain a frequency pertaining to a peak other than said peak, as a noise band which is the frequency band of the noise signal; a noise signal estimation means for estimating the intensity of the noise signal on the basis of the time-frequency intensity in the noise band; and a determination means for determining whether or not an event has occurred on the basis of the ratio of the intensity of the target signal to the intensity of the noise signal.

Description

[Technical Field]

[0001] The present disclosure relates to a technique for detecting the occurrence of an event from a signal. [Background technology]

[0002] Patent Document 1 discloses a technique for detecting the occurrence of a specific event from a time-series signal, which is a signal related to sound.

[0003] Patent Document 1 discloses a technology that defines a specific sound as a detection target sound and detects the occurrence of the detection target sound from a mixture of the detection target sound and background sounds different from the detection target sound. Specifically, the technology disclosed in Patent Document 1 defines a target frequency band, which is the frequency band of the detection target sound, and a control frequency band, which is not the frequency band of the detection target sound, and detects the detection target sound by comparing the average frequency intensity in each band.

[0004] Furthermore, as a related document, Patent Document 2 discloses a technique for determining whether a target sound is not occurring in a target frame of an input signal. Specifically, the technique disclosed in Patent Document 2 acquires the power of an emphasis signal, which is the target sound, from one frame in the past, and determines that the target sound is not occurring in the target frame if the power is equal to or less than a certain value. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Laid-Open No. 2015-64502 [Patent Document 2] International Publication No. 2014-136629 Summary of the Invention [Problem to be solved by the invention]

[0006] One example of a situation where an event occurrence is detected from an input signal is when detecting abnormal noises generated by various types of equipment, such as vehicles such as bullet trains, automobiles, and airplanes, or machine tools. The abnormal noises are, for example, sounds that are different from the sounds generated by the equipment's normal operation. Here, the abnormal noises may include friction noises and sliding noises of components used in the equipment. The frequencies of such friction noises and sliding noises vary depending on the type of component.

[0007] The technology of Patent Document 1 requires that a target frequency band and a control frequency band be set in advance to detect the target sound. Therefore, as described above, when the frequency of the target sound to be detected varies depending on the type of component, it is difficult to set the frequency band appropriately. In other words, the technology of Patent Document 1 may not be able to properly detect the occurrence of an event.

[0008] The technology of Patent Document 2 is a technology for cases where the frequency identification of the target sound signal is known in advance. In other words, Patent Document 2 does not disclose detecting a sound when the frequency of the sound to be detected varies depending on the situation, as described above.

[0009] The present disclosure has been made in consideration of the above-mentioned problems, and one of its objects is to provide a signal processing device and the like that can appropriately detect the occurrence of an event. [Means for solving the problem]

[0010] A signal processing device according to one aspect of the present disclosure includes a conversion means for converting an input signal into a predetermined signal which is a signal in the time-frequency domain; a target signal estimation means for estimating a peak of the time-frequency intensity of the predetermined signal as the intensity of a target signal which is a signal related to the occurrence of an event; a band estimation means for estimating a band which includes at least a bandwidth from a frequency related to the peak to a predetermined frequency and which does not include a frequency related to a peak different from the peak as a noise band which is the frequency band of a noise signal; a noise signal estimation means for estimating the intensity of the noise signal based on the time-frequency intensity in the noise band; and a determination means for determining whether or not an event has occurred based on the ratio between the intensity of the target signal and the intensity of the noise signal.

[0011] A signal processing method according to one aspect of the present disclosure converts an input signal into a predetermined signal that is a signal in the time-frequency domain, estimates a peak in the time-frequency intensity of the predetermined signal as the intensity of a target signal that is a signal related to the occurrence of an event, estimates a band that includes at least a bandwidth from a frequency related to the peak to a predetermined frequency and that does not include a frequency related to a peak different from the peak as a noise band that is a frequency band of a noise signal, estimates the intensity of the noise signal based on the time-frequency intensity in the noise band, and determines whether or not an event has occurred based on the ratio between the intensity of the target signal and the intensity of the noise signal.

[0012] A computer-readable storage medium according to one aspect of the present disclosure stores a program that causes a computer to execute the following processes: converting an input signal into a predetermined signal that is a signal in the time-frequency domain; estimating a peak in the time-frequency intensity of the predetermined signal as the intensity of a target signal that is a signal related to the occurrence of an event; estimating a band that includes at least a bandwidth from a frequency related to the peak to a predetermined frequency and that does not include a frequency related to a peak different from the peak as a noise band that is the frequency band of a noise signal; estimating the intensity of the noise signal based on the time-frequency intensity in the noise band; and determining whether or not an event has occurred based on the ratio between the intensity of the target signal and the intensity of the noise signal. [Effects of the Invention]

[0013] According to the present disclosure, it is possible to appropriately detect the occurrence of an event. [Brief explanation of the drawings]

[0014] [Figure 1] 1 is a block diagram illustrating an example of a functional configuration of a signal processing device according to a first embodiment of the present disclosure. [Figure 2] 5 is a flowchart illustrating an example of the operation of the signal processing device according to the first embodiment of the present disclosure. [Figure 3] FIG. 2 is a diagram illustrating an example of a signal in the time-frequency domain according to the first embodiment of the present disclosure. [Figure 4] FIG. 2 is a diagram illustrating an example of a noise band according to the first embodiment of the present disclosure. [Figure 5] FIG. 1 is a diagram illustrating an example of a relationship between a noise band and a frequency associated with a peak according to the first embodiment of the present disclosure; [Figure 6] 10 is a flowchart showing another example of the operation of the signal processing device according to the first embodiment of the present disclosure. [Figure 7] FIG. 10 is a block diagram illustrating an example of a functional configuration of a signal processing system according to a second embodiment of the present disclosure. [Figure 8] FIG. 10 is a block diagram illustrating an example of a functional configuration of a signal processing device according to a modified example of the present disclosure. [Figure 9] FIG. 2 is a block diagram showing an example of the hardware configuration of a computer device that realizes each device of the first and second embodiments of the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0015] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings.

[0016] First Embodiment A signal processing device according to a first embodiment will be described.

[0017] 1 is a block diagram showing an example of the functional configuration of a signal processing device 100 according to the first embodiment. As shown in FIG. 1, the signal processing device 100 includes a conversion unit 110, a target signal estimation unit 120, a band estimation unit 130, a noise signal estimation unit 140, and a determination unit 150.

[0018] The conversion unit 110 converts the input signal into a predetermined signal, which is a signal in the time-frequency domain. The input signal is, for example, a time-series signal input to the signal processing device 100, and is a signal that includes signals of multiple types of sounds. The conversion unit 110 converts the input signal into a signal in the time-frequency domain, for example, by using a discrete Fourier transform. The conversion unit 110 is an example of conversion means. In the present disclosure, the signal in the time-frequency domain converted from the input signal is also referred to as a predetermined signal.

[0019] The target signal estimation unit 120 estimates the peak of the time-frequency intensity of a predetermined signal, which is a signal in the time-frequency domain, as the intensity of a target signal, which is a signal related to the occurrence of an event. For example, the target signal estimation unit 120 detects the peak of the time-frequency intensity of the predetermined signal. Then, the target signal estimation unit 120 determines, for example, the detected peak of the time-frequency intensity as the intensity of the target signal. Here, the target signal is a signal related to the occurrence of an event. The target signal is, for example, a signal of a target sound that the user wants to detect, such as an abnormal sound generated by a device. In other words, the target signal estimation unit 120 assumes that the peak of the time-frequency intensity of the predetermined signal is the intensity of the target signal. The target signal estimation unit 120 is an example of a target signal estimation means.

[0020] The band estimation unit 130 estimates a noise band, which is the frequency band of the noise signal, as a band that includes at least a bandwidth from a frequency associated with a peak to a predetermined frequency, but does not include a frequency associated with a peak other than the peak. The frequency associated with a peak may be, for example, a frequency corresponding to the peak, or a frequency corresponding to a portion from the peak to its base. The base of a peak refers to the rising and falling portions of a waveform representing the peak. Hereinafter, the frequency corresponding to a peak is also referred to as the peak frequency. The noise signal is a signal representing a sound different from the target signal. The band estimation unit 130 may estimate the noise band to include, for example, a band from the peak frequency to an intermediate frequency between the target peak and a peak adjacent to the target peak. Alternatively, for example, the band estimation unit 130 may set the minimum frequency of the noise band to be less than the peak frequency and the maximum frequency of the noise band to be greater than the peak frequency. The band estimation unit 130 is an example of a band estimation means.

[0021] The noise signal estimation unit 140 estimates the intensity of the noise signal based on the time frequency intensity in the noise band. For example, the noise signal estimation unit 140 estimates a representative value of the time frequency intensity in the noise band as the intensity of the noise signal. The representative value may be any of the time frequency intensities in the noise band, or may be a value obtained by performing a predetermined calculation on the time frequency intensities in the noise band. When the noise band includes a frequency associated with a peak, the noise signal estimation unit 140 may estimate the intensity of the noise signal based on, for example, the time frequency intensity at a frequency in the noise band that is different from the frequency associated with the peak. The noise signal estimation unit 140 is an example of a noise signal estimation means.

[0022] The determination unit 150 determines whether an event has occurred based on the ratio between the intensity of the target signal and the intensity of the noise signal. For example, the determination unit 150 determines that an event has occurred when the ratio between the intensity of the target signal and the intensity of the noise signal is equal to or greater than a predetermined threshold. In this case, the determination unit 150 may determine that an event has occurred when a value obtained by performing a predetermined calculation on the ratio between the intensity of the target signal and the intensity of the noise signal is equal to or greater than a predetermined threshold. The determination unit 150 is an example of a determination means.

[0023] Next, an example of the operation of the signal processing device 100 will be described with reference to Fig. 2. In this disclosure, each step in a flowchart will be expressed using a number assigned to each step, such as "S1".

[0024] 2 is a flowchart illustrating an example of the operation of the signal processing device 100. The conversion unit 110 converts an input signal into a predetermined signal, which is a signal in the time-frequency domain (S1). The target signal estimation unit 120 estimates a peak in the time-frequency intensity of the predetermined signal as the intensity of the target signal, which is a signal related to the occurrence of an event (S2). The band estimation unit 130 estimates, as the noise band, a band that is a bandwidth from a frequency related to the peak to a predetermined frequency and that includes at least a bandwidth that does not include a frequency related to the peak other than the peak (S3). The noise signal estimation unit 140 estimates the intensity of the noise signal based on the time-frequency intensity in the noise band (S4). Then, the determination unit 150 determines whether or not an event has occurred based on the ratio between the intensity of the target signal and the intensity of the noise signal (S5).

[0025] The signal processing device 100 of the present disclosure converts an input signal into a predetermined signal, which is a signal in the time-frequency domain, and estimates a peak in the time-frequency intensity of the predetermined signal as the intensity of a target signal, which is a signal related to the occurrence of an event. The signal processing device 100 also estimates a noise band, which is the frequency band of a noise signal, that includes at least a bandwidth from a frequency related to the peak to a predetermined frequency but does not include a frequency related to a peak different from the peak. The signal processing device 100 then estimates the intensity of the noise signal based on the time-frequency intensity in the noise band and determines whether an event has occurred based on the ratio between the intensity of the target signal and the intensity of the noise signal. In this way, the signal processing device 100 of the present disclosure can dynamically determine the target signal, which is a signal related to the occurrence of an event, and the noise signal. Therefore, the signal processing device 100 of the present disclosure can detect the occurrence of unknown sounds, such as frictional sounds and sliding sounds, whose frequency changes depending on the type of material generating the sound. In other words, the signal processing device 100 of the present disclosure can appropriately detect the occurrence of an event.

[0026] [Another example of the signal processing device 100] Next, another example of the signal processing device 100 of the present disclosure will be described in detail. Specifically, an example of further functions and operations of each component of the signal processing device 100 of FIG.

[0027] The signal processing device 100 of the present disclosure determines whether an event has occurred based on information about a target signal and a noise signal estimated from an input signal. The target signal refers to a target sound signal. The occurrence of a target sound is an example of the occurrence of an event. That is, the signal processing device 100 of the present disclosure determines whether a target sound has occurred based on information about the signal estimated as the target signal. For example, in the above-mentioned scene of detecting abnormal sounds generated from various devices, the abnormal sound is the target sound.

[0028] The conversion unit 110 acquires an input signal input to the signal processing device 100. Then, the conversion unit 110 converts the input signal into a signal in the time-frequency domain. Specifically, the conversion unit 110 cuts out the input signal for each predetermined interval. At this time, the conversion unit 110 may use a predetermined window function to cut out the signal for the predetermined interval from the input signal. For example, the conversion unit 110 can cut out the signal for the predetermined interval using a rectangular window. This example is not limiting, and the conversion unit 110 may cut out the signal using other window functions such as a Gaussian window, a Hanning window, or a Hamming window. Furthermore, the cut-out intervals may be different from one another.

[0029] The transform unit 110 calculates a frequency domain signal, i.e., a frequency spectrum, for a signal extracted from the input signal. At this time, the transform unit 110 calculates the frequency spectrum by, for example, using a discrete Fourier transform for the extracted signal. Note that, without being limited to this example, the transform unit 110 may calculate the frequency spectrum using a known method such as a wavelet transform. In the present disclosure, a frequency domain signal in a certain section of the input signal is referred to as a time-frequency domain signal. FIG. 3 is a diagram showing an example of a time-frequency domain signal. FIG. 3 shows a frequency spectrum at time t, which indicates a certain section. The horizontal axis of the spectrum in FIG. 3 represents frequency, and the vertical axis represents intensity. Here, the intensity is, for example, amplitude. Note that the intensity is not limited to amplitude, and logarithmic amplitude, power, logarithmic power, etc. may also be used. Also, in the present disclosure, the intensity of a frequency f at time t is referred to as time-frequency intensity, and is denoted as S(t,f).

[0030] The target signal estimation unit 120 estimates the intensity of the target signal from the signal in the time-frequency domain. Specifically, the target signal estimation unit 120 detects peaks in the time-frequency intensity. In this case, the target signal estimation unit 120 may perform peak detection by calculating a local maximum value using differentiation in the time-frequency domain signal, or by searching for a maximum value in a local region. The target signal estimation unit 120 may also perform peak detection based on the envelope of the time-frequency intensity. By performing peak detection based on the envelope, it is possible to reduce erroneous detection of peaks due to noise. Note that the peak detection method is not limited to this example, and other known methods may also be used.

[0031] The target signal estimation unit 120 estimates the peak of the time-frequency intensity as the intensity of the target signal. Pm (t,f Pm ) where Pm indicates the peak index. m is an integer equal to or greater than 1 and takes a value up to the number of detected peaks. f Pm indicates the peak frequency. That is, the target signal estimation unit 120 estimates the intensity of the target signal for each detected peak. For example, if five peaks are detected, the target signal estimation unit 120 estimates S P1 (t,f P1 ) to S P5 (t,f P5 ) to identify the value.

[0032] The number of target signal strengths to be estimated may be set in advance. For example, if the number of target signal strengths to be estimated is M (M is an integer equal to or greater than 1), target signal estimation unit 120 may detect up to M peaks and estimate the target signal strength for each of the M detected peaks. That is, target signal estimation unit 120 may detect a number of time-frequency strength peaks of a predetermined signal that is equal to or less than a preset number, and estimate each of the detected peaks as the target signal strength. This makes it possible to prevent degradation of peak detection performance due to signal noise.

[0033] The band estimation unit 130 estimates the noise band for each peak. For example, the band estimation unit 130 sets the minimum and maximum frequencies of the noise band based on the peak frequency. The minimum and maximum frequencies may be intermediate frequencies between the frequency of the target peak and the frequency of a peak adjacent to the target peak. In this case, the minimum frequency is calculated, for example, as follows:

[0034]

number

[0035] Nm is the noise index. When m=1, f P0 may be 0, or f P1 Alternatively, the maximum frequency may be calculated by subtracting a predetermined number from the maximum frequency.

[0036]

number

[0037] f Pm+1 If f does not exist, Pm+1 may be set to 0 or a predetermined number. The predetermined number may be, for example, the maximum frequency F of the target frequency spectrum. FIG. 4 is a diagram showing an example of a noise band. Specifically, FIG. 4 shows an example of a noise band when m=2, and the minimum frequency is f P1 and f P2 The intermediate frequency and the maximum frequency are f P2 and f P3 In this example, the frequency is set to an intermediate frequency between the peaks. The band estimation unit 130 estimates the noise band for each peak.

[0038] The method for estimating the noise band is not limited to the above example. For example, the minimum frequency of the noise band can be calculated by Pm and f Pm―1 For example, the maximum frequency of the noise band may be a frequency corresponding to a minimum value of the time frequency intensity in the band between f Pmand f Pm+1 Alternatively, the band estimation unit 130 may set the minimum frequency to a frequency corresponding to a minimum value of the time frequency intensity in the band between f Pm―1 The maximum frequency is f Pm+1 The frequency of the base of the rising part of the peak of f Pm The noise band may not include f Pm A bandwidth from the base frequency of the rising part of the peak to the first predetermined frequency, and f Pm The first predetermined frequency is f Pm The frequency of the tail of the rising part of the peak and f Pm-1 The second predetermined frequency is any frequency between the frequency of the foot of the falling part of the peak of f Pm The frequency of the trailing edge of the peak and f Pm+1 The frequency is any frequency between the frequency at the base of the rising part of the peak.

[0039] The method for detecting the base of a peak may be a method for detecting it based on the variation of the slope of the signal in the time-frequency domain, or a method for detecting the minimum value immediately adjacent to the peak as the base of the peak. The detection of the base of the peak may be performed by the band estimation unit 130 or the noise signal estimation unit 140.

[0040] The noise signal estimation unit 140 estimates the noise signal strength for each peak based on the time frequency strength of the noise band estimated for each peak. For example, the noise signal estimation unit 140 determines the average of the time frequency strength in the noise band as the noise signal strength. Here, if the noise band includes a frequency related to the peak, the noise signal estimation unit 140 calculates the average of the time frequency strength at frequencies excluding the peak frequency from the noise band. The noise signal strength is expressed as S Nm If (t), then the intensity of the noise signal is calculated as follows, for example:

[0041]

number

[0042] In this way, when the minimum frequency of the noise band is less than the frequency associated with the peak and the maximum frequency of the noise band is greater than the frequency associated with the peak, the noise signal estimation unit 140 may estimate the intensity of the noise signal based on the intensity at frequencies in the band excluding the peak frequency, which is the frequency corresponding to the peak, from the noise band.

[0043] In the example of Equation 3, only the peak frequency is excluded from the noise band, but all frequencies related to the peak may be excluded from the noise band. That is, the noise signal estimation unit 140 may calculate the average of the time-frequency intensity at frequencies excluding the band from the base frequency of the rising part of the peak to the base frequency of the falling part of the peak in the noise band. In this case, for example, the intensity of the noise signal is calculated as follows:

[0044]

number

[0045] FIG. 5 is a diagram illustrating an example of the relationship between a noise band and a frequency associated with a peak. Specifically, FIG. 5 shows the noise band and the frequency band associated with the peak when calculating the intensity of a noise signal when m=2. Here, band A shown in FIG. 5 is the band between the minimum frequency of the noise band and the frequency at the base of the rising part of the peak, and band B is the band between the maximum frequency of the noise band and the frequency at the base of the falling part of the peak. That is, in the example shown in FIG. 5, when using equation 4, the noise signal estimation unit 140 calculates the average of the time-frequency intensity in band A and band B. In this way, the noise signal estimation unit 140 estimates the intensity of the noise signal based on the intensity at the frequency in the band excluding the frequencies associated with the peak (i.e., the frequencies between the bases of the rising part of the peak) from the noise band. This makes it possible to eliminate the influence of the peak when calculating the intensity of the noise signal, thereby enabling the intensity of the noise signal to be estimated with high accuracy.

[0046] In the above example, the method of calculating the average of the time-frequency intensity in the noise band or in the band obtained by excluding the frequencies associated with the peaks from the noise band has been described, but the method of calculating the intensity of the noise signal is not limited to this example. For example, the noise signal estimation unit 140 may calculate, as the intensity of the noise signal, any one of the minimum value, the mode, and the median value of the time-frequency intensity in the noise band or in the band obtained by excluding the frequencies associated with the peaks from the noise band.

[0047] The determination unit 150 determines whether an event has occurred based on the ratio between the intensity of the estimated target signal and the intensity of the noise signal. Specifically, the determination unit 150 calculates the ratio between the intensity of the estimated target signal and the intensity of the noise signal for each peak. The intensity ratio for each peak is denoted as S Rm If (t), then the ratio is calculated as follows:

[0048]

number

[0049] Then, the determining unit 150 calculates a score using the ratio calculated for each peak. For example, the determining unit 150 calculates a score using the ratio S Rm The score may be the average of (t). Rm The score may be a weighted average, maximum value, or minimum value of (t). Rm If the weighted average of (t) is used as the score, and the score is D(t), the score is calculated as follows:

[0050]

number

[0051] w m indicates a weight. The weight may be, for example, the ratio of the intensity of each target signal when the sum of the intensities of the target signals is 1, or the reciprocal of the peak frequency. In this way, the determination unit 150 may weight the ratio for each peak according to the magnitude of the intensity of the target signal calculated for each peak, and may use the average of the weighted ratios for each peak as the score.

[0052] The determination unit 150 determines whether an event has occurred based on the calculated score. For example, if the threshold value is θ, the determination unit 150 determines that an event has occurred if D(t)>θ, and determines that an event has not occurred if D(t)≦θ. The threshold value θ may be set to any value by the user. For example, the threshold value θ may be set to the average of scores calculated from input signals in which no event has occurred, but is not limited to this example. In this way, the determination unit 150 calculates the ratio between the intensity of the target signal and the intensity of the noise signal for each peak, and determines that an event has occurred if the score calculated from the ratio for each peak is greater than the threshold value.

[0053] Determining that an event has occurred indicates that the input signal contains a signal of the sound that the user wants to detect. For example, in the above-mentioned situation where abnormal sounds generated from various devices are to be detected, if the score is greater than the threshold value, it indicates that the input signal contains a signal indicating an abnormal sound, that is, that an abnormal sound is occurring.

[0054] [Operation of signal processing device 100] Next, another example of the operation of the signal processing device 100 will be described with reference to Fig. 6. Fig. 6 is a flowchart illustrating another example of the operation of the signal processing device 100.

[0055] First, the conversion unit 110 converts the input signal into a predetermined signal, which is a signal in the time-frequency domain (S101). The target signal estimation unit 120 detects peaks in the time-frequency intensity of the predetermined signal (S102). Then, the target signal estimation unit 120 estimates the detected peaks as the intensity of the target signal (S103). If multiple peaks are detected, the target signal estimation unit 120 estimates each of the peaks as the intensity of the target signal.

[0056] Next, the band estimation unit 130 estimates the noise band based on the peak frequency (S104). If multiple peaks are detected, the band estimation unit 130 estimates the noise band for each peak. The noise signal estimation unit 140 estimates the noise signal intensity for each peak based on the time-frequency intensity of the noise band (S105). Here, if the noise band includes a frequency associated with the peak, the noise signal estimation unit 140 estimates the noise signal intensity based on the time-frequency intensity of a band excluding the frequency associated with the peak from the noise band.

[0057] Then, the determination unit 150 calculates the ratio between the intensity of the target signal and the intensity of the noise signal for each peak (S106). The determination unit 150 calculates a score based on the calculated ratio (S107). If the score is greater than the threshold (Yes in S108), the determination unit 150 determines that an event has occurred (S109). If the score is less than the threshold, smallIf so (No in S108), the determination unit 150 determines that an event has not occurred (S110).

[0058] As described above, the signal processing device 100 of the first embodiment converts an input signal into a predetermined signal, which is a signal in the time-frequency domain, and estimates a peak in the time-frequency intensity of the predetermined signal as the intensity of a target signal, which is a signal related to the occurrence of an event. Furthermore, the signal processing device 100 estimates a band, which is a frequency band of a noise signal, that is a bandwidth from a frequency related to a peak to a predetermined frequency and that includes at least a bandwidth that does not include a frequency related to a peak different from the peak. The signal processing device 100 then estimates the intensity of the noise signal based on the time-frequency intensity in the noise band, and determines whether an event has occurred based on the ratio between the intensity of the target signal and the intensity of the noise signal.

[0059] For example, when detecting the occurrence of a specific sound from a sound signal, a method may be used in which the frequency pattern of the specific sound is modeled and the occurrence of the specific sound is detected based on the model. However, it is difficult to model the frequency pattern of unknown sounds whose frequency changes depending on the type of material generating the sound, such as the friction noise and sliding noise described above. In contrast, the signal processing device 100 of the first embodiment can dynamically determine a target signal, which is a signal related to the occurrence of an event, and a noise signal. Therefore, the signal processing device 100 can detect the occurrence of even unknown sounds whose frequency changes depending on the type of material generating the sound. In other words, the signal processing device 100 of the present disclosure can appropriately detect the occurrence of an event.

[0060] <Second embodiment> Next, a signal processing system including the signal processing device of the second embodiment will be described.

[0061] Fig. 7 is a block diagram showing an example of the functional configuration of a signal processing system 1000 according to the second embodiment. As shown in Fig. 7, the signal processing system 1000 includes a signal processing device 100, a learning device 200, and a storage device 300. The storage device 300 may be integrated with either the learning device or the signal processing device 100. Note that a description of the configuration and operation of the signal processing system 1000 shown in Fig. 6 that overlap with the description of the first embodiment will be omitted.

[0062] When a signal including a target signal is input, the learning device 200 determines whether an event has occurred in the input signal. That is, the learning device 200 determines whether a target sound has occurred in a signal that is known in advance to contain a target sound signal (e.g., an abnormal sound) that the user wants to detect. The learning device 200 determines whether an event has occurred using a method similar to that of the signal processing device 100 described in the first embodiment. At this time, the learning device 200 changes parameters used to determine whether an event has occurred and performs each process. Examples of parameters include, but are not limited to, a number M for estimating the target signal strength, a minimum frequency and a maximum frequency set when estimating the noise band, and a threshold θ. The parameters indicate the processing method performed by the signal processing device 100 and numerical values set for each method. The learning device 200 then outputs statistical information related to each parameter. The statistical information may include information associating the value of each parameter with the determination result and information indicating whether the determination result is correct. The statistical information may include the average, variance, maximum value, minimum value, etc. of the numerical values of the parameters used when the determination result is correct. The statistical information is stored in the storage device 300 .

[0063] Next, a description will be given of the functional configuration of the learning device 200. As shown in Fig. 7, the learning device 200 includes a second conversion unit 210, a second target signal estimation unit 220, a second band estimation unit 230, a second noise signal estimation unit 240, a second determination unit 250, and a management unit 260. The second conversion unit 210, the second target signal estimation unit 220, the second band estimation unit 230, the second noise signal estimation unit 240, and the second determination unit 250 each have the same functions as the conversion unit 110, the target signal estimation unit 120, the band estimation unit 130, the noise signal estimation unit 140, and the determination unit 150, respectively.

[0064] For example, the second conversion unit 210 performs processing by changing parameters related to the method of converting the input signal into a signal in the time-frequency domain, such as the section to be extracted from the input signal and the method of calculating the frequency spectrum of the extracted signal. Furthermore, for example, the second target signal estimation unit 220 performs processing by changing the peak detection method and the number M for estimating the intensity of the target signal. Furthermore, for example, the second band estimation unit 230 performs processing by changing the method of setting the minimum and maximum frequencies of the noise band to be estimated, i.e., the width of the noise band. Furthermore, for example, the second noise signal estimation unit 240 performs processing by changing the method of calculating the time-frequency intensity in the noise band when estimating the intensity of the noise signal. Furthermore, for example, the second determination unit 250 performs processing by changing the method of calculating the score D(t) and the threshold θ.

[0065] The management unit 260 outputs statistical information related to the parameters applied when the second conversion unit 210, the second target signal estimation unit 220, the second band estimation unit 230, the second noise signal estimation unit 240, and the second determination unit 250 performed their processes. For example, the management unit 260 may associate the value of each parameter, the determination result, and information indicating whether the determination result is correct or not, and store this as statistical information in the storage device 300. Furthermore, the management unit 260 may store the average, variance, maximum value, minimum value, etc. of the numerical values of the parameters used when the determination result is correct, in the storage device 300 as statistical information.

[0066] In this way, the learning device 200 outputs statistical information of the parameters used when determining whether or not an event has occurred for a second input signal that includes at least another target signal, in the same manner as the determination means of the signal processing device 100.

[0067] The signal processing device 100 performs each process based on the statistical information stored in the storage device 300. For example, the target signal estimation unit 120 may set the number M for estimating the target signal strength as the maximum value of M included in the statistical information. Also, for example, the band estimation unit 130 may set the minimum value of the noise band width included in the statistical information as the upper limit of the noise band width to be estimated. Also, for example, the determination unit 150 may set the threshold θ as the maximum value of the threshold θ included in the statistical information.

[0068] In this way, the signal processing system 1000 of the second embodiment outputs statistical information of parameters used when determining whether or not an event has occurred for a second input signal including at least another target signal, similarly to the determination unit 150 of the signal processing device 100. Then, the signal processing system 1000 sets parameters used when determining whether or not an event has occurred for the input signal, based on the statistical information. This allows the signal processing system 1000 to determine whether or not an event has occurred for an unknown input signal, for example, using parameters with high determination accuracy.

[0069] [Variations] In the second embodiment, an example has been described in which the learning device 200 is a device separate from the signal processing device 100. The signal processing device may perform the same operation as the learning device 200. FIG. 8 is a block diagram showing an example of the functional configuration of a signal processing device 101 according to a modified example. As shown in FIG. 8, the signal processing device 101 is communicably connected to a storage device 300. Note that the storage device 300 and the signal processing device 101 may be integrated into one device. The signal processing device 101 includes a conversion unit 111, a target signal estimation unit 121, a band estimation unit 131, a noise signal estimation unit 141, a determination unit 151, and a management unit 260. The conversion unit 111, the target signal estimation unit 121, the band estimation unit 131, the noise signal estimation unit 141, and the determination unit 151 have the same functions as the conversion unit 110, the target signal estimation unit 120, the band estimation unit 130, the noise signal estimation unit 140, and the determination unit 150, respectively. In addition, the conversion unit 111, the target signal estimation unit 121, the band estimation unit 131, the noise signal estimation unit 141, and the judgment unit 151 each have the same functions as the second conversion unit 210, the second target signal estimation unit 220, the second band estimation unit 230, the second noise signal estimation unit 240, and the second judgment unit 250, respectively.

[0070] <Example of hardware configuration of signal processing device> The hardware constituting the signal processing device of the first and second embodiments described above will be described. Fig. 9 is a block diagram showing an example of the hardware configuration of a computer device that realizes the signal processing device in each embodiment. The signal processing device and signal processing method described in each embodiment and modification are realized in a computer device 10. Note that the learning device and storage device described in the second embodiment may each have the hardware configuration shown in Fig. 9.

[0071] 9, a computer device 10 includes a processor 11, a RAM (Random Access Memory) 12, a ROM (Read Only Memory) 13, a storage device 14, an input / output interface 15, a bus 16, and a drive device 17. Note that the signal processing device may be realized by a plurality of electric circuits.

[0072] The storage device 14 stores a program (computer program) 18. The processor 11 executes the program 18 of the present signal processing device using the RAM 12. Specifically, for example, the program 18 includes a program that causes a computer to execute the processing of the signal processing device described in each embodiment, such as FIGS. 2 and 6 . The processor 11 executes the program 18 to realize the functions of each component of the present signal processing device. The program 18 may be stored in the ROM 13. Alternatively, the program 18 may be recorded on a storage medium 20 and read out using the drive device 17, or may be transmitted to the computer device 10 from an external device (not shown) via a network (not shown).

[0073] The input / output interface 15 exchanges data with peripheral devices (such as a keyboard, a mouse, and a display device) 19. The input / output interface 15 functions as a means for acquiring or outputting data. The bus 16 connects the various components.

[0074] There are various variations in the implementation of the signal processing device. For example, the signal processing device can be implemented as a dedicated device. Alternatively, the signal processing device can be implemented based on a combination of multiple devices.

[0075] The scope of each embodiment also includes a processing method for recording a program for realizing each component of the function of each embodiment on a storage medium, reading the program recorded on the storage medium as code, and executing it on a computer. That is, a computer-readable storage medium is also included in the scope of each embodiment. Furthermore, the storage medium on which the above-mentioned program is recorded and the program itself are also included in each embodiment.

[0076] The storage medium may be, but is not limited to, a floppy disk, a hard disk, an optical disk, a magneto-optical disk, a CD (Compact Disc)-ROM, a magnetic tape, a non-volatile memory card, or a ROM. The programs recorded on the storage medium are not limited to standalone programs that execute processes, but also include programs that run on an OS (Operating System) in cooperation with other software and functions of an expansion board.

[0077] Although the present invention has been described above with reference to the embodiments, the present invention 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 invention within the scope of the present invention.

[0078] Some or all of the above-described embodiments can be described as, but are not limited to, the following supplementary notes.

[0079] <Additional Notes> [Appendix 1] A conversion means for converting an input signal into a predetermined signal which is a signal in the time-frequency domain; a target signal estimation means for estimating a peak of the time frequency intensity of the predetermined signal as the intensity of a target signal that is a signal related to the occurrence of an event; a band estimation means for estimating a band including at least a band from a frequency associated with the peak to a predetermined frequency and not including a frequency associated with a peak different from the peak, as a noise band, which is a frequency band of a noise signal; a noise signal estimation means for estimating the intensity of the noise signal based on the time frequency intensity in the noise band; a determination unit that determines whether an event has occurred based on the ratio between the intensity of the target signal and the intensity of the noise signal. Signal processing device.

[0080] [Appendix 2] if the minimum frequency of the noise band is less than the frequency associated with the peak and the maximum frequency of the noise band is greater than the frequency associated with the peak, the noise signal estimation means estimates the intensity of the noise signal based on the intensity at a frequency in a band excluding a peak frequency, which is a frequency corresponding to the peak, from the noise band; 2. The signal processing device of claim 1.

[0081] [Appendix 3] the noise signal estimation means estimates the intensity of the noise signal based on the intensity at a frequency in a band excluding a frequency band associated with the peak from the noise band; 3. The signal processing device of claim 2.

[0082] [Appendix 4] the determination means calculates a ratio between the intensity of the target signal and the intensity of the noise signal for each of the peaks, and determines that an event has occurred if a score calculated based on the ratio is greater than a threshold value. 4. A signal processing device according to any one of claims 1 to 3.

[0083] [Appendix 5] the determination means weights the ratio for each peak according to the magnitude of the intensity of the target signal calculated for each peak, and determines the average of the weighted ratios for each peak as the score. 5. The signal processing device of claim 4.

[0084] [Appendix 6] the target signal estimation means detects a predetermined number of peaks of the time frequency intensity of the predetermined signal, and estimates each of the detected peaks as the intensity of the target signal; 6. A signal processing device according to any one of claims 1 to 5.

[0085] [Appendix 7] the determination means sets the score calculation method and the threshold value based on statistical information of parameters used in determining whether or not the event has occurred, in the same manner as the determination means, for a second input signal including at least another target signal; 6. The signal processing device according to claim 4 or 5.

[0086] [Appendix 8] the target signal estimation means sets the predetermined number based on statistical information of parameters used when determining whether or not the event has occurred, in the same manner as the determination means, for a second input signal including at least another target signal; 7. The signal processing device according to claim 6.

[0087] [Appendix 9] the band estimation means estimates the noise band based on statistical information of parameters used in determining whether or not the event has occurred, in the same manner as the determination means, for a second input signal including at least another target signal; 9. A signal processing device according to any one of appendices 1 to 8.

[0088] [Appendix 10] a learning device that outputs statistical information of parameters used when determining whether or not the event has occurred, in the same manner as the determination means, for a second input signal that includes at least another target signal; a storage device for storing information output by the learning device; and a signal processing device according to any one of Supplementary Notes 1 to 9. Signal processing system.

[0089] [Appendix 11] Converting the input signal into a predetermined signal that is a time-frequency domain signal; a peak in the time frequency intensity of the predetermined signal is estimated as the intensity of a signal of interest, the signal being related to the occurrence of an event; a bandwidth from a frequency associated with the peak to a predetermined frequency, the bandwidth including at least a bandwidth not including a frequency associated with a peak different from the peak, being a noise bandwidth that is a frequency band of a noise signal; estimating the intensity of the noise signal based on the time-frequency intensity in the noise band; determining whether an event has occurred based on a ratio between the intensity of the target signal and the intensity of the noise signal; Signal processing methods.

[0090] [Appendix 12] if the minimum frequency of the noise band is less than the frequency associated with the peak and the maximum frequency of the noise band is greater than the frequency associated with the peak, estimating the intensity of the noise signal based on the intensity at frequencies in a band excluding a peak frequency, which is a frequency corresponding to the peak, from the noise band; 12. A signal processing method according to claim 11.

[0091] [Appendix 13] estimating the intensity of the noise signal based on the intensity at frequencies in a band excluding a frequency band associated with the peak from the noise band; 13. A signal processing method according to claim 12.

[0092] [Appendix 14] calculating a ratio between the intensity of the target signal and the intensity of the noise signal for each of the peaks, and determining that an event has occurred if a score calculated based on the ratio is greater than a threshold value; 14. A signal processing method according to any one of appendices 11 to 13.

[0093] [Appendix 15] The ratio for each peak is weighted according to the magnitude of the intensity of the target signal calculated for each peak, and the average of the weighted ratios for each peak is set as the score. 15. A signal processing method as recited in claim 14.

[0094] [Appendix 16] detecting a predetermined number of peaks of the time frequency intensity of the predetermined signal, the number of peaks being equal to or less than a predetermined number, and estimating each of the detected peaks as the intensity of the target signal; 16. A signal processing method according to any one of Supplementary Notes 11 to 15.

[0095] [Appendix 17] setting the score calculation method and the threshold value based on statistical information of parameters used in determining whether or not the event has occurred, in the same manner as in the determination process, for a second input signal including at least another target signal; 16. A signal processing method according to claim 14 or 15.

[0096] [Appendix 18] setting the predetermined number based on statistical information of parameters used when determining whether or not the event has occurred, in the same manner as in the determination process, for a second input signal including at least another target signal; 17. A signal processing method according to claim 16.

[0097] [Appendix 19] For a second input signal including at least another target signal, the noise band is estimated based on statistical information of parameters used in determining whether or not the event has occurred, in the same manner as in the determination process. A signal processing method according to any one of Supplementary Notes 11 to 18.

[0098] [Appendix 20] A process of converting an input signal into a predetermined signal that is a time-frequency domain signal; a process of estimating a peak of the time frequency intensity of the predetermined signal as the intensity of a target signal, the target signal being a signal related to the occurrence of an event; a process of estimating a band including at least a bandwidth from a frequency associated with the peak to a predetermined frequency, the bandwidth not including a frequency associated with a peak different from the peak, as a noise band, which is a frequency band of a noise signal; a process of estimating the intensity of the noise signal based on the time frequency intensity in the noise band; and a process of determining whether or not an event has occurred based on the ratio between the intensity of the target signal and the intensity of the noise signal. A computer-readable storage medium.

[0099] [Appendix 21] if the minimum frequency of the noise band is less than the frequency associated with the peak and the maximum frequency of the noise band is greater than the frequency associated with the peak, In the process of estimating the intensity of the noise signal, the intensity of the noise signal is estimated based on the intensity at a frequency in a band excluding a peak frequency, which is a frequency corresponding to the peak, from the noise band. 21. The computer-readable storage medium of claim 20.

[0100] [Appendix 22] In the process of estimating the intensity of the noise signal, the intensity of the noise signal is estimated based on the intensity at a frequency in a band excluding a frequency band associated with the peak from the noise band. 22. The computer-readable storage medium of claim 21.

[0101] [Appendix 23] In the determination process, a ratio between the intensity of the target signal and the intensity of the noise signal is calculated for each of the peaks, and if a score calculated based on the ratio is greater than a threshold, it is determined that an event has occurred. 22. A computer-readable storage medium according to any one of claims 20 to 21.

[0102] [Appendix 24] In the determination process, the ratio for each peak is weighted according to the magnitude of the intensity of the target signal calculated for each peak, and the average of the weighted ratios for each peak is set as the score. 24. The computer-readable storage medium of claim 23.

[0103] [Appendix 25] In the process of estimating the intensity of the target signal, the peaks of the time frequency intensity of the predetermined signal are detected a number of times equal to or less than a predetermined number, and each of the detected peaks is estimated as the intensity of the target signal. 25. A computer-readable storage medium according to any one of claims 20 to 24.

[0104] [Appendix 26] In the determination process, for a second input signal including at least another target signal, the score calculation method and the threshold value are set based on statistical information of parameters used when determining whether or not the event has occurred, similarly to the determination process. 25. The computer-readable storage medium of claim 23 or 24.

[0105] [Appendix 27] In the process of estimating the intensity of the target signal, the predetermined number is set based on statistical information of parameters used in determining whether or not the event has occurred, in the same manner as in the determination process, for a second input signal including at least another target signal. 26. The computer-readable storage medium of claim 25.

[0106] [Appendix 28] In the process of estimating the noise band, the noise band is estimated for a second input signal including at least another target signal based on statistical information of parameters used in determining whether or not the event has occurred, in the same manner as in the process of determining. 28. A computer-readable storage medium according to any one of appendices 20 to 27. [Explanation of symbols]

[0107] 100, 101 signal processing device 110, 111 conversion unit 120, 121 Target signal estimation unit 130, 131 Bandwidth estimation unit 140, 141 Noise signal estimation unit 150, 151 Judgment section

Claims

1. A conversion means for converting an input signal into a predetermined signal which is a signal in the time-frequency domain; a target signal estimation means for detecting a peak in the time frequency intensity of the predetermined signal and determining that peak as the intensity of a target signal that is a signal related to the occurrence of an event; a band estimation means for setting a peak frequency, which is a frequency corresponding to the target peak, and a frequency between the target peak and a frequency corresponding to an adjacent peak as a minimum frequency and a maximum frequency of a noise band, which is a frequency band of a noise signal; a noise signal estimation means for calculating a representative value of the time frequency intensity in the noise band and estimating the representative value as the intensity of the noise signal; a determination means for determining whether an event has occurred based on a ratio between the intensity of the target signal and the intensity of the noise signal, Signal processing device.

2. the noise signal estimation means estimates the intensity of the noise signal based on the intensity at a frequency in a band excluding the peak frequency from the noise band; The signal processing device according to claim 1 .

3. the noise signal estimation means estimates the intensity of the noise signal based on the intensity at frequencies in a band excluding, from the noise band, a band from the base frequency of the rising portion of the peak to the base frequency of the falling portion of the peak. The signal processing device according to claim 2 .

4. the determination means calculates a ratio between the intensity of the target signal and the intensity of the noise signal for each of the peaks, and determines that an event has occurred if a score calculated based on the ratio is greater than a threshold value. The signal processing device according to any one of claims 1 to 3.

5. the determination means weights the ratio for each peak according to the magnitude of the intensity of the target signal calculated for each peak, and determines the average of the weighted ratios for each peak as the score. The signal processing device according to claim 4 .

6. the determination means sets the score calculation method and the threshold value based on statistical information of parameters used in determining whether or not the event has occurred, in the same manner as the determination means, for a second input signal including at least another target signal; the parameters include at least one of the number of peaks to be detected, the minimum frequency and the maximum frequency of the noise band, and the threshold value; 6. The signal processing device according to claim 4 or 5.

7. the target signal estimation means detects a predetermined number of peaks of the time frequency intensity of the predetermined signal, and estimates each of the detected peaks as the intensity of the target signal; the predetermined number is set based on a maximum value of the number of peaks to be detected, which is a parameter used when determining whether or not the event has occurred, in the same manner as the determination means, for a second input signal including at least another target signal.

6. The signal processing device according to claim 4 or 5.

8. the band estimation means sets the noise band based on statistical information of the minimum frequency and the maximum frequency of the noise band, which are parameters used when determining whether or not the event has occurred, in the same manner as the determination means, for a second input signal including at least another target signal; A signal processing device according to any one of claims 1 to 7.

9. Converting the input signal into a predetermined signal that is a time-frequency domain signal; Detecting a peak in the time frequency intensity of the predetermined signal, and determining the peak as the intensity of a signal of interest, which is a signal related to the occurrence of an event; a peak frequency corresponding to the target peak and a frequency between the target peak and a frequency corresponding to a peak adjacent to the target peak are set as a minimum frequency and a maximum frequency of a noise band, which is a frequency band of a noise signal; calculating a representative value of the time frequency intensity in the noise band, and estimating the representative value as the intensity of the noise signal; determining whether an event has occurred based on a ratio between the intensity of the target signal and the intensity of the noise signal; Signal processing methods.

10. A process of converting an input signal into a predetermined signal that is a time-frequency domain signal; A process of detecting a peak in the time frequency intensity of the predetermined signal and determining that peak as the intensity of a target signal, which is a signal related to the occurrence of an event; a process of setting a peak frequency, which is a frequency corresponding to the target peak, and a frequency between the target peak and a frequency corresponding to an adjacent peak, as a minimum frequency and a maximum frequency of a noise band, which is a frequency band of a noise signal; a process of calculating a representative value of the time frequency intensity in the noise band and estimating the representative value as the intensity of the noise signal; and determining whether or not an event has occurred based on the ratio between the intensity of the target signal and the intensity of the noise signal.

Images