DISTURBANCE NOISE ELIMINATING DEVICE, ECHOL EXTINGUISHER DEVICE, AND ANORMAL NOISE DETECTION DEVICE

By aligning acoustic sensors with the bisector of their connecting line segment to the noise source, the device achieves efficient noise cancellation and improved sound clarity in compact designs.

DE112016007079B4Active Publication Date: 2024-11-07MITSUBISHI ELECTRIC CORP
View PDF 8 Cites 0 Cited by

Patent Information

Application Number
DE112016007079
Authority / Receiving Office
DE · DE
Patent Type
Patents
Current Assignee / Owner
Filing Date
2016-07-21
Publication Date
2024-11-07
Estimated Expiration
2036-07-21

AI Technical Summary

Technical Problem

Existing noise cancellation technologies require a large distance between acoustic sensors, increasing device size and limiting their applicability in compact designs.

Method used

A noise canceling device with acoustic sensors arranged such that the direction of the bisector of the line segment connecting two sensors aligns with the noise source direction, using signal processing to amplify and eliminate noise from the target direction.

Benefits of technology

This arrangement achieves high noise cancellation performance with a small sensor distance, enhancing clarity of target sounds and improving echo cancellation and abnormal noise detection accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 00000000_0000_ABST
    Figure 00000000_0000_ABST
Patent Text Reader

Abstract

Noise elimination device (1), comprising: an acoustic sensor arrangement (3) comprising three acoustic sensors (2) arranged at vertices of an isosceles triangle: a first acoustic sensor (2a) located at one of the vertices forming a vertex angle of the isosceles triangle, a second acoustic sensor (2b) located at one of the vertices forming an end of a base opposite the vertex angle, and a third acoustic sensor (2c) located at one of the vertices forming the other end of the base, wherein the first acoustic sensor (2a) is arranged on a bisecting line between two boundaries between a directional region of a target noise source, which is a region in which the target noise source may be present, and a directional region of an interference noise source, which is a region in which the interference noise source (7) may be present; and the second acoustic sensor (2b) and the third acoustic sensor (2c) are arranged such that the boundaries are aligned with a direction of a perpendicular midpoint of a line segment connecting the second acoustic sensor (2b) and the first acoustic sensor (2a), and with a direction of a perpendicular midpoint of a line segment connecting the third acoustic sensor (2c) and the first acoustic sensor (2a), and a noise elimination processing means (4) for using signals output by the acoustic sensors (2) to output a signal obtained by amplifying noise from the target noise source and eliminating the noise from the noise source (7).
Need to check novelty before this filing date? Find Prior Art

Description

Technical area

[0001] The present invention relates to a noise canceling device that amplifies only noise coming from a target direction of noise coming from a plurality of noise sources arranged at different positions and cancels other noise. Background to the state of the art

[0002] A noise canceler removes noise from sound data recorded with an acoustic sensor, such as a microphone, thereby making the target sound (desired sound) more audible. This device can, for example, clarify speech that is difficult to hear due to noise generated by devices such as an air conditioner, or extract only speech from a target speaker when multiple speakers are speaking simultaneously.

[0003] Such a technique can not only help people better understand speech, but can also improve robustness against noise in a speech recognition system or the like. In addition to clarifying human utterances, it can also be used for purposes such as preventing deterioration in detection accuracy due to ambient noise in a facility monitoring system that automatically detects whether the operating noise of a facility contains abnormal noise.

[0004] As a method for eliminating noise, there is a method that forms an acoustic sensor array with multiple acoustic sensors and performs signal processing through software on the observed signals obtained by the respective acoustic sensors, thereby forming a directivity pattern to a target noise source. Since this method can generate a sharp directivity pattern using inexpensive acoustic sensors, such as omnidirectional microphones, it has the advantage of reducing hardware costs. Since it can dynamically change the formed directivity pattern through software, it can also handle situations where a noise source is moving.

[0005] In the above noise elimination method, it is known that the noise elimination performance depends on the arrangement of the acoustic sensors that constitute the acoustic sensor array. In this regard, Non-Patent Literatures 1 and 2 experimentally demonstrate that the noise elimination performance depends on the positional relationship between acoustic sensors, a target noise source, and a noise source. However, Non-Patent Literatures 1 and 2 do not refer to a specific positional relationship between acoustic sensors and noise sources that provides high noise elimination performance.

[0006] Non-Patent Literatures 3 and 4 disclose acoustic sensor arrays for noise cancellation. The noise cancellation methods disclosed in Non-Patent Literatures 3 and 4 perform noise cancellation using a time difference between observed signals obtained from multiple acoustic sensors, so that distances between the acoustic sensors can be reduced. However, they are silent on whether the acoustic sensor arrays disclosed in these literatures have higher noise cancellation performance than other acoustic sensor arrays.

[0007] Patent Literature 1 discloses a specific positional relationship between acoustic sensors, a target sound source, and a noise source for performing noise elimination using a sound volume difference between the acoustic sensors.

[0008] Patent Literature 2 describes a sound recording device that amplifies sound in a specific range and suppresses sound outside the range, and a corresponding method.

[0009] Patent Literature 3 discloses a built-in microphone device which can detect sound in response to a zoom mechanism of a camera and which can suppress noises such as vibration or wind.

[0010] Patent Literature 4 discloses a speech processing method and apparatus for detecting a speech interval of an input speech signal and improving the speech signal by noise suppression. Patent Literature 1: Japanese Patent Application Publication No. 2006-349723 Patent Literature 2: Publication of US Patent Application US 2016 / 0198258 A1 Patent Literature 3: Publication of Japanese Patent Application JP H06 - 54242 A Patent Literature 4: Publication of US Patent Application US 2002 / 0138254 A1

[0011] Non-patent literature 1: Nobuaki Tanaka, Tetsuji Ogawa, Tetsunori Kobayashi, “Spatial filter calibration based on minimization of modified LSD”, Proc. Interspeech 2011, pp. 1761-1764, Aug. 2011.

[0012] Non-patent literature 2: Nobuaki Tanaka, Tetsuji Ogawa, Tetsunori Kobayashi, "Spatial filter calibration based on minimization of modified LSD", lecture document collection of the Acoustical Society of Japan, pp. 33-36, Sep. 2011.

[0013] Non-patent literature 3: Nobuaki Tanaka, Kosuke Hosoya, Tetsuji Ogawa, Tetsunori Kobayashi, “Development of zone enhancement beamformer,” Proceedings of the Acoustical Society of Japan, pp. 153-154, Sep. 2009.

[0014] Non-patent literature 4: Nobuaki Tanaka, Tetsuji Ogawa, Kenzo Akagiri and Tetsunori Kobayashi, “Development of zonal beamformer and its application to robot audition,” Proc. EUSIPCO 2010, pp. 1529-1533, Aug. 2010. Summary of the inventionTechnical problem

[0015] However, in Patent Literature 1, there is a problem that for the purpose of performing noise elimination by means of the noise volume difference between the acoustic sensors, a relatively large distance between the acoustic sensors is required, which increases the size of the device.

[0016] To solve this problem, the present invention aims to achieve high noise elimination performance with a small distance between the acoustic sensors. Solution to the problem

[0017] A noise canceling device according to the present invention comprises: an acoustic sensor array including two or more acoustic sensors, the acoustic sensors being arranged so that a direction of a perpendicular bisector of a line segment connecting two adjacent acoustic sensors of the two or more acoustic sensors coincides with a direction from which noise comes; and noise canceling processing means for using signals output from the acoustic sensors to output a signal obtained by amplifying noise from a target direction. Advantageous effects of the invention

[0018] A noise canceling device according to the present invention accurately estimates a direction from which noise comes, with an acoustic sensor array including two or more acoustic sensors arranged so that a direction of a perpendicular bisector of a line segment connecting two adjacent acoustic sensors of the two or more acoustic sensors coincides with the direction from which the noise comes, and thus can provide high noise canceling performance with a small distance between the acoustic sensors. Short description of the drawings Fig. 1 is a configuration diagram of a noise canceling device 1 according to a first embodiment of the present invention. Fig. 2 is an explanatory diagram regarding an arrangement of microphones 2 according to the first embodiment of the present invention. Fig. 3 is a relationship diagram between an arrival direction θ of noise observed by a microphone pair 6 and a time difference in the first embodiment of the present invention. Fig. 4 is a representation obtained by applying the Fig. 3 is obtained on a circle centered on the microphone pair 6 in the first embodiment of the present invention. Fig. Figure 5 is a representation obtained by plotting observed values ​​of the time difference for each discrete frequency with respect to two noise signals with a time difference between them. Fig. 6 is a configuration diagram of a noise elimination processing means 4 according to the first embodiment of the present invention. Fig. 7 is a hardware configuration diagram of the noise canceling device 1 according to the first embodiment of the present invention. Fig. 8 is another hardware configuration diagram of the noise canceling device 1 according to the first embodiment of the present invention. Fig. 9 is a configuration diagram of an echo canceller 14 according to a second embodiment of the present invention. Fig. 10 is a configuration diagram of an abnormal noise detection device 19 according to a third embodiment of the present invention. Fig. 11A and Fig. 11B are configuration diagrams of the area where a target noise source may be present, the area where a noise source may be present, and an arrangement of microphones 2 in a fourth embodiment of the present invention. Description of the embodiments

[0019] Embodiments of a noise canceling device, an echo canceling device, an abnormal noise detection device, and a noise canceling method according to the present invention will now be described in detail with reference to the drawings. The present invention is not limited to the embodiments. First embodiment

[0020] Fig. 1 is a configuration diagram of a noise canceling device 1 according to a first embodiment of the present invention.

[0021] In this embodiment, the description is made using microphones as a specific example of acoustic sensors; an acoustic sensor pair is described as a microphone pair; an acoustic sensor array is described as a microphone array. However, acoustic sensors in the present invention are not limited to microphones and may be, for example, ultrasonic sensors.

[0022] The noise canceler 1 includes a microphone array 3 with two or more microphones 2 and a noise canceler processing means (or noise canceler processor) 4. Signals (observed signals) of noise observed by the microphones 2 of the noise canceler 1 are output to the noise canceler processing means 4. The noise canceler processing means 4 then outputs an output signal obtained by amplifying only noise from a direction from which a target noise in the observed signals comes and canceling noise in the observed signals to a loudspeaker 5 connected to the noise canceler 1.

[0023] An arrangement of the microphones 2 is now described.

[0024] In Fig. 1, when pairs of adjacent microphones 2 are referred to as microphone pairs 6, the microphones 2 are arranged so that a direction of a perpendicular bisector of a line segment connecting at least one microphone pair 6 coincides with a noise source 7 (coincides with a direction from which noise originates). The observed signals are output from the microphones 2 to the noise cancellation processing means 4. It is assumed that the positional relationship of the noise source 7 to the microphones 2 is known.

[0025] Here, the positional relationship between the microphones 2 and the noise source 7 is described, which maximizes the noise elimination performance when the noise elimination is performed with a pair of microphones 6.

[0026] Fig. 2 is an explanatory diagram of an arrangement of the microphones 2 in the first embodiment of the present invention.

[0027] As in Fig. 2, the direction (in radians) from which sound comes from the sound source is denoted by θ with a direction of a perpendicular bisector of a line segment connecting the microphone pair 6 as a reference.

[0028] Fig. 3 is a relationship diagram between the arrival direction θ of the noise observed by the microphone pair 6 according to the first embodiment of the present invention and a time difference.

[0029] In Fig. 3, grid lines are drawn at regular intervals along a vertical axis representing the time difference, and points are placed at noise arrival directions θ that correspond to the values ​​of the time difference at the grid lines. The noise arrival directions θ at the positions of these points are unevenly spaced, as shown in Fig. 3 shown.

[0030] Fig. 4 is a representation obtained by plotting the Fig. 3 is obtained on a circle centered on the microphone array 3 of the first embodiment of the present invention.

[0031] Fig. Figure 4 shows that for noise arrival directions θ around 0, the unevenly spaced points are densely distributed, and for noise arrival directions θ around ± π / 2, the unevenly spaced points are sparsely distributed. If an observed time difference is increased by an amount corresponding to the grid line spacing in Fig. 3, deviates from the actual time difference, due to, for example, the effect of the noise, one of the two points next to the point in Fig. 4 is calculated according to the actual noise arrival direction as an observation value of the noise arrival direction. In a directional range in Fig. 4, in which the points are densely distributed, the observed value of the noise arrival direction does not deviate greatly even if a deviation in the time difference occurs.

[0032] On the other hand, it can be seen that in a directional range where the points are sparsely distributed, a small deviation in the time difference leads to a large deviation in the observed value of the noise arrival direction. Therefore, when the noise arrival direction is observed using the microphone pair, it can be observed most accurately when the angle of the noise arrival direction is close to 0, that is, when the noise source 7 is located in the direction of the perpendicular bisector of the line segment connecting the microphone pair.

[0033] During noise cancellation, the noise source 7 (the arrival direction of the noise), which is a component to be eliminated, must be carefully observed. Therefore, from the above consideration, it can be seen that the noise cancellation performance is maximized when the noise source 7 is present in the direction of the perpendicular bisector of the line segment connecting the microphone pair 6. The requirement that the direction of the perpendicular bisector of the line segment connecting the microphone pair 6 coincides with the arrival direction of the noise only requires that the arrival direction of the noise be closer to the direction of the perpendicular bisector of the line segment connecting the microphone pair 6 than the arrival direction of the target noise.When the microphone pair 6 is arranged so that the direction of the perpendicular bisector of the line segment connecting the microphone pair 6 coincides with the arrival direction of the noise, the noise to be eliminated can be observed more accurately than in the case where the microphone pair 6 is arranged so that the arrival direction of the target noise is closer to the direction of the perpendicular bisector of the line segment connecting the microphone pair 6 than the arrival direction of the noise.

[0034] Fig. Figure 5 is a representation obtained by plotting observed values ​​of the time difference for each discrete frequency with respect to two noise signals with a time difference between them.

[0035] There is another reason why the positional relationship between the microphone pair 6 and the noise source 7 described above is advantageous for noise elimination. Fig. Figure 5 shows that as the time difference between the two noise signals increases, the variance of the observed time difference increases, preventing the time difference from being accurately determined. This is because when performing a discrete short-time Fourier transform, the boundaries between short-time frames for the two signals do not coincide. Thus, for the microphone pair 6 formed by the two microphones 2, when the time difference between the noise reaching the two microphones 2 is close to 0—that is, when the arrival direction of the noise coincides with the direction of the perpendicular bisector of the line segment connecting the adjacent microphones 2—the time difference can be observed most accurately.

[0036] As described above, from both the viewpoints that even if a time difference occurs between noises, the deviation of the observed value between the two points is small, and that the variance of the time difference between the noise signals is small, by arranging the microphones 2 so that the direction of the perpendicular bisector of the line segment connecting the adjacent microphones 2 coincides with the noise source 7, it is possible to maximize the noise elimination performance with a small distance between the acoustic sensors.

[0037] In a case where the microphones 2 are mounted on a vehicle and observe the speech of a vehicle occupant, when the direction of the noise from the engine, which may be the noise source 7, is known and the engine noise is eliminated, the microphone pair 6 is arranged so that the direction of the perpendicular bisector of the line segment connecting the adjacent microphones 2 coincides with the direction of the engine, which is the noise source 7. By arranging the microphones 2 in this manner, it is possible to maximize the noise elimination performance and amplify the speech of the vehicle occupant with a small distance between the microphones 2.

[0038] Returning to the description of the configuration of Fig. 1, the noise elimination processing means 4 will now be described.

[0039] The noise elimination processing means 4 outputs an output signal obtained from the observed signals input through the microphones 2 by amplifying noise from the arrival direction of the target noise and eliminating noise, to the speaker 5 connected to the noise elimination device 1.

[0040] Generally, to amplify the incoming sound from a specific direction using the microphone array 3, the sound arrival direction is first observed for each time-frequency pair based on time differences between the observed signals obtained by the plurality of microphones 2. Subsequently, a method of applying a filter to the observed signal of an observed sound is employed, wherein the filter is configured to amplify time-frequency components constituting the incoming sound from the target direction in the observed signal and eliminate the other components in the observed signal.

[0041] Fig. 6 is a configuration diagram of the noise elimination processing means 4 according to the first embodiment of the present invention.

[0042] The noise elimination processing means 4 includes a Discrete Fourier Transform (DFT) means (or Discrete Fourier Transform) 41, a band selection means (or band selector) 42, a multiplication means (or multiplier) 43, and an Inverse Discrete Fourier Transform (IDFT) means (or Inverse Discrete Fourier Transform) 44. Although the noise elimination processing means 4 is here provided with the Fig. 6, it does not necessarily have to be the configuration shown in Fig. 6 and may have other configurations.

[0043] For simplicity of description, the following describes a case where the microphone array 3 is formed of two microphones 2. It is easy to extend this to configurations with three or more microphones 2, and the present invention also encompasses cases where the microphone array 3 is formed of three or more microphones 2. It is assumed that the two microphones 2 constituting the microphone array 3 are microphones 2a and 2b, and these two microphones 2 form a microphone pair 6.

[0044] The hardware configuration of the noise canceller 1 will now be described.

[0045] Fig. 7 is a hardware configuration diagram of the noise canceling device 1 of the first embodiment of the present invention.

[0046] The noise canceling device 1 comprises microphones 2 that collect noise; an AD converter 8 that converts output signals of the microphones 2 into digital signals; a memory 9 that stores programs of the DFT means 41, the band selection means 42, the multiplication means 43, and the IDFT means 44 of the noise canceling processing means 4; and a processor 10 that reads the programs of the respective means of the noise canceling processing means 4 from the memory 9 and executes them. An output signal resulting from the processing by the processor 10 is supplied to an output device 11 (e.g., the loudspeaker 5 in Fig. 1) issued. Fig. 7 shows an example in which the microphone array 3 has n microphones 2.

[0047] Fig. 8 is another hardware configuration diagram of the noise canceling device 1 of the first embodiment of the present invention.

[0048] Fig. Figure 8 shows a configuration in which the signal processing of the noise cancellation processing means 4 is implemented by hardware. The noise cancellation device 1 includes microphones 2 that collect noise; an AD converter 8 that converts output signals of the microphones 2 into digital signals; and a processing circuit 12 that stores programs of the DFT means 41, the band selection means 42, the multiplication means 43, and the IDFT means 44 of the noise cancellation processing means 4, and reads and stores the stored programs. An output signal resulting from the processing by the processing circuit 12 is supplied to an output device 11 (e.g., the loudspeaker 5 in Fig. 1) issued. Fig. 8 shows an example in which the microphone array 3 has n microphones 2.

[0049] The operation of the noise elimination device 1 will now be described with reference to Fig. 6 described.

[0050] Noises collected by the microphones 2a and 2b in the noise cancelling device 1 are Fig. 7 or Fig. 8, the AD converter 8 shown in FIG. 8 converts the observed signals into digital form and inputs them to the DFT means 41a and 41b, respectively, as time-domain observed signals. They are then converted into frequency-domain observed signal spectra by the DFT means 41a and 41b and output to the band selection means 42. The band selection means 42 generates a filter configured to amplify only time-frequency components constituting the target noise in the observed signals based on the time differences. Then, the multiplication means 43 multiplies an observed signal spectrum obtained from the filter and the observed signal spectrum obtained from the DFT means 41a or DFT means 41b, thus generating an output signal spectrum. The output signal spectrum is then input to the IDFT means 44 and converted into a time-domain output signal by an inverse discrete Fourier transform, and the converted output signal is output to the speaker 5.

[0051] The operation of the noise elimination device 1 is described in detail below.

[0052] The noise elimination device 1 first arranges the microphones 2 so that the noise source 7 is present in the direction of the perpendicular bisector of the line segment connecting the microphones 2a and 2b, as shown in Fig. 2 (a noise removal step). The noise collected by the microphones 2 arranged in this way has the most accurately observed noise signal; the noise collected by the microphones 2a and 2b is input to the DFT means 41a and 41b of the Fig. 6 shown noise elimination processing means 4 are each input as an observation signal.

[0053] In the DFT means 41a and 41b, the time-domain observation signals obtained from the microphones 2a and 2b at time t are denoted as x1(t) and x2(t), respectively. The DFT means 41a and 41b perform discrete short-time Fourier transforms on the input observation signals x1(t) and x2(t), and output the resulting frequency-domain observation signal spectra X1(ω,τ) and X2(ω,τ), where ω denotes a discrete frequency and τ denotes a short-time frame.

[0054] Then, the band selection means 42 calculates the noise arrival direction for each discrete frequency based on the observed signal spectra X1(ω,τ) and X2(ω,τ) input from the DFT means 41a and 41b. Assuming that the noise source is sufficiently far from the microphone array 3, the noise arrival direction θ(ω,τ) can be calculated based on the phase difference between X1(ω,τ) and X2(ω,τ) as in Equation 1: θ(ω,τ)=arcsin{c2πωdarg(K2(ω,τ)K1(ω,τ))} where c is the speed of the sound, d is the distance between the microphones, and arg is the argument of a complex number.

[0055] The noise arrival direction θ(ω,τ) calculated according to equation 1 is represented as the angle (in radians) defined as 0 at the direction perpendicular to the line segment connecting the microphone pair 6 formed by the microphones 2a and 2b, as shown in Fig. 2 shown.

[0056] Then, based on the noise arrival direction θ(ω,τ), the band selection means 42 generates, for each discrete frequency, a filter b(ω,τ) configured to retain only time-frequency components of the noise arriving from the direction of the target noise as in equation 2, and outputs it to the multiplication means 43. b(ω,τ)={1(θ(ω,τ)∈θ)0(otherwise) where Θ is a set of arrival directions of the target sound.

[0057] Equation 2 creates a filter configured to multiply the time-frequency components of noise coming from the desired directions by 1 as a coefficient and multiply the other noise components by 0. This filter can only amplify the time-frequency components of the target noise contained in the observed signals.

[0058] The multiplication means 43 multiplies the observed signal spectrum X1(ω,τ) of the microphone 2a by the filter b(ω,τ) to generate an output signal spectrum Y(ω,τ) with noise removed, and outputs it to the IDFT means 44.

[0059] Although it is described here that the observed signal spectrum X1(ω,τ) of microphone 2a is multiplied by the filter b(ω,τ), that of microphone 2b or any microphone 2 can be multiplied.

[0060] Then, the IDFT means 44 converts the input output signal spectrum Y(ω, τ) into a time-domain output signal y(t) through an inverse discrete Fourier transform and outputs the converted output signal y(t) to the loudspeaker 5 (a noise removal step). Thus, the loudspeaker 5 outputs a sound in which noise is maximally removed and the target sound is enhanced.

[0061] Although described here as the speaker 5, it may be a headphone, a memory, a hard disk, or the like. When the output destination is a storage medium such as a memory or hard disk, digital data of the noise-corrected sound is stored on the storage medium.

[0062] When the microphone array 3 is formed of three or more microphones 2, it is possible to eliminate noise more precisely, for example, by generating the filter using average values ​​of the noise arrival directions observed by several microphone pairs 6.

[0063] By arranging the microphones 2 as above so that the direction of the perpendicular bisector of the line segment connecting at least one pair of microphones 6 coincides with the noise source 7, the deviation of the observed value when a time difference occurs in the noise to be eliminated is reduced, the variance of the time difference between the observed signals with respect to the noise to be eliminated is reduced, the estimation of the noise arrival direction based on the time difference can be performed with high accuracy, and thus, it is possible to achieve high noise elimination performance. This allows noise to be eliminated precisely compared to other microphone arrangements, which improves the clarity of the target sound. Second embodiment

[0064] The following describes an example in which the microphone assembly 3 described in the first embodiment is applied to an echo canceller 14 that outputs a signal subjected to echo cancelling processing based on a reference signal from a speech receiver 13 that receives speech (telephone speech) from a communication partner.

[0065] The echo canceller 14 described here is used, for example, in a hands-free telephone system to avoid a situation in which telephone speech received by the speech receiver 13 is reproduced by a loudspeaker 15 and the telephone speech reproduced by the loudspeaker 15 penetrates telephone microphones (microphones 2) for a speaking person, so that the telephone speech is repeatedly reproduced like an echo.

[0066] The echo canceller 14 observes the telephone voice output from the speaker 15 and the voice of a speaker at a communication source, and outputs a signal obtained by performing, based on the reference signal from the voice receiver 13, echo cancelling processing on an output signal obtained by removing (noise removing) the telephone voice as noise and amplifying the voice, which is the target sound, of the speaker at the communication source.

[0067] Fig. 9 is a configuration diagram of the echo canceller 14 of a second embodiment of the present invention. The echo canceller 14 includes a microphone array 3 and a noise cancellation processing means 4 configured as described in the first embodiment, and an echo canceller 16.

[0068] The microphone array 3 includes microphones 2 arranged such that the direction of the perpendicular bisector of the line segment connecting at least one microphone pair 6 coincides with the loudspeaker 15 as a noise source that outputs speech received by the speech receiver 13. Observed signals of noise observed by the microphone array 3 are output to the noise cancellation processing means 4.

[0069] The noise removal processing means 4 is configured as described in the first embodiment, and outputs an output signal obtained from the observed signals input from the microphone array 3 by amplifying target noise and removing noise to the echo cancellation means 16.

[0070] The echo canceling means 16 cancels the echo from the output signal input from the noise canceling processing means 4 based on the reference signal from the speech receiver 13, and outputs the echo-cancelled speech signal to a speech transmitter 17 connected to the echo canceling means 14 (echo canceling step).

[0071] The speech transmitter 17 sends the echo-corrected speech signal to the communication partner.

[0072] The echo canceler 14 eliminates noise from the loudspeaker 15 as noise (echo) in the same way as the noise canceler 1 of the first embodiment. At this time, the echo is generally not completely eliminated due to reverberation or other disturbing factors. Thus, the echo canceler 16 cancels the echo remaining in the output signal of the noise canceler processing means 4 using the reference signal obtained from the speech receiver 13. As a method of canceling echo based on the reference signal, the LMS algorithm and the affine projection algorithm are known.

[0073] By arranging the microphones 2 so that the loudspeaker 15, which is a noise source causing echo, is located in the direction of the perpendicular bisector of the line segment connecting the microphone pair 6, and inserting the high-performance noise elimination processing before the echo cancellation processing described above, it is possible to achieve excellent echo cancellation performance and improve the clarity of the target sound compared with conventional echo cancellation devices. Third embodiment

[0074] The following describes an abnormal noise detection device 19 which constantly monitors operating noise of a monitoring target device 18, which is a device to be monitored, with microphones 2 arranged as described in the first embodiment, and automatically detects abnormal noise due to disturbances or malfunctions of the device.

[0075] In continuous monitoring based on the operating noise of a device, the problem is that the operating noise of a neighboring device or other noise acts as noise and impairs the accuracy of detecting abnormal noise. By eliminating the noise and amplifying the operating noise of the monitoring target device 18 as the target noise in the same way as in the first embodiment, it is possible to improve the accuracy of detecting abnormal noise.

[0076] Fig. 10 is a configuration diagram of the abnormal noise detection device 19 of a third embodiment of the present invention.

[0077] The abnormal noise detection device 19 includes a microphone array 3 and a noise elimination processing means 4 constructed as described in the first embodiment, and an abnormal noise detection means (abnormal noise detector) 20.

[0078] The microphone array 3 includes the microphones 2 arranged such that the direction of the perpendicular bisector of the line segment connecting at least one microphone pair 6 coincides with a noise source 7. Observed signals of noise observed by the microphone array 3 are output to the noise cancellation processing means 4. Here, the noise source 7 is, for example, a device other than the monitoring target device 18, which is known to act as the noise source 7, and the microphones 2 are arranged with respect to the position of the noise source 7 as described above.

[0079] The noise removal processing means 4 is configured as described in the first embodiment, and outputs an output signal obtained from the observed signals input from the microphone array 3 arranged as described in the first embodiment by amplifying noise input from the monitoring target device 18 as the target noise and removing noise, to the abnormal noise detection means 20.

[0080] The abnormal noise detection means 20 detects abnormal noise of the monitoring target device 18 from the output signal input from the noise removal processing means 4. Abnormal noise of the monitoring target device 18 is detected, for example, by abnormal noise detection processing described in Japanese Patent Application Publication No. 2010-271073 or 2008-76246 (abnormal noise detection step).

[0081] In this way, it becomes possible to detect abnormal noise in a noise environment, so that it is possible to improve the accuracy in detecting abnormal noise in various environments, compared with a case where the noise elimination processing is performed without employing the arrangement of the microphone array 3 as described in the first embodiment.

[0082] Even if an abnormal noise is detected by the abnormal noise detecting means 20, it is possible to prevent the device from operating in an unstable state for a long period of time, for example, by automatically stopping the monitoring target device 18 or notifying an operator of a malfunction of the device through an alarm, a mail, or the like. Fourth embodiment

[0083] The following describes an arrangement of microphones 2 for accurately eliminating the noise in a situation in which the area in which a target noise source is present and the area in which a noise source is present can move.

[0084] Fig. 11A and Fig. 11B are configuration diagrams of the area where the target noise source may be present, the area where the noise source may be present, and an arrangement of the microphones 2 in a fourth embodiment of the present invention.

[0085] Fig. 11a is an arrangement diagram of the area in which the target noise source may be present, the area in which the noise source may be present, and a microphone arrangement 3. Fig. 11b is an arrangement diagram of the microphone array 3 formed of three microphones 2.

[0086] It is assumed that, as in Fig.11a, the area in which the target noise source may be present (direction area of ​​the target noise source) 21 lies radially around the microphone array 3 over a certain angular range. The area outside the direction area 21 of the target noise source is assumed to be the area in which the noise source may be present (direction area of ​​the noise source) 22; a boundary between the direction area 21 of the target noise source and a direction area 22a of the noise source is referred to as the boundary 23a; a boundary between the direction area 21 of the target noise source and a direction area 22b of the noise source is referred to as the boundary 23b. In this case, multiple target noise sources may be present in the target noise source direction area 21 and multiple noise sources may be present in the noise source direction area 22.

[0087] In such a situation, a microphone 2a of the microphone array 3 is arranged on an angle bisector formed by the two boundaries 23a and 23b of the target noise source directional range. A microphone 2b adjacent to the microphone 2a is arranged such that a direction of a median bisector of a line segment connecting the microphones 2a and 2b coincides with the boundary 23a. Furthermore, a microphone 2c adjacent to the microphone 2a is arranged such that a direction of a perpendicular bisector of a line segment connecting the microphones 2a and 2c coincides with the boundary 23b.

[0088] The microphone 2a corresponds to a first acoustic sensor, the microphone 2b corresponds to a second acoustic sensor and the microphone 2c corresponds to a third acoustic sensor.

[0089] Arranging the microphones 2b and 2c with the microphone 2a as a vertex in this manner results in an arrangement in which the microphone 2a is located at a vertex forming a vertex of an isosceles triangle, the microphone 2b is located at one end of a base opposite the vertex angle of the isosceles triangle, and the microphone 2c is located at the other end of the base. It is assumed that the distance between the microphone array 3 and the target noise source or the distance between the microphone array 3 and the noise source is sufficiently greater than the distances between the microphones 2a, 2b, and 2c. Although the microphone array 3 has been described as being formed of the three microphones 2 arranged as described above, it may be configured to include the three microphones arranged as described above.

[0090] The noise removal processing means 4 performs noise removal and target noise amplification on observed signals from the noise observed by the microphone array 3 arranged as described above, and outputs an output signal as described in the first embodiment. It is also possible to perform noise removal on the observed signals of the microphone array 3 arranged as described above and cancel the echo in the echo canceler 16 as described in the second embodiment. It is further possible to perform noise removal on the observed signals of the microphone array 3 arranged as described above and detect abnormal noise in the abnormal noise detection means 20.

[0091] In this way, for example, the three microphones 2 are arranged at the respective vertices of the isosceles triangle, and the microphone 2a, which is arranged at one of the vertices forming the vertex angle of the isosceles triangle, is arranged on the angle bisector formed by the two boundaries 23a and 23b between the target sound source direction area and the noise source direction area. The microphones 2b and 2c are arranged so that the boundaries between the target sound source direction area and the noise source direction area coincide with the direction of the perpendicular bisector of the line segment connecting the microphone 2a and the microphone 2b, which is arranged at one end of the base of the isosceles triangle, and the direction of the perpendicular bisector connecting the microphone 2a and the microphone 2c, which is arranged at the other end of the base of the isosceles triangle.By arranging the microphones 2 in this way, in a situation where the target noise is most difficult to detect, i.e., in a case where the noise source is present at one of the boundaries where the noise source is closest to the target noise source, the arrival direction of the noise coincides with the direction of the perpendicular bisector of the line segment connecting at least one pair of microphones 6, so that the noise cancellation performance is maximized. Thus, when the noise source does not enter the target noise source direction range, it is possible to achieve stable noise cancellation performance wherever the noise source is present. Commercial applicability

[0092] As explained above, the noise canceling device 1 according to the present invention can accurately estimate the noise arrival direction based on the time difference, with microphones 2 arranged so that the direction of the perpendicular bisector of the line segment connecting at least one microphone pair 6 coincides with the noise source 7, and thus high noise canceling performance can be provided. List of reference symbols

[0093] 1 noise canceling device, 2 microphone, 3 microphone array, 4 noise canceling processing means, 5 loudspeaker, 6 microphone pair, 7 noise source, 8 AD converter, 9 memory, 10 processor, 11 output device, 12 processing circuit, 13 speech receiver, 14 echo canceller, 15 loudspeaker, 16 echo canceller, 17 speech transmitter, 18 monitoring target device, 19 abnormal noise detection device, 20 abnormal noise detection means, 21 target noise source direction area, 22 noise source direction area, 23 boundary between target noise source direction area and noise source direction area.

Claims

[1] Noise elimination device (1), comprising: an acoustic sensor arrangement (3) comprising three acoustic sensors (2) arranged at vertices of an isosceles triangle: a first acoustic sensor (2a) arranged at one of the vertices forming a vertex angle of the isosceles triangle, a second acoustic sensor (2b) arranged at one of the vertices being one end of a base opposite the vertex angle, and a third acoustic sensor (2c) arranged at one of the vertices being the other end of the base, wherein the first acoustic sensor (2a) is arranged on a bisecting line between two boundaries between a directional range of a target noise source, which is a range in which the target noise source may be present, and a directional range of an interference noise source, which is a range in which the interference noise source (7) may be present; and the second acoustic sensor (2b) and the third acoustic sensor (2c) are arranged such that the boundaries coincide with a direction of a perpendicular bisector of a line segment connecting the second acoustic sensor (2b) and the first acoustic sensor (2a), and a direction of a perpendicular bisector of a line segment connecting the third acoustic sensor (2c) and the first acoustic sensor (2a), and a noise elimination processing means (4) for using signals output from the acoustic sensors (2) to output a signal obtained by amplifying noise from the target noise source and eliminating the noise from the noise source (7). [2] Echo cancellation device (14) comprising: an acoustic sensor arrangement (3) comprising three acoustic sensors (2) arranged at vertices of an isosceles triangle: a first acoustic sensor (2a) arranged at one of the vertices forming a vertex angle of the isosceles triangle, a second acoustic sensor (2b) arranged at one of the vertices being one end of a base opposite the vertex angle, and a third acoustic sensor (2c) arranged at one of the vertices being the other end of the base, wherein the first acoustic sensor (2a) is arranged on a bisecting line between two boundaries between a directional range of a target noise source, which is a range in which the target noise source may be present, and a directional range of an interference noise source, which is a range in which the interference noise source (7) may be present, wherein the noise source outputs speech received by a speech receiver (13) receiving speech from a communication partner; and the second acoustic sensor (2b) and the third acoustic sensor (2c) are arranged such that the boundaries coincide with a direction of a perpendicular bisector of a line segment connecting the second acoustic sensor (2b) and the first acoustic sensor (2a) and a direction of a perpendicular bisector of a line segment connecting the third acoustic sensor (2c) and the first acoustic sensor (2a), and a noise elimination processing means (4) for using signals output from the acoustic sensors (2) to output a signal obtained by amplifying noise from the target noise source and eliminating the noise from the noise source (7); and an echo canceller (16) for cancelling echo of the output signal output from the noise canceller processing means (4) based on a reference signal from the speech receiver (13). [3] Abnormal noise detection device (19) comprising: an acoustic sensor arrangement (3) comprising three acoustic sensors (2) arranged at vertices of an isosceles triangle: a first acoustic sensor (2a) arranged at one of the vertices forming a vertex angle of the isosceles triangle, a second acoustic sensor (2b) arranged at one of the vertices being one end of a base opposite the vertex angle, and a third acoustic sensor (2c) arranged at one of the vertices being the other end of the base, wherein the first acoustic sensor (2a) is arranged on a bisecting line between two boundaries between a directional range of a target noise source, which is a range in which the target noise source may be present, and a directional range of an interference noise source, which is a range in which the interference noise source (7) may be present; and the second acoustic sensor (2b) and the third acoustic sensor (2c) are arranged such that the boundaries coincide with a direction of a perpendicular bisector of a line segment connecting the second acoustic sensor (2b) and the first acoustic sensor (2a) and a direction of a perpendicular bisector of a line segment connecting the third acoustic sensor (2c) and the first acoustic sensor (2a), and; a noise elimination processing means (4) for using signals output from the acoustic sensors (2) to output a signal obtained by amplifying noise from the target noise source and eliminating the noise from the noise source (7); and an abnormal noise detection means (20) for detecting abnormal noise from a device to be monitored, which is the target noise source, on a basis of the output signal output from the noise elimination processing means (4).

Citation Information

Patent Citations

  • Abnormal sound detecting device and abnormal sound detection method

    JP2008076246A

  • Diagnosis device of abnormality in equipment

    JP2010271073A

  • Microphone device

    JP1994054242A

  • Acoustic model creating device, method, and program, speech recognition device, method, and program, and recording medium

    JP2006349723A

  • Method and apparatus for processing speech signals

    US20020138254A1