Close-talking microphone array and method for setting the close-talking microphone array

Abstract

To provide a close-talking microphone array and a setting method for the close-talking microphone array capable of preferentially detecting sound from a near sound source.SOLUTION: The sensitivity and terminal polarity of each of a plurality of microphones 20-1, 20-2, 20-3, ..., and 20-N are set on the basis of the calculated eigenvalues and eigenvectors calculated from a generalized eigenvalue problem for a matrix whose elements are real parts of the specific acoustic impedance on the surface of a microphone array 20, which is a set of the plurality of microphones 20-1, 20-2, 20-3, ..., and 20-N, and a matrix whose elements are the imaginary parts of the specific acoustic impedance.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a close-talking microphone array and a method for configuring a close-talking microphone array. [Background technology]

[0002] When talking on a mobile phone in a noisy place, the surrounding noise mixes with the speaker's voice, making it difficult for the other party to hear the speaker, so technology has been proposed to remove noise from voice. For example, there is signal processing in which the frequency spectrum of noise is estimated from a microphone measurement signal containing noise mixed into the voice, and the estimated frequency spectrum value is subtracted from the microphone measurement signal. In such signal processing, the noise is removed from the voice by subtracting the "estimated value of noise" from the microphone measurement signal, which is "voice + noise."

[0003] However, it is generally not easy to estimate the frequency spectrum of noise from a signal containing a mixture of speech and noise, and estimation errors may occur. If an estimated value including estimation errors is subtracted from the measurement signal of the microphone, the speech may be distorted and the clarity of the call may deteriorate. In addition, if the deduction of noise components from the measurement signal of the microphone is suppressed for fear of deteriorating the clarity of the call, there is a risk that the noise may not be sufficiently removed. Therefore, when the noise components are subtracted from the measurement signal of the microphone to clarify the speech, the technical challenge is to improve the estimation accuracy of the frequency spectrum of noise.

[0004] Patent Document 1 discloses an invention relating to a microphone array that estimates the distance between a microphone and each sound source based on the fact that the direct-indirect ratio, which is the ratio of direct sound to indirect sound (reverberation sound) contained in the received sound, changes monotonically depending on the distance between the microphone and the sound source, and removes noise by filtering to emphasize or suppress the amplitude of the components of the sound source that are determined to be within a certain distance range. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] JP 2011-53062 A Summary of the Invention [Problem to be solved by the invention]

[0006] However, the invention described in Patent Document 1 has a problem in that it requires estimating the distance from the microphone to each sound source and emphasizing or suppressing the amplitude of the signal related to each sound source depending on the estimated distance, which is a complicated process.

[0007] An object of the present invention is to provide a close-talking microphone array and a setting method for the close-talking microphone array that are capable of detecting sounds from nearby sound sources with priority. [Means for solving the problem]

[0008] In order to solve the above problem, the invention described in claim 1 sets the sensitivity and terminal polarity of each of the multiple microphones based on eigenvalues ​​and eigenvectors calculated from a generalized eigenvalue problem relating to a matrix whose elements are the real part of the specific acoustic impedance on the surface of a microphone array, which is a collection of multiple microphones, and a matrix whose elements are the imaginary part of the specific acoustic impedance.

[0009] According to the present invention, the sensitivity and terminal polarity of each of a plurality of microphones can be optimized based on the eigenvalues ​​and eigenvectors calculated from a generalized eigenvalue problem.

[0010] As described in claim 2, the sensitivity and terminal polarity of each of the multiple microphones may be set based on the element of the eigenvector for which the eigenvalue is maximum when the product of a matrix whose elements are the imaginary part of the specific acoustic impedance and a real matrix whose columns are the eigenvectors is equal to the product of a matrix whose elements are the real part of the specific acoustic impedance, a real matrix whose columns are the eigenvectors, and a diagonal matrix whose elements are the eigenvalues.

[0011] As described in claim 3, the sensitivity and terminal polarity of each of the multiple microphones may be set based on the element of the eigenvector for which the eigenvalue is the smallest when the product of a matrix whose elements are the real part of the specific acoustic impedance and a real matrix whose columns are the eigenvectors is equal to the product of a matrix whose elements are the imaginary part of the specific acoustic impedance, a real matrix whose columns are the eigenvectors, and a diagonal matrix whose elements are the eigenvalues.

[0012] In the invention described in claim 4, the plurality of microphones are arranged in a line.

[0013] In the invention described in claim 5, the plurality of microphones are disposed at intervals equal to or less than a half wavelength of an upper limit frequency that each of the plurality of microphones is to measure.

[0014] The method for setting a close-talking microphone array, which is the invention described in claim 6, includes the steps of calculating eigenvalues ​​and eigenvectors from a generalized eigenvalue problem relating to a matrix whose elements are the real part of the specific acoustic impedance on the surface of the microphone array, which is a collection of multiple microphones, and a matrix whose elements are the imaginary part of the specific acoustic impedance, and setting the sensitivity and terminal polarity of each of the multiple microphones based on the calculated eigenvalues ​​and eigenvectors. Effect of the Invention

[0015] The present invention has an advantage in that it is possible to provide a close-talking microphone array and a setting method for the close-talking microphone array that are capable of detecting the sound of a nearby sound source with priority. [Brief description of the drawings]

[0016] [Figure 1] FIG. 2 is a block diagram showing an example of the configuration of a microphone array. [Diagram 2] FIG. 1 is a block diagram showing a microphone array with five microphones. [Diagram 3](A) is a layout diagram for two microphones, (B) is a layout diagram for three microphones, (C) is a layout diagram for four microphones, and (D) is a layout diagram for five microphones. [Figure 4] FIG. 13 is an explanatory diagram showing the change in eigenvalue with respect to frequency when (A) there are two microphones, (B) there are three microphones, (C) there are four microphones, and (D) there are five microphones. [Diagram 5] FIG. 13 is an explanatory diagram showing an example of the change in the first-order eigenvector with respect to frequency when (A) there are two microphones, (B) there are three microphones, (C) there are four microphones, and (D) there are five microphones. [Figure 6] FIG. 13 is an explanatory diagram showing an example of the sensitivity distribution of a microphone array in which two microphones are arranged in a line, where (A) is a frequency of 300 Hz, (B) is a frequency of 1850 Hz, and (C) is a frequency of 3400 Hz. [Figure 7] FIG. 13 is an explanatory diagram showing an example of the sensitivity distribution of a microphone array in which three microphones are arranged in a line, where (A) is a frequency of 300 Hz, (B) is a frequency of 1850 Hz, and (C) is a frequency of 3400 Hz. [Figure 8] FIG. 13 is an explanatory diagram showing an example of the sensitivity distribution of a microphone array in which four microphones are arranged in a line, where (A) is a frequency of 300 Hz, (B) is a frequency of 1850 Hz, and (C) is a frequency of 3400 Hz. [Figure 9] FIG. 13 is an explanatory diagram showing an example of the sensitivity distribution of a microphone array in which five microphones are arranged in a line, where (A) is a frequency of 300 Hz, (B) is a frequency of 1850 Hz, and (C) is a frequency of 3400 Hz. [Figure 10] FIG. 2 is an explanatory diagram showing an example of the sensitivity distribution of one microphone. [Figure 11]FIG. 13 is an explanatory diagram showing an example of sensitivity distribution in a secondary gradient microphone composed of four microphones, where (A) is a frequency of 300 Hz, (B) is a frequency of 1850 Hz, and (C) is a frequency of 3400 Hz. [Figure 12] FIG. 1A is an explanatory diagram showing the sensitivity characteristics versus frequency of one microphone, and FIG. 1B is an explanatory diagram showing the sensitivity characteristics versus frequency of a microphone array in which five microphones are arranged in a line. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0017] An embodiment of the present invention will be described below. In this embodiment, a case where the present invention is applied to a close-talking microphone array will be described.

[0018] In this embodiment, instead of removing noise components from the measurement signal of the microphone by post-processing, an approach is proposed to configure a microphone that is less likely to pick up noise. In close-talking communication devices such as mobile phones and headsets, the distance between the speaker's mouth and the microphone is close, whereas the distance between the microphone and surrounding noise sources is far. Therefore, if a microphone measures sounds emitted from nearby sound sources with high sensitivity and sounds emitted from distant sound sources with low sensitivity, the speaker's voice can be preferentially detected.

[0019] An example of a technology that has already been put to practical use is the gradient microphone (differential microphone). A gradient microphone arranges two microphones and detects the difference between the measurement signals of both microphones. The detected difference is proportional to the gradient of the sound pressure. The gradient of the sound pressure, rather than the sound pressure itself, is a sensitivity characteristic that emphasizes nearby sound sources. Furthermore, by preparing more microphones and obtaining a higher-order gradient, nearby sound sources can be emphasized even further. However, while it is true that the gradient of the sound pressure emphasizes nearby sound sources, it does not necessarily emphasize nearby sound sources to the maximum extent.

[0020] In this embodiment, eigenvalues ​​and eigenvectors related to the specific acoustic impedance of the surface of the microphone array are calculated, and the amplitude (hereinafter referred to as "sensitivity") indicating the sensitivity of each microphone that makes up the microphone array and the phase (hereinafter referred to as "polarity") indicating the electrical polarity of the terminals of each microphone are set based on the calculated eigenvalues ​​and eigenvectors, thereby effectively detecting sound from nearby sound sources while making it difficult to pick up ambient noise.

[0021] FIG. 1 is an explanatory diagram showing a configuration of an audio input device 10 according to the present embodiment. The audio input device 10 includes a microphone array 20 and an amplifier array 30. The microphone array 20 includes N microphones 20-1, 20-2, 20-3, ..., 20-N arranged in a line. Here, N is an integer of 2 or more. As a preparation for driving the microphone array 20, a predetermined eigenvector of the microphone array 20 is assumed to have been calculated. The eigenvector has N vector elements corresponding to the sensitivity and polarity of each of the N microphones 20-1, 20-2, 20-3, ..., 20-N. For example, the i-th vector element of the N vector elements of the eigenvector indicates the sensitivity and polarity of the i-th microphone 20-i of the microphone array 20. Here, i is an integer of 1 to N. The amplifier array 30 includes N amplifiers 30-1, 30-2, 30-3, ..., 30-N. The i-th amplifier 30-i in the amplifier array 30 controls the sensitivity and polarity of the i-th microphone 20-i in the microphone array 20 in accordance with the i-th vector element of the N vector elements of the eigenvector.

[0022] In this embodiment, the sensitivity and polarity of each of the microphones 20-1, 20-2, 20-3, ..., 20-N are adjusted by the amplifier array 30, but it may also be possible to arbitrarily set the sensitivity and polarity of each of the microphones 20-1, 20-2, 20-3, ..., 20-N without using the amplifier array 30.

[0023] Each of the microphones 20-1, 20-2, 20-3, . . . , 20-N is any one of a dynamic microphone, a condenser microphone, a carbon microphone, and a piezoelectric microphone.

[0024] First, consider a speaker array instead of a microphone array. The sound (complex sound power) emitted from a speaker array is divided into a component that reaches a long distance (the real part of the complex sound power) and a component that remains in a short distance (the imaginary part of the complex sound power). The real part of the complex sound power is called active sound power. On the other hand, the imaginary part of the complex sound power is called reactive sound power. If the ratio of reactive sound power to active sound power can be increased, sound can be delivered only to nearby places. The eigenvalue of the generalized eigenvalue problem described later corresponds to the ratio of reactive sound power to active sound power. The eigenvector of the generalized eigenvalue problem described later corresponds to the amplitude and polarity of each speaker that constitutes the speaker array. In other words, if the amplitude and polarity of each speaker are set according to the eigenvector with the largest eigenvalue, a speaker array that easily delivers sound only to nearby places can be realized. Due to reciprocity, this idea can be applied to the design of microphone arrays. In other words, if the sensitivity and polarity of each microphone are set according to the eigenvector with the maximum eigenvalue, a microphone array that can easily measure only sounds emitted from nearby sound sources can be realized.

[0025] Here, assume that N speakers are used instead of microphones. Let p be a complex sound pressure vector, which is a column vector with the complex sound pressure on the surface of each of the N speakers as vector elements, v be a complex velocity vector, which is a column vector with the complex velocity (vibration velocity in the normal direction of the vibration surface) of each of the N speakers as vector elements, and Z be a specific acoustic impedance matrix with the specific acoustic impedance of the speaker surface as a matrix element, and the following formula (1) is established. The specific acoustic impedance matrix Z is a symmetric matrix due to acoustic reciprocity. Here, if p is a vector with N rows and 1 column, and v is a vector with N rows and 1 column, Z is an N-row, N-column matrix. The i-th row and j-th column element of Z is the specific acoustic impedance between the i-th speaker surface and the j-th speaker surface. JPEG0007671497000001.jpg1583

[0026] The complex sound power W is expressed by the following formula (2). In formula (2), s is the area of ​​the vibration surface of each speaker. For convenience of explanation, formula (2) assumes that the area of ​​the vibration surface of each speaker is the same, but the area of ​​the vibration surface of each speaker does not necessarily have to be the same. Also, v H is the conjugate transpose of the complex velocity vector v. JPEG0007671497000002.jpg24133

[0027] Complex sound power W is expressed as follows: r and the imaginary part W i In the formula below, W r denotes the active sound power, and W i denotes the reactive sound power. JPEG0007671497000003.jpg1367

[0028] The specific acoustic impedance matrix Z has a real part Z r and the imaginary part Z i In the formula below, Z r denotes the acoustic resistance matrix, and Z i denotes the acoustic reactance matrix. JPEG0007671497000004.jpg1362

[0029] Therefore, the real part W of the complex sound power W r and the imaginary part W i is expressed as the following equations (3) and (4). JPEG0007671497000005.jpg41105

[0030] In a speaker array, the active acoustic power W r Reactive sound power W i Ratio (W i / W rBy increasing the RA ratio, it is possible to obtain a configuration suitable for listening in the near field. As described below, the RA ratio corresponds to an eigenvalue in a generalized eigenvalue problem, and the sensitivity and polarity of each microphone constituting the microphone array 20 are set based on the eigenvector that maximizes the eigenvalue.

[0031] As shown in the following equation (5), a generalized eigenvalue problem for the acoustic resistance matrix and the acoustic reactance matrix is ​​defined. In the following equation, Λ is a diagonal matrix whose elements are eigenvalues, and Φ is a real matrix whose columns are eigenvectors. In addition, the acoustic resistance matrix Z r is a real, positive definite, symmetric matrix, and the acoustic reactance matrix Z i is a symmetric matrix that is a real number. In the generalized eigenvalue problem, the eigenvalues ​​are real numbers and the eigenvectors are real vectors. The generalized eigenvalue problem is expressed by the following equation (5): i and the acoustic resistance matrix Z r However, the acoustic reactance matrix Z i and the acoustic resistance matrix Z r The eigenvalue of the generalized eigenvalue problem when the eigenvalues ​​are replaced with the eigenvalues ​​of the RA ratio in the original equation (5) is the inverse of the RA ratio in the original equation (5). Therefore, in such a case, the sensitivity and polarity of each microphone constituting the microphone array 20 are set based on the eigenvector when the eigenvalue is minimized. JPEG0007671497000006.jpg1383

[0032] When solving the generalized eigenvalue problem, the acoustic resistance matrix Z r First, we solve the standard eigenvalue problem for . The diagonal matrix Λ has real, positive eigenvalues. r and an orthogonal matrix Φ whose columns are real eigenvectors. r satisfies the following formula (6). JPEG0007671497000007.jpg1888

[0033] In the above equation (6), ΦT r is Φ r is the transpose of Φ r Λ r -1 / 2 By using the acoustic reactance matrix Z i is converted into the following equation (7). JPEG0007671497000008.jpg30137

[0034] In the above formula (7), -1 / 2 means the reciprocal of the square root of the diagonal element. Z i ' is a real symmetric matrix. Then, Z i Solve the standard eigenvalue problem for '. Given a diagonal matrix Λ with real eigenvalues ​​arranged on each diagonal element, and an orthogonal matrix Φ with real eigenvectors arranged on each column. i satisfies the following formula (8). JPEG0007671497000009.jpg2086

[0035] When Φ, which is a real matrix, is defined as in the following equation (9), diagonalization as shown in equations (10) and (11) becomes possible. JPEG0007671497000010.jpg2298 JPEG0007671497000011.jpg2993

[0036] In equation (10), I is a unit matrix. Multiplying both sides of equation (10) by Λ from the right and comparing the result with equation (11) gives equation (5). By diagonalizing equations (10) and (11), the complex sound power W can be written as equation (13).

[0037] In a speaker array consisting of multiple speakers, there are multiple eigenvalues ​​and eigenvectors. m When designing a speaker array based only on , the complex velocity vector v is expressed by the following equation (12). m is an arbitrary coefficient. JPEG0007671497000012.jpg1591

[0038] In this case, the complex acoustic power W is expressed by the following equation (13). JPEG0007671497000013.jpg18131

[0039] m-th eigenvalue λ m is the active sound power W r Reactive sound power W i The RA ratio is the ratio of the eigenvector φ m means the amplitude and phase of each speaker (this eigenvector is also called the generalized radiation mode). Therefore, to maximize the RA ratio, the amplitude and phase of each speaker are set according to the eigenvector with the largest eigenvalue. Note that the eigenvalue λ m Also, eigenvector φ m is also a real number, so setting the phase simply requires setting the positive or negative electrical polarity of the speaker terminals.

[0040] We will now apply the theory of loudspeaker arrays described above to close-talking microphone arrays. First, considering equation (12) for loudspeaker arrays, to maximize the RA ratio, the amplitude and phase of each loudspeaker are set according to the eigenvector with the largest eigenvalue.

[0041] Here, at any position r o The complex sound pressure p(r o ) is expressed by the following equation (14). In the following equation (14), z is a row vector of the specific acoustic impedance, which is the acoustic transfer function between the sound source and the observation point. The sound pressure level is expected to be high near the speaker array and low far away. JPEG0007671497000014.jpg20116

[0042] Next, if we consider the sound source and the observation point interchangeably, the following equation (15) holds due to reciprocity. o )=Sv is r o The complex volume velocity (m 3 / s), where v is the complex velocity corresponding to the surface area of ​​the source, S. JPEG0007671497000015.jpg3896

[0043] In equation (15), p is a column vector whose vector elements are the complex sound pressures observed by the microphones that make up the microphone array. Then, a filter process represented by the following equation (16) is performed. JPEG0007671497000016.jpg26126

[0044] φ on the left side of equation (16) T m p is φ T m z T is proportional to and φ T m z T =zφ m Furthermore, from equation (14), the complex sound pressure p(r o ) is zφ m Therefore, the sensitivity of a microphone array is expected to be higher when the microphone array is closer to the sound source and lower when it is farther away.

[0045] In this embodiment, the acoustic resistance matrix Z r and the acoustic reactance matrix Z i The sensitivity and polarity of each microphone constituting the microphone array are determined by solving a generalized eigenvalue problem related to the optimal filter φ T m is obtained as the eigenvector with the maximum eigenvalue. The specific acoustic impedance depends on the frequency, and the eigenvalue and eigenvector depend on the frequency.

[0046] In this embodiment, the eigenvector associated with the maximum eigenvalue indicates the sensitivity (amplitude) and polarity (phase) of each microphone constituting the microphone array. As described above, since the eigenvalues ​​and eigenvectors are composed of real numbers, the phase of each microphone can be adjusted by simply setting either a positive or negative sign.

[0047] Next, the operation of the microphone array 20 according to the present embodiment will be described. In the present embodiment, a simulation model is used to analyze the characteristics of the microphone array 20. In the simulation, the specific acoustic impedance z of the surface of the microphone array 20 is calculated as the specific acoustic impedance of a speaker array surface having the same geometric shape as the microphone array 20 by the following formula (17). JPEG0007671497000017.jpg26113

[0048] In the above equation (17), ρ is the air density, ρ=1.21kg / m 3 where k is the wave number, k = ω / c, ω is the angular frequency, c is the speed of sound in air, c = 340 m / s, and r is the distance between the sound source and the observation point.

[0049] As described above, the i-th row, j-th column element of the specific acoustic impedance matrix Z is the specific acoustic impedance between the i-th speaker surface and the j-th speaker surface. z on the left side of equation (17) indicates a matrix element of the specific acoustic impedance matrix Z. r on the right side of equation (17) is the distance between the i-th speaker surface and the j-th speaker surface. When calculating the off-diagonal elements (i ≠ j) of matrix Z, r ≠ 0, so equation (17) having r in the denominator on the right side can be applied. However, when calculating the diagonal elements (i = j) of matrix Z, r = 0, so the calculation result diverges in equation (17) having r in the denominator on the right side. In such a case, z is calculated using the following equation (18) instead of the above equation (17).

[0050] In a special case, the equation (18) for calculating the specific acoustic impedance z at r = 0 is as follows. In equation (18), l p is the length of one side of the microphone diaphragm, which is considered to be a square, and S=l p 2 (As an example, p =0.005m). JPEG0007671497000018.jpg25164

[0051] Fig. 2 is an explanatory diagram showing an example of a microphone array 20 according to an embodiment of the present invention. The microphone array 20 shown in Fig. 2 is a simulation model for performing numerical analysis of eigenvalues ​​and eigenvectors, and in this example, the microphone array 20 includes five microphones 20-1, 20-2, ..., 20-5 arranged on the X-axis of an XY plane (infinite baffle) in an XYZ space.

[0052] The distance between each microphone in the X direction is L x As will be described later, in the numerical analysis simulation, L x =0.015m.

[0053] Fig. 3 is an explanatory diagram showing an example of the configuration of microphone array 20 in which microphones are arranged linearly in a simulation in this embodiment. Fig. 3(A) shows a case where there are two microphones, Fig. 3(B) shows a case where there are three microphones, Fig. 3(C) shows a case where there are four microphones, and Fig. 3(D) shows a case where there are five microphones.

[0054] In all cases of Figures 3(A) to (D), the center of the microphone array coincides with the origin O of the x-axis. The distance between adjacent microphones is 0.015 m, which is half the wavelength of the sound wave at 3400 Hz, which is the upper limit frequency to be measured. In general, the frequency band for voice communication is from 300 Hz to 3400 Hz, and 3400 Hz corresponds to the upper limit of that frequency band.

[0055] Fig. 4 is an explanatory diagram showing the change in eigenvalue with respect to frequency of microphone array 20 in which microphones are arranged linearly in a simulation of this embodiment. Fig. 4(A) shows the case of two microphones, Fig. 4(B) shows the case of three microphones, Fig. 4(C) shows the case of four microphones, and Fig. 4(D) shows the case of five microphones.

[0056] The generalized radiation mode is composed of radiation modes of orders according to the number of microphones constituting the microphone array 20. For example, when the microphone array 20 includes two microphones, it has a first-order radiation mode and a second-order radiation mode, and when the microphone array includes three microphones, it has a first-order radiation mode, a second-order radiation mode, and a third-order radiation mode.

[0057] 4, the eigenvalues ​​are numbered in descending order of size as 1st order, 2nd order, 3rd order, etc. Therefore, the smaller the order, the larger the eigenvalue, and the 1st order eigenvalue is the largest eigenvalue. Also, the greater the number of microphones that make up microphone array 20, the greater the maximum eigenvalue.

[0058] Therefore, the greater the number of microphones that configure the microphone array 20, the greater the RA ratio in the microphone array 20 can be.

[0059] Figure 5(A) is an explanatory diagram showing an example of the change in eigenvectors at frequencies of 300 Hz, 1850 Hz, and 3400 Hz when there are two microphones, Figure 5(B) is when there are three microphones, Figure 5(C) is when there are four microphones, and Figure 5(D) is when there are five microphones.

[0060] 5(A) to (D) are normalized by the maximum absolute value of the vector element of each eigenvector. Each eigenvector indicates the amplitude (sensitivity) and phase (polarity) of each microphone constituting microphone array 20. Specifically, the numerical value of the vector element of the eigenvector indicates the amplitude (sensitivity) of each microphone, and the sign of the vector element of the eigenvector indicates the phase (polarity) of each microphone. For example, if the absolute value of the numerical value of the vector element of the eigenvector is large, high sensitivity is required for that microphone, and if the sign of the vector element of the eigenvector of that microphone is negative, it indicates that the polarity of the electrical terminal is opposite to that of a microphone whose sign is positive.

[0061] FIG. 5(A) shows that a microphone array 20 consisting of two microphones results in a first order gradient microphone.

[0062] FIG. 5B shows that a microphone array 20 consisting of three microphones results in a quadratic gradient microphone consisting of three microphones.

[0063] FIG. 5C shows that the microphone array 20 consisting of four microphones is different from the quadratic gradient microphone consisting of four microphones. The output p of the quadratic gradient microphone is as follows. In the following equation, x1, x2, x3, and x4 are the x coordinates of each microphone as shown in FIG. 3C. The difference in sensitivity between the microphone array 20 consisting of four microphones and the quadratic gradient microphone will be described later. p=p(x1,0,0)-p(x2,0,0)-p(x3,0,0)+p(x4,0,0)

[0064] FIG. 5D shows the sensitivity and polarity of each microphone constituting the microphone array 20 made up of five microphones.

[0065] As shown in Figures 5(A) to (D), the eigenvector with the maximum eigenvalue is almost independent of frequency. Therefore, the eigenvector at a certain frequency can be used to preferentially detect the sound of a nearby sound source at a practically sufficient level in a certain frequency range including the frequency. This makes it unnecessary to control the amplitude (sensitivity) and phase (polarity) of each microphone according to the eigenvector that differs for each frequency, and simplifies the procedure for constructing the microphone array 20.

[0066] Figures 6, 7, 8, and 9 are explanatory diagrams showing the sensitivity distribution of the microphone array 20 at frequencies of 300 Hz, 1850 Hz, and 3400 Hz on the xz plane where y = 0 in the positional relationship shown in Figure 2. In each of Figures 6, 7, 8, and 9, the dark areas indicate low sensitivity, and the bright areas indicate high sensitivity.

[0067] Each of FIG. 6, FIG. 7, FIG. 8, and FIG. 9 shows |φ T m The plot is in dB units normalized by the maximum value of |φ T m p| is set to MAX(|φ T m p|), then 20log 10 [|φ T m p| / MAX(|φ T m p|)] dB are plotted.

[0068] Fig. 6(A) is an explanatory diagram showing an example of the sensitivity distribution of a microphone array in which two microphones are arranged linearly when the frequency is 300 Hz, Fig. 6(B) is a diagram showing the case of a frequency of 1850 Hz, and Fig. 6(C) is a diagram showing the case of a frequency of 3400 Hz. As shown in Fig. 6(A) to (C), at each frequency, there is an undesirable knot-like region of low sensitivity at the center 0 m of the x-axis, but the sensitivity is highest in the central region of the x-axis excluding 0 m, and the sensitivity gradually decreases as the distance from the central region in the x-axis direction or z-axis direction increases. The knot-like region of low sensitivity can also be seen in Fig. 8(A), Fig. 8(B), and Fig. 8(C) described later.

[0069] FIG. 7(A) is an explanatory diagram showing an example of the sensitivity distribution of a microphone array in which three microphones are arranged in a line, where the frequency is 300 Hz, FIG. 7(B) is a diagram showing the sensitivity distribution when the frequency is 1850 Hz, and FIG. 7(C) is a diagram showing the sensitivity distribution when the frequency is 3400 Hz.

[0070] FIG. 8(A) is an explanatory diagram showing an example of the sensitivity distribution of a microphone array in which four microphones are arranged in a line, where the frequency is 300 Hz, FIG. 8(B) is a diagram showing the sensitivity distribution when the frequency is 1850 Hz, and FIG. 8(C) is a diagram showing the sensitivity distribution when the frequency is 3400 Hz.

[0071] FIG. 9(A) is an explanatory diagram showing an example of the sensitivity distribution of a microphone array in which five microphones are arranged in a line, where the frequency is 300 Hz, FIG. 9(B) is a diagram showing the sensitivity distribution when the frequency is 1850 Hz, and FIG. 9(C) is a diagram showing the sensitivity distribution when the frequency is 3400 Hz.

[0072] 7 to 9, there is a slight difference in sensitivity due to differences in frequency, and especially at a frequency of 300 Hz, the sensitivity is remarkably high at a short distance from the microphone at the position of 0 m where the microphone is placed. As the number of microphones constituting microphone array 20 increases to 3, 4, 5, etc., the area with high sensitivity is limited to the area closer to the microphone.

[0073] 10 is an explanatory diagram showing an example of the sensitivity distribution of one microphone placed at the origin O. When there is one microphone, no significant change due to frequency is observed, so only the case of a frequency of 300 Hz is shown as an example.

[0074] Comparing Figure 10, which shows the case of one microphone, with Figures 6(A)-(C), which shows the case of two microphones, it can be seen that while Figure 10 shows a gradual decrease in sensitivity around the origin O where the microphones are placed, Figures 6(A)-(C) show the same as a linear gradient microphone. However, in the cases of Figures 6(A)-(C), as mentioned above, an undesirable knot-shaped region of low sensitivity occurs around the origin O. Such a region of low sensitivity can also be seen in Figures (8), 8(B) and 8(C), which show the cases of four microphones.

[0075] When there are an even number of microphones, the occurrence of a region of low sensitivity centered on the origin O as shown in Figures 6(A)-(C) and Figures 8(A)-(C) is related to the fact that the eigenvectors on either side of the origin O have the same amplitude and opposite phase as shown in Figure 5(A) for the case of two microphones and Figure 5(C) for the case of four microphones. Also, as described above, as the number of microphones constituting microphone array 20 increases to three, four, five, etc., the region of high sensitivity is limited to the region closer to the microphones, so it is preferable that microphone array 20 includes an odd number of microphones, and furthermore, a greater number of microphones constituting microphone array 20, such as five rather than three, is expected to provide better characteristics as a close-talking microphone array.

[0076] Fig. 11(A) is an explanatory diagram showing an example of sensitivity distribution in a quadratic gradient microphone composed of four microphones when the frequency is 300 Hz, Fig. 11(B) is a diagram showing when the frequency is 1850 Hz, and Fig. 11(C) is a diagram showing when the frequency is 3400 Hz. The output of the quadratic gradient microphone is p(x1,0,0)-p(x2,0,0)-p(x3,0,0)+p(x4,0,0).

[0077] Comparing Figures 8(A) to (C), which show the sensitivity characteristics of the microphone array 20 composed of four microphones in this embodiment, with Figure 11, it can be seen that the microphone array 20 composed of four microphones in this embodiment has a more pronounced contrast in sensitivity between near and far near the origin O than the quadratic gradient microphone shown in Figures 11(A) to (C), and the microphone array 20 shown in Figures 8(A) to (C) has areas of high sensitivity concentrated in the vicinity of the origin O.

[0078] FIG. 12(A) is an explanatory diagram showing the sensitivity characteristic of one microphone placed at the origin O with respect to frequency, and FIG. 12(B) is an explanatory diagram showing the sensitivity characteristic of a microphone array in which five microphones are arranged in a line, with the third microphone from the left and right placed at the origin O as shown in FIG. 3(D). FIG. 12(A) and FIG. 12(B) show the change in sensitivity of the microphone array with respect to frequency when one point sound source is present at a different distance from the microphone array in the z-axis direction. The coordinates of the point sound source are (0m, 0m, 0.01m), (0m, 0m, 0.05m), (0m, 0m, 0.1m), (0m, 0m, 0.5m), and (0m, 0m, 1m). If the output of the sound source is constant, the sound pressure is proportional to the frequency as shown in the above-mentioned formula (17), so in this embodiment, the output of the sound source is operated so that it is inversely proportional to the frequency. As a result, in FIG. 12(A), the response over frequency for one microphone is flat.

[0079] As shown in FIG. 12(A), the sensitivity of a single microphone decreases with the distance from the sound source, showing the distance attenuation of a spherical wave.

[0080] In the case of Fig. 12(B), similarly to the case of Fig. 12(A), the output of the sound source is operated so that it is inversely proportional to the frequency. As shown in Fig. 12(A) and Fig. 12(B), in Fig. 12(B) showing the characteristics of the microphone array 20 according to this embodiment, the distance attenuation is more prominent than in Fig. 12(A). Also, at close ranges from the sound source, the response to frequency is almost flat, but as the sound source moves away, the sensitivity in the low frequency range decreases, showing characteristics like a high-pass filter.

[0081] To flatten the frequency response, an additional equalizer is required.

[0082] As described above, the microphone array 20 of this embodiment can set the sensitivity and polarity of each microphone according to the eigenvector with the maximum eigenvalue, thereby realizing a microphone array 20 with characteristics suitable for close-contact use, in which the sensitivity is high when the microphone array 20 is close to the sound source and the sensitivity is low when the microphone array 20 is far from the sound source.

[0083] The generalized eigenvalue problem is solved by using the acoustic reactance matrix Z i and the acoustic resistance matrix Z r The above equation can also be satisfied by swapping the left and right sides of the equation, but the eigenvalue of the generalized eigenvalue problem when such a swap is performed will be the reciprocal of the RA ratio in the original equation (5). Therefore, in such a case, the sensitivity and polarity of each microphone constituting the microphone array 20 are set based on the eigenvector when the eigenvalue is minimized. In this embodiment, even when the generalized eigenvalue problem is handled based on equation (5), the acoustic reactance matrix Z i and the acoustic resistance matrix Z r Even when the left and right sides of the above expressions are interchanged, the sensitivity and polarity of each microphone are set based on the eigenvector when the eigenvalue becomes an extreme value.

[0084] In summary, in the generalized eigenvalue problem in equation (5), the acoustic reactance matrix Z i The product of and a real matrix Φ whose columns are eigenvectors is the acoustic resistance matrix Z r When the eigenvalue is equal to the product of a real matrix Φ with eigenvectors as columns and a diagonal matrix Λ with eigenvalues ​​as elements, the sensitivity and polarity of each microphone are set based on the element of the eigenvector with the maximum eigenvalue. The eigenvalue is maximum when the active acoustic power W r Reactive sound power W i is the maximum.

[0085] In addition, in equation (5), the acoustic reactance matrix Zi and the acoustic resistance matrix Z r In the generalized eigenvalue problem, the acoustic resistance matrix Z r The product of and the real matrix Φ is the acoustic reactance matrix Z i The sensitivity and polarity of each microphone are set based on the element of the eigenvector that has the smallest eigenvalue when the product of the real matrix Φ and the diagonal matrix Λ is equal to the reactive sound power W i Active acoustic power W r is the smallest case.

[0086] Microphone arrays composed of multiple microphones include first-order gradient microphones and second-order gradient microphones, but first-order gradient microphones and second-order gradient microphones do not take into account the optimal settings of each microphone from the eigenvalues ​​and eigenvectors, as does the microphone array 20 of this embodiment, and do not have characteristics suitable for close-talking comparable to the microphone array 20 of this embodiment, which has four or five microphones.

[0087] In the microphone array 20 of this embodiment, the number of microphones is also important. If an even number of microphones are arranged symmetrically with respect to the origin O, as shown in Figures 5(A) and 5(C), the eigenvectors on either side of the origin O will have the same amplitude and opposite phase, which may result in a phenomenon in which the sensitivity of the microphone array 20 is significantly reduced at the origin O, as shown in Figures 6(A) to (C) and Figures 8(B) and (C).

[0088] To avoid this phenomenon, microphone array 20 is configured with an odd number of microphones, with one microphone placed at origin O and the remaining microphones placed symmetrically with respect to origin O. With this arrangement, it is possible to avoid the phenomenon in which the sensitivity of microphone array 20 at origin O drops significantly.

[0089] Furthermore, even if the number of microphones constituting microphone array 20 is the same odd number, five microphones will have characteristics more suitable for close talking than three.

[0090] The microphone array 20 according to this embodiment has characteristics suitable for close-talking applications in telephones, headsets, etc., but is not limited thereto. In addition to close-talking applications, the microphone array also has characteristics as a near-sound source microphone array for picking up sounds such as the murmuring of a river or the flapping of insects without detecting surrounding noise.

[0091] The above-described embodiments are intended to facilitate understanding of the present invention, and are not intended to limit the present invention. The present invention may be modified or improved without departing from the spirit of the present invention, and equivalents are also included in the present invention. In other words, even if a person skilled in the art appropriately modifies the design of the embodiments, the embodiment is included in the scope of the present invention as long as it has the characteristics of the present invention. In addition, the elements of the embodiments can be combined to the extent technically possible, and the combination of these elements is included in the scope of the present invention as long as it has the characteristics of the present invention. [Explanation of symbols]

[0092] 10. Audio input device 20 Microphone Array 20-1, 20-2, 20-3, 20-4, 20-5 Microphones 30 Amplifier Array 30-1, 20-2, 20-3, 20-4, 20-5 Amplifiers

Claims

1. A close-talking microphone array in which the sensitivity and terminal polarity of each of a plurality of microphones are set based on eigenvalues ​​and eigenvectors calculated from a generalized eigenvalue problem relating to a matrix whose elements are the real part of the specific acoustic impedance on the surface of the microphone array, which is a collection of a plurality of microphones, and a matrix whose elements are the imaginary part of the specific acoustic impedance.

2. 2. The close-talking microphone array according to claim 1, wherein the sensitivity and terminal polarity of each of the plurality of microphones are set based on an element of an eigenvector that provides a maximum eigenvalue when a product of a matrix having the imaginary part of the specific acoustic impedance as its elements and a real matrix having the eigenvectors as its columns is equal to a product of a matrix having the real part of the specific acoustic impedance as its elements, a real matrix having the eigenvectors as its columns, and a diagonal matrix having the eigenvalues ​​as its elements.

3. 2. The close-talking microphone array according to claim 1, wherein the sensitivity and terminal polarity of each of the plurality of microphones are set based on an element of an eigenvector that provides a minimum eigenvalue when a product of a matrix having real parts of the specific acoustic impedance as elements and a real matrix having eigenvectors as columns is equal to a product of a matrix having imaginary parts of the specific acoustic impedance as elements, a real matrix having eigenvectors as columns and a diagonal matrix having eigenvalues ​​as elements.

4. 4. The close-talking microphone array according to claim 1, wherein the plurality of microphones are arranged in a line.

5. 4. The close-talking microphone array according to claim 3, wherein the plurality of microphones are arranged at intervals equal to or less than a half wavelength of an upper limit frequency to be measured by each of the plurality of microphones.

6. A step of calculating eigenvalues ​​and eigenvectors from a generalized eigenvalue problem for a matrix whose elements are the real part of the specific acoustic impedance on a surface of a microphone array, which is a set of multiple microphones, and a matrix whose elements are the imaginary part of the specific acoustic impedance; setting a sensitivity and a terminal polarity of each of the plurality of microphones based on the calculated eigenvalues ​​and eigenvectors; A method for configuring a close-talking microphone array, comprising:

Images