JP6982966B2 - Sound source exploration device - Google Patents

Description

The present invention also relates to the sound source exploration equipment.

Conventionally, it is generally performed to estimate the arrival direction of a sound source by using the MUSIC (Multiple Signal Classification) method. The MUSIC method is an algorithm used for frequency estimation and wave source estimation, and is generally widely known.

For example, a robot has been developed for the purpose of communicating with humans by conversation using the technology of the MUSIC method (see Patent Document 1). The invention described in Patent Document 1 estimates the direction of a sound source by a MUSIC algorithm based on the positional relationship between the microphones constituting the microphone array and the sound source signals of a plurality of channels output from the microphone array.

Japanese Unexamined Patent Publication No. 2011-22701

However, in the invention described in Patent Document 1, no consideration has been given to the arrangement of microphones. Therefore, depending on the positional relationship between each microphone, it may be difficult to search for a sound source, or even if a sound source coming from any direction in the 4π space can be searched, more microphones than necessary are required. There was a problem.

From this point of view, the present invention provides a sound source localization equipment capable of the direction of arrival of the sound source to efficiently exploration.

In order to solve the above problems, the sound source search device according to the present invention is a sound source search device that searches the position of a sound source by using the MUSIC method. The sound source search device includes a microphone array composed of a plurality of microphones, and a signal processing unit that receives sound source signals from the plurality of microphones and estimates the arrival direction of the sound source by a MUSIC algorithm. By arranging the microphones in a plurality of layers in the similar model of the C80 fullerene molecular structure, the microphones are arranged on a plurality of concentric circles having different radii and shifting the center point in the direction of the central axis. When the similarity model is numbered in ascending order from the upper layer with respect to the central axis of the concentric circles, and the nodes in the same layer are numbered in ascending order clockwise, the first layer is assigned. The microphone is located at the positions of the first, third, and fourth sections, the first, third, and fourth sections of the second layer, and the first, fifth, and seventh sections of the fifth layer. The microphone is arranged, and the microphone is not arranged at any other position.

In order to solve the above problems, the sound source search device according to the present invention is a sound source search device that searches the position of a sound source by using the MUSIC method. The sound source search device includes a microphone array composed of a plurality of microphones, and a signal processing unit that receives sound source signals from the plurality of microphones and estimates the arrival direction of the sound source by a MUSIC algorithm. By arranging the microphones in a plurality of layers in a similar model of the C80 fullerene molecular structure, the microphones are arranged on a plurality of concentric circles having different radii and shifting the center point in the direction of the central axis. When the similarity model is numbered in ascending order from the upper layer with respect to the central axis of the concentric circles, and the nodes in the same layer are numbered in ascending order clockwise, the first layer is assigned. The first, third, and fourth sections, the first, third, and fourth sections of the second layer, the first, fifth, and seventh sections of the third layer, and the second of the fourth layer. The microphone is arranged at the positions of the sixth, sixth, and eighth nodes, and the microphone is not arranged at other positions.

In the present invention, not all microphones are located on the same plane, and the microphones are arranged three-dimensionally. Therefore, the problem that the arrival direction of the sound source cannot be estimated is unlikely to occur, and it is possible to search for a sound source existing in all directions in the 4π space.

According to the present invention, the arrival direction of the sound source can be efficiently searched.

It is a schematic block diagram of the sound source exploration apparatus which concerns on 1st Embodiment of this invention. It is a figure for demonstrating a three-dimensional coordinate system. It is a figure for demonstrating the structure of the microphone array which concerns on 1st Embodiment of this invention. It is a figure for demonstrating the reference circle which becomes the reference of the arrangement of a microphone, (a) shows the case where a reference circle is based on a cone, and (b) shows the case where a reference circle is based on a hemisphere. It is a figure for demonstrating the 1st arrangement pattern of a microphone, (a) shows the perspective view, (b) shows the front view, (c) shows the plan view. It is a figure for demonstrating the 2nd arrangement pattern of a microphone, (a) shows the perspective view, (b) shows the front view, (c) shows the plan view. It is a figure for demonstrating the 3rd arrangement pattern of a microphone, (a) shows the perspective view, (b) shows the front view, (c) shows the plan view. It is a figure for demonstrating the 4th arrangement pattern of a microphone, (a) shows the perspective view, (b) shows the front view, (c) shows the plan view. It is a figure for demonstrating the effect of the sound source exploration apparatus which concerns on 1st Embodiment of this invention, (a) shows the arrangement of the microphone which concerns on this Embodiment, (b) is the conventional microphone as a comparative example. Shows the placement. It is a figure for demonstrating the effect of the sound source exploration apparatus which concerns on 1st Embodiment of this invention, (a) shows the CRLB value of the microphone which concerns on this embodiment, (b) is the conventional microphone as a comparative example. The CRLB value of is shown. It is a schematic perspective view of the C80 fullerene molecular structure which is the basis of the microphone array which concerns on 2nd Embodiment of this invention. It is a similar model of the C80 fullerene molecular structure, (a) shows a plan view, (b) shows a front view, and (c) shows a side view. It is a figure for demonstrating the 5th arrangement pattern of a microphone, (a) shows the perspective view, (b) shows the front view, (c) shows the plan view. It is a figure for demonstrating the sixth arrangement pattern of a microphone, (a) shows the perspective view, (b) shows the front view, (c) shows the plan view. It is a figure for demonstrating the 7th arrangement pattern of a microphone, (a) shows the perspective view, (b) shows the front view, (c) shows the plan view. It is a figure for demonstrating the 8th arrangement pattern of a microphone, (a) shows the perspective view, (b) shows the front view, (c) shows the plan view. It is a figure for demonstrating the 9th arrangement pattern of a microphone, (a) shows the perspective view, (b) shows the front view, (c) shows the plan view. It is a figure for demonstrating the frequency characteristic of the signal radiated from a speaker in an estimation experiment. It is a figure for demonstrating the time waveform of the signal emitted from a speaker in an estimation experiment. It is a figure for demonstrating the estimation result of the 5th arrangement pattern of a microphone, (a) shows the estimation result of the frequency 90Hz, (b) shows the estimation result of the frequency 270Hz. It is a figure for demonstrating the estimation result of the 6th arrangement pattern of a microphone, (a) shows the estimation result of the frequency 90Hz, (b) shows the estimation result of the frequency 270Hz. It is a figure for demonstrating the estimation result of the 7th arrangement pattern of a microphone, (a) shows the estimation result of the frequency 90Hz, (b) shows the estimation result of the frequency 270Hz. It is a figure for demonstrating the estimation result of the 8th arrangement pattern of a microphone, (a) shows the estimation result of the frequency 90Hz, (b) shows the estimation result of the frequency 270Hz. It is a figure for demonstrating the estimation result of the 9th arrangement pattern of a microphone, (a) shows the estimation result of the frequency 90Hz, (b) shows the estimation result of the frequency 270Hz. It is a figure for demonstrating the tenth arrangement pattern of a microphone, (a) shows the perspective view, (b) shows the front view, (c) shows the plan view. It is a figure for demonstrating the eleventh arrangement pattern of a microphone, (a) shows the perspective view, (b) shows the front view, (c) shows the plan view. It is a figure for demonstrating the twelfth arrangement pattern of a microphone, (a) shows the perspective view, (b) shows the front view, (c) shows the plan view. It is a figure for demonstrating the thirteenth arrangement pattern of a microphone, (a) shows the perspective view, (b) shows the front view, (c) shows the plan view. It is a figure for demonstrating the estimation result of the tenth arrangement pattern of a microphone, (a) shows the estimation result of the frequency 90Hz, (b) shows the estimation result of the frequency 270Hz. It is a figure for demonstrating the estimation result of the eleventh arrangement pattern of a microphone, (a) shows the estimation result of the frequency 90Hz, (b) shows the estimation result of the frequency 270Hz. It is a figure for demonstrating the estimation result of the twelfth arrangement pattern of a microphone, (a) shows the estimation result of the frequency 90Hz, (b) shows the estimation result of the frequency 270Hz. It is a figure for demonstrating the estimation result of the 13th arrangement pattern of a microphone, (a) shows the estimation result of the frequency 90Hz, (b) shows the estimation result of the frequency 270Hz.

Hereinafter, embodiments for carrying out the present invention will be described in detail with reference to the drawings as appropriate.
Each figure is merely schematically shown to the extent that the present invention can be fully understood. Therefore, the present invention is not limited to the illustrated examples. In each figure, common components and similar components are designated by the same reference numerals, and duplicate description thereof will be omitted.

[First Embodiment]
(Configuration of sound source exploration device according to the first embodiment)
The configuration of the sound source exploration device 1 according to the first embodiment will be described with reference to FIG. The sound source exploration device 1 is a device for exploring the arrival direction of a sound source, and estimates, for example, the generation direction of an abnormal sound when an abnormal sound is generated in a room (4π space).
The sound source search device 1 mainly includes a microphone array 3, an A / D converter 4, and a signal processing unit 10. The signal processing unit 10 is realized by a program execution process by a CPU (Central Processing Unit), a dedicated circuit, or the like.

The microphone array 3 is a three-dimensional arrangement of a plurality of omnidirectional microphones M. Details of the number of microphones M and the arrangement method will be described later. Each microphone M outputs an observation signal (including a sound source signal) to the A / D converter 4. That is, the microphone array 3 outputs analog observation signals for the number of microphones M to the A / D converter 4.
The A / D converter 4 receives analog observation signals for the number of microphones M from the microphone array 3 and performs analog / digital conversion on the received analog observation signals. Then, the A / D converter 4 outputs the digital observation signal to the signal processing unit 10.

The signal processing unit 10 receives digital observation signals for the number of microphones from the A / D converter 4 and estimates the arrival direction of the sound source by the MUSIC algorithm. Then, the signal processing unit 10 outputs an estimated value (horizontal angle φ, elevation angle θ) in the arrival direction of the sound source to the outside.
The signal processing unit 10 may have a general configuration capable of estimating the arrival direction of the sound source by the MUSIC algorithm. The signal processing unit 10 here includes a frame processing unit 20, an FFT (Fast Fourier Transformation) 21, an observation space correlation matrix calculation unit 22, an eigenvalue decomposition unit 23, a MUSIC spectrum calculation unit 24, and an arrival direction estimation unit. It includes 25, a model space correlation matrix storage unit 30, and an arrival direction storage unit 31. The configuration of the signal processing unit 10 here is merely an example.

The frame processing unit 20 frames the digital observation signal output from the A / D converter 4 by the frame length (for example, one frame is 4 milliseconds).
The FFT 21 performs FFT processing and calculates an observation signal vector from each framed digital observation signal.
The observation space correlation matrix calculation unit 22 calculates the current observation space correlation matrix based on the observation signal vector obtained from the FFT 21 and the past correlation matrix.
The eigenvalue decomposition unit 23 decomposes the current observation space correlation matrix into eigenvalues and obtains a noise level eigenvector corresponding to an eigenvalue other than a large eigenvalue.

The MUSIC spectrum calculation unit 24 calculates the MUSIC spectrum based on the model space correlation matrix previously stored in the model space correlation matrix storage unit 30 and the noise level eigenvector calculated by the eigenvalue decomposition unit 23.
The arrival direction estimation unit 25 averages the estimation directions obtained in each frequency domain (bin), and sets the peak value as the final arrival direction estimation value.
The model space correlation matrix is stored in the model space correlation matrix storage unit 30. The model space correlation matrix is a matrix obtained by the phase difference of the output of each microphone M with respect to one microphone M.
The arrival direction storage unit 31 stores the estimated arrival direction estimated value.

Next, the configuration of the microphone array 3 will be described. The microphone array 3 of the present embodiment is a three-dimensional arrangement of a plurality of microphones M. Three-dimensional here means that not all microphones M are located on the same plane.
Here, the arrangement of the microphone M will be described using the three-dimensional coordinate system shown in FIG. In the three-dimensional coordinate system shown in FIG. 2, an arbitrary point in 4π space (for example, the center of the microphone array 3) is set as the origin O, the vertical direction is set as the z-axis, and an arbitrary direction in the horizontal direction (for example, the latitude direction) is set. The x-axis is used, and the direction orthogonal to the z-axis and the x-axis (for example, the longitude direction) is the y-axis. In the figure, the symbol “φ” is a horizontal angle, and the symbol “θ” is an elevation angle. The three-dimensional coordinate system is defined for convenience of explanation, and does not limit the present invention.

As shown in FIG. 3, a plurality of reference circles are set in the three-dimensional coordinate system. The reference circle is a virtual figure that serves as a reference for the arrangement of the microphone M, and is composed of a circle whose center is located on the z-axis. The plurality of reference circles are concentric circles having different radii offset in the z-axis direction. The number of reference circles may be plural, and is not particularly limited. Here, three reference circles, a first reference circle, a second reference circle, and a third reference circle are set. The first reference circle is a circle having a radius R _{1 centered on the point Q 1} on the z-axis. The second reference circle is a circle having a radius R _{2 centered on the point Q 2} on the z-axis. The third reference circle is a circle with a radius R _{3 centered on the point Q 3} on the z-axis. Here, the coordinates of the point Q ₃ are the same as the origin O coordinates.

First center point of the reference circle Q ₁ and the distance dz ₁₂ between the center Q ₂ of the second reference circle, the distance dz ₂₃ of the center point of the second reference circle Q ₂ and the center point Q ₃ of the third reference circle May be the same or different. That is, the deviation amount dz of each reference circle in the z-axis direction (central axis direction) may be evenly spaced or different. If the deviation amount dz of the reference circle is made too wide, it becomes difficult to estimate the position of the sound source that generates high frequency sound. Therefore, the maximum width of the deviation amount dz of the reference circle is determined by the wavelength of the sound source. Is good.

The difference between the radius R ₁ of the first reference circle and the difference between the radius R ₂ of the second reference circle (R ₁ -R _2), the radius R ₃ of the radius R ₂ of the second reference circle third reference circle ( It may be the same as or different from _{R 2-} R _3). The radius R of each reference circle can be set arbitrarily, and preferably not the same value. The reference circle is, for example, the outer circumference of a cut surface obtained by cutting a cone (see FIG. 4 (a)) or a hemisphere (see FIG. 4 (b)) set on the xy plane on a plane parallel to the bottom surface (xy plane). It may be there. In that case, the radius R is determined by the position of the cut surface.

As shown in FIG. 3, the microphone M is arranged on the circumference of each reference circle. In the present embodiment, the microphone M arranged in the reference circle is represented _{by "microphone M ij" from the relationship between the reference circle and the microphone M.} The symbol "i" is a positive integer that identifies the reference circle, and here, the origin O is numbered in ascending order from the reference circle far from the reference circle (i = 1, 2, ...). Further, the symbol "j" is a positive integer that identifies the microphone M on the same reference circle, and here, a number (j = 1, 2, ...) In ascending order clockwise from any microphone M is used. It is attached. In addition, ascending numbers (i = 1, 2, ...) may be assigned from the reference circle near the origin O, or ascending numbers (j = 1, ...) counterclockwise from any microphone M. 2, ...) may be added.

The total number of microphones M and the position where the microphones M are arranged may be such that the microphones M can be arranged three-dimensionally, and a total of four microphones M may be arranged on at least two reference circles.
Further, since the position where the microphones M are arranged is related to the accuracy of estimating the arrival direction of the sound source, it is desirable that the microphones M are arranged evenly. That is, it is preferable that the number of microphones M arranged on each reference circle is the same. Further, the microphones M on the same reference circle are preferably arranged at equal intervals.

The arrangement pattern of the microphone M will be described with reference to FIGS. 5 to 8. 5 to 7 show the case where the microphone M is arranged on the concentric circles created from the cone (on the surface of the cone), and FIG. 8 shows the microphone M on the concentric circles created from the hemisphere (on the spherical surface of the hemisphere). Shows the case where.
In the first arrangement pattern shown in FIG. 5, two reference circles, a first reference circle and a second reference circle, are set in the three-dimensional coordinate system, and three microphones M (six in total) are arranged in each reference circle. Has been done. As shown in FIG. 5 (c), the microphones M ₁₁ , M ₁₂ , and M ₁₃ are arranged at equal intervals on the first reference circle, and the microphones M ₂₁ , M ₂₂ , and M ₂₃ are the second reference circles. They are evenly spaced on top. In addition, the microphone M ₁₁ and the microphone M ₂₁ are arranged at corresponding positions, and the angle (horizontal angle φ) from the x-axis to the microphone M ₁₁ and the angle from the x-axis to the microphone M ₂₁ with the z-axis as the center. It is the same as the angle (horizontal angle φ). The same applies to the relationship between the microphone M ₁₂ and the microphone M _22, and the relationship between the microphone M ₁₃ and the microphone M _23.

In the second arrangement pattern shown in FIG. 6, three reference circles, a first reference circle, a second reference circle, and a third reference circle are set in the three-dimensional coordinate system. _{The distance dz 12} between the first reference circle and the second reference circle and _{the distance dz 23} between the second reference circle and the third reference circle are the same (distance dz ₁₂ = distance dz ₂₃ ). Two microphones M (six in total) are arranged in each reference circle. As shown in FIG. 6 (c), the microphones M ₁₁ and M ₁₂ are arranged at equal intervals on the first reference circle, and the microphones M ₂₁ and M ₂₂ are arranged at equal intervals on the second reference circle. The microphones M ₃₁ and M ₃₂ are arranged at equal intervals on the third reference circle. Microphone M ₂₁ is a microphone M ₁₁ around the z axis is arranged at a position rotated clockwise a predetermined angle, the microphone M ₃₁ is further a microphone M ₂₁ only clockwise a predetermined angle about the z axis It is placed in the rotated position. That is, the microphone M ₂₁ has a different horizontal angle φ with respect to the reference plane including the z-axis (central axis) and the microphone M _{11 , and the microphone M 31} has a reference plane including the z-axis and the microphone M _21. The horizontal angle φ is different. The same applies to the relationship between the microphone M ₁₂ and the microphone M ₂₂ and the microphone M _32. The rotation angle (the amount of deviation of the horizontal angle φ) between the microphones M may be the same or different. As a result, as shown in FIG. 6 (c), the corresponding group of microphones M ₁₁ , M ₂₁ , M ₃₁ and the microphones M ₁₂ , M ₂₂ , M ₃₂ are arranged in a spiral.

In the third arrangement pattern shown in FIG. 7, four reference circles, a first reference circle, a second reference circle, a third reference circle, and a fourth reference circle are set in the three-dimensional coordinate system. _{The distance dz 12} between the first reference circle and the second reference circle, the distance _{dz 23} between the second reference circle and the third reference circle, _{and the distance dz 34} between the third reference circle and the fourth reference circle are the same. (Distance dz ₁₂ = Distance dz ₂₃ = Distance dz ₃₄ ). One microphone M (four in total) is arranged in each reference circle. The microphone M ₁₁ is located on the first reference circle, the microphone M ₂₁ is located on the second reference circle, the microphone M ₃₁ is located on the third reference circle, and the microphone M ₄₁ is located on the third reference circle. It is arranged on the four reference circles. The microphone M ₂₁ is arranged at a position where the microphone M ₁₁ is rotated clockwise by a predetermined angle about the z-axis. The microphone M ₃₁ is arranged at a position where the microphone M ₂₁ is further rotated clockwise by a predetermined angle about the z-axis. The microphone M ₄₁ is arranged at a position where the microphone M ₃₁ is further rotated clockwise by a predetermined angle about the z-axis. The rotation angle (the amount of deviation of the horizontal angle φ) between the microphones M may be the same or different. As a result, as shown in FIG. 7 (c), the corresponding group of microphones M ₁₁ , M ₂₁ , M ₃₁ and M ₄₁ are arranged in a spiral shape.

In the fourth arrangement pattern shown in FIG. 8, four reference circles, a first reference circle, a second reference circle, a third reference circle, and a fourth reference circle are set in the three-dimensional coordinate system. The reference circle of the third arrangement pattern shown in FIG. 7 is created from the cut surface of the cone, whereas the reference circle of the fourth arrangement pattern shown in FIG. 8 is created from the cut surface of the hemisphere. different. _{The distance dz 12} between the first reference circle and the second reference circle, the distance _{dz 23} between the second reference circle and the third reference circle, _{and the distance dz 34} between the third reference circle and the fourth reference circle are the same. (Distance dz ₁₂ = Distance dz ₂₃ = Distance dz ₃₄ ). One microphone M (four in total) is arranged in each reference circle. The microphone M ₁₁ is located on the first reference circle, the microphone M ₂₁ is located on the second reference circle, the microphone M ₃₁ is located on the third reference circle, and the microphone M ₄₁ is located on the third reference circle. It is arranged on the four reference circles. The microphone M ₂₁ is arranged at a position where the microphone M ₁₁ is rotated clockwise by a predetermined angle about the z-axis. The microphone M ₃₁ is arranged at a position where the microphone M ₂₁ is further rotated clockwise by a predetermined angle about the z-axis. The microphone M ₄₁ is arranged at a position where the microphone M ₃₁ is further rotated clockwise by a predetermined angle about the z-axis. The rotation angle between the microphones M may be the same or different. As a result, as shown in FIG. 8 (c), the corresponding group of microphones M ₁₁ , M ₂₁ , M ₃₁ and M ₄₁ are arranged in a spiral shape.

(Effect of sound source exploration device according to the first embodiment)
The effect of the sound source exploration device 1 according to the first embodiment will be described. Here, the estimation accuracy of the arrival direction of the sound source is defined as the Cramér-Rao lower limit (CRLB) and considered. The Cramér-Rao lower limit is a statistical estimation limit in an estimation problem, and is explained, for example, in the following literature.
・ Moriya, Ichige et al., "Improving Elevation Optimization Accuracy in DOA Optimization: How Planner Arrays Can Be Modified into 3-D Configuration", IEICE TRANS. FUNDAMENTALS, 2012, Vol.E95.A, No.10, p.1667 --p.1675
By applying the Kramer-Lao lower limit to the sound source arrival direction estimation, the calculation result by the Kramer-Lao lower limit is considered to be a numerical value equivalent to the root mean square error of the estimated value of the sound source direction coming to any microphone array. That is, when a sound wave of an arbitrary frequency arrives from a direction in an arbitrary microphone array arrangement, the root mean square error between the estimated value by the MUSIC method and the true arrival direction is given by the Cramér-Rao lower limit. Therefore, it is possible to consider the microphone arrangement, the incoming sound source frequency, the incoming sound source direction, and the like.

(Estimation used to evaluate estimation accuracy)
First, the signal model input to the microphone array will be described. In the coordinate system shown in FIG. 2, it is assumed that K uncorrelated waves arrive at the microphone array of P elements from infinity under a white noise environment. The position vector r _p (p = 1, ..., P) of the p- _{th element and the array steering vector a p} (p = 1, ..., P) are expressed by the following equation (1). The vector.

Further, the time waveform and the arrival direction of the kth wave are given by _sk (t) and (θ _k , φ _k). Array steering vector a (θ _k , φ _k ) = [a ₁ (θ _k , φ _k ), ..., a _P (θ _k , φ _k )] ^T , and the array input vector x (t) is as follows. It is expressed as the equation (2) of. The vector n (t) in the equation (2) is a thermal noise vector.

Subsequently, a method for deriving CRLB will be described. It is assumed that the _kth arrival signal having a complex amplitude _sk (t) comes from (θ k, φ _k). The estimated CRLB of θ and φ is represented by the coupling shown in the following equation (3).

In the case of one sound source, the CRLB of the elevation angle θ is expressed by the following equation (4). _{The CRLB S} of the plurality of sound sources is calculated from the equation (4). CRLB with a horizontal angle φ is also derived.

(Calculation conditions for estimation accuracy)
Here, as the microphone array 3 according to the present embodiment, the arrangement (three-dimensional spiral shape) of the microphones M shown in FIG. 9A is assumed. Further, as a comparative example, the arrangement (planar spiral shape) of the microphone M shown in FIG. 9B was assumed.

The microphone array 3 shown in FIG. 9A is composed of nine microphones M, and the arrangement of the microphones M is determined based on the first reference circle to the third reference circle. _{The radius R 1} of the first reference circle is "67.5 (mm)", and the distance dz _{1 in the} z-axis direction is "135.0 (mm)". _{The radius R 2} of the second reference circle is "130.5 (mm)", and the distance dz _{2 in the} z-axis direction is "67.5 (mm)". _{The radius R 3} of the third reference circle is "168.0 (mm)", and the distance dz _{3 in the} z-axis direction is "0.0 (mm)".
Three microphones M are arranged at equal intervals in each of the first reference circle to the third reference circle. Further, the microphones M on the adjacent reference circles are arranged so as to be displaced in the horizontal angle direction. That is, the microphone M on the second reference circle is arranged at a position rotated by 40 ° clockwise with respect to the microphone M on the first reference circle. Further, the microphone M on the third reference circle is arranged at a position rotated by 40 ° clockwise with respect to the microphone M on the second reference circle.

The microphone array 103 shown in FIG. 9B is composed of nine microphones M, and the arrangement of the microphones M is determined based on the first reference circle to the third reference circle. _{The radius R 1} of the first reference circle is "67.5 (mm)", the radius R ₂ of the second reference circle is "130.5 (mm)", and the radius R ₃ of the third reference circle is "168.0 (mm)". _{, And the distances dz 1} , dz ₂ , and dz _{3 in} the z-axis direction of all the reference circles are "0.0 (mm)".
Three microphones M are arranged at equal intervals in each of the first reference circle to the third reference circle. Further, the microphones M on the adjacent reference circles are arranged so as to be displaced in the horizontal angle direction. That is, the microphone on the second reference circle is arranged at a position rotated by 40 ° clockwise with respect to the microphone M on the first reference circle. Further, the microphone on the third reference circle is arranged at a position rotated by 40 ° clockwise with respect to the microphone M on the second reference circle.

Then, the horizontal angle φ in the arrival direction was fixed to an arbitrary value, and the elevation angle θ was changed to “0 °, 20 °, 45 °, 75 °, 90 °” to calculate the CRLB. Other conditions are as shown in [Table 1] below.

The calculation result of the estimation accuracy is shown in FIG. The horizontal axis shown in FIG. 10 represents the elevation angle θ, and the vertical axis represents the value of CRLB. The graph shows the relationship between the elevation angle θ and CRLB for each frequency “63Hz, 125Hz, 250Hz, 500Hz, 1000Hz, 2000Hz”. 10 (a) is a calculation result of the estimation accuracy in the arrangement of the microphone M shown in FIG. 9 (a), and FIG. 10 (b) is a calculation of the estimation accuracy in the arrangement of the microphone M shown in FIG. 9 (b). The result.
As can be seen from this graph, in the arrangement of the comparative example (planar spiral shape) shown in FIG. 10B, the estimation error increases as the angle approaches the horizontal plane (elevation angle θ = 90 °). On the other hand, in the arrangement (three-dimensional spiral shape) of the present embodiment shown in FIG. 10A, an increase in estimation error near the horizontal plane (elevation angle θ = 90 °) is suppressed. As for the arrival from directly above (the elevation angle θ is near 0 °), the same high estimation accuracy as the arrangement of the comparative example is obtained.

As described above, in the sound source exploration device 1 according to the first embodiment, it is easy to arrange the microphones M three-dimensionally, and it is possible to reduce the possibility that all the microphones M are located on the same plane. can. That is, by increasing the number of microphones M, it is possible to eliminate a region where it is difficult to estimate the arrival direction of the sound source. Therefore, the problem that the arrival direction of the sound source cannot be estimated is unlikely to occur, and the sound source existing in all directions of the 4π space can be efficiently searched using the minimum necessary microphone.
Further, in the first to fourth arrangement patterns shown in FIGS. 5 to 8, not all the microphones M are located on the same plane, and the microphones M are arranged three-dimensionally. Therefore, it is extremely unlikely that the problem of not being able to estimate the direction of arrival of the sound source will occur.

[Second Embodiment]
The microphone array 3A (not shown) of the sound source exploration device 1A (not shown) according to the second embodiment has a microphone M arranged based on the C80 fullerene molecular structure. The C80 fullerene molecular structure can generally be defined as a cluster composed of 80 carbon atoms. Here, paying attention to the structure, the microphone M is arranged at a position corresponding to the position of the carbon atom by using a similar model simulating the C80 fullerene molecular structure. The molecular structure of C80 fullerene is shown in FIG. The C80 fullerene molecular structure is composed of a combination of a hexagon and a pentagon having a carbon atom as an apex, and has a substantially spherical shape. As shown in FIG. 11, the C80 fullerene molecular structure has a rotational symmetry in which the positions of carbon atoms are the same even when rotated by 72 ° with the z-axis as the rotation axis in a state where a pair of opposing pentagons are arranged vertically. It has become. In the similarity model, the carbon atom of the C80 fullerene molecular structure is represented by a representative point, and this representative point is referred to as a “node”. The representative point may be, for example, the center of a carbon atom. A similar model of the C80 fullerene molecular structure is shown in FIG.

As shown in FIG. 12, in a state where a pair of opposing pentagons are arranged vertically, the node F in the similarity model of the C80 fullerene molecular structure is classified into a plurality of layers based on the vertical direction (z-axis direction). That is, the hierarchy here is a name representing a set of clauses F having the same z value. In the present embodiment, the section F is _{represented by "section F vw} " in relation to the hierarchy. The symbol "v" is a positive integer that identifies the hierarchy, and here, numbers in ascending order (v = 1, 2, ...) Are assigned from the upper hierarchy to the lower hierarchy. Further, the symbol "w" is a positive integer that identifies the clause F in the same layer, and here, a number in ascending order (w = 1, 2, ...) Is added clockwise from any clause F. ing. In addition, ascending numbers (v = 1, 2, ...) may be assigned from the lower layer to the upper layer, or ascending numbers (counterclockwise from any section F). w = 1, 2, ...) may be added.

As shown in FIG. 12 (a), there are five sections F _{11 to} F ₁₅ in the first layer. Similarly, as shown in FIG. 12 (b), the second layer has five sections F _{21 to} F ₂₅ , and the third layer has ten sections F _{31 to} F _{310. There are 10 sections F 41 to} F ₄₁₀ in the 4th layer, and 10 _{sections F 51 to} F ₅₁₀ in the 5th layer. The sixth layer has ten sections F _{61 to} F ₆₁₀ , the seventh layer has ten sections F _{71 to} F ₇₁₀ , and the eighth layer has ten sections F _81. There are ~ F _{810, there are five sections F 91} ~ F ₉₅ in the ninth layer, and there are five _{sections F 101} ~ F ₁₀₅ in the tenth layer.

Here, as shown in FIG. 12 (a), the nodes F _{11 to} F ₁₅ of the first layer are located at the same distance from the z axis in the xy plane, and thus are located on a circle centered on the z axis. It can be said that. Further, since the sections F _{21 to} F ₂₅ of the second layer are at the same distance from the z-axis in the xy plane, it can be said that they are located on a circle centered on the z-axis. In this way, since the nodes F in the same layer are at the same distance from the z-axis in the xy plane, all the nodes F have the z-axis as the central axis and ten different radii shifted in the z-axis direction. It exists on any of the concentric circles (thick dashed line in FIG. 12).

The arrangement pattern of the microphone M based on the C80 fullerene molecular structure will be described with reference to FIGS. 13 to 17. As above described, sound source localization apparatus 1A according to the second embodiment, since those of arranging the microphone M to a node F _vw in similar models of C80 fullerene molecular structure, the microphone M from the relationship between the nodes F _vw " It will be represented by "microphone M _vw".

In the fifth arrangement pattern shown in FIG. 13, the microphone M is arranged in _{all the nodes F vw corresponding to the C80 fullerene molecular structure.} That is, _{there are a total of 80 microphones M vw} (80 channels (ch)) from the first layer to the tenth layer. The size of the microphone array 3A is "about 0.2 m" in the vertical and horizontal directions. The size is the same for the other arrangement patterns of the second embodiment.

In the sixth arrangement pattern shown in FIG. 14, the microphone M _vw is arranged in a _{total of 50 nodes F vw from} the first layer to the sixth layer corresponding to the C80 fullerene molecular structure. Specifically, the microphones M _{11 to} M ₁₅ of the first layer, the microphones M _{21 to} M ₂₅ of the second layer, the microphones M _{31 to} M ₃₁₀ of the third layer, and the microphones M _{41 to} M of the fourth layer. _It consists of 410, fifth-tier microphones M _{51 to} M _510, and sixth-tier microphones M _{61 to} M ₆₁₀ .

In the seventh arrangement pattern shown in FIG. 15, the microphone M _vw is arranged in a _{total of 40 nodes F vw from} the first layer to the fifth layer (that is, the hemisphere) corresponding to the C80 fullerene molecular structure. Specifically, the first-tier microphones M _{11 to} M ₁₅ , the second-tier microphones M _{21 to} M ₂₅ , the third-tier microphones M _{31 to} M ₃₁₀ , and the fourth-tier microphones M _{41 to M. It} consists of 410 and fifth-level microphones M _{51 to} M ₅₁₀ .

In the eighth arrangement pattern shown in FIG. 16, the microphone M _vw is arranged in a _{part of the node F vw from} the first layer to the fifth layer (that is, the hemisphere) corresponding to the C80 fullerene molecular structure. From Specifically, the microphone M _11, M _13, M ₁₄ of the first layer, a microphone M _21, M _23, M ₂₄ of the second layer, the microphone M ₅₁ of the fifth hierarchy, M _55, M ₅₇ Prefecture Become. In the eighth arrangement pattern, as shown in FIG. 16A, the microphone M _vw is arranged in a Y shape on the surface of the C80 fullerene molecular structure.

In the ninth arrangement pattern shown in FIG. 17, the microphone M _vw is arranged in a _{part of the node F vw from} the first layer to the fifth layer (that is, the hemisphere) corresponding to the C80 fullerene molecular structure. Specifically, the microphone M _11, M _13, M ₁₄ of the first layer, a microphone M _21, M _23, M ₂₄ of the second layer, a microphone M _31, M _35, M ₃₇ of the third hierarchy, It consists of the fourth level microphones M ₄₂ , M ₄₆ , and M ₄₈ . In the ninth arrangement pattern, as shown in FIG. 17A, the microphone M _vw is spirally arranged on the surface of the C80 fullerene molecular structure.

(Effect of sound source exploration device according to the second embodiment)
The effect of the sound source exploration device 1A according to the second embodiment will be described. Here, an experiment was conducted to estimate the direction of arrival using impact-induced abnormal noise as a noise source. The microphone array 3A of the 5th to 9th arrangement patterns shown in FIGS. 13 to 17 was used for recording the abnormal sound. In a small laboratory with a diameter of 2 m and a height of 2 m, a microphone array 3A is installed near the center of the chamber, and the horizontal angle φ = 233 °, elevation angle θ = 35 °, and distance “from the origin, which is the center of the microphone array 3A. A shocking abnormal noise was emitted from the speaker installed at "1.05m". In the laboratory, the walls and ceiling were sound-absorbed with "120 mm" thick polywool, and the background noise level was "24.4db". The signal radiated from the speaker has the frequency characteristics shown in FIG. 18 and the time waveform shown in FIG. 19, and was adjusted so that the noise level was "44.2db" at the center of the microphone array 3A.

FIG. 20 shows the estimation result of the fifth arrangement pattern (80ch) of the microphone. FIG. 20A shows the estimation result of the frequency 90Hz, and FIG. 20B shows the estimation result of the frequency 270Hz. As shown in FIG. 20 (b), in the region of 270 Hz, the estimation result is not blurred, and the peak of the MUSIC spectrum is in the direction close to the true sound source direction (horizontal angle φ = 233 °, elevation angle θ = 35 °). Is formed. On the other hand, as shown in FIG. 20A, in the low frequency range of 90 Hz, the estimation result tends to be blurred as a whole compared to the frequency of 270 Hz, and the true sound source direction (horizontal angle φ = 233 °, elevation angle θ =). The peak of the MUSIC spectrum is likely to be formed in a direction different from 35 °). Therefore, the fifth arrangement pattern (80ch) of the microphone is effective for searching for a sound source near a frequency of 270 Hz.

FIG. 21 shows the estimation result of the sixth arrangement pattern (50ch) of the microphone. FIG. 21A shows the estimation result of the frequency 90Hz, and FIG. 21B shows the estimation result of the frequency 270Hz. As shown in FIG. 21 (b), in the region of 270 Hz, the estimation result is not blurred, and the peak of the MUSIC spectrum is in the direction close to the true sound source direction (horizontal angle φ = 233 °, elevation angle θ = 35 °). Is formed. On the other hand, as shown in FIG. 21 (a), in the low frequency range of 90 Hz, the estimation result tends to be blurred as a whole compared to the frequency of 270 Hz, and the true sound source direction (horizontal angle φ = 233 °, elevation angle θ =). The peak of the MUSIC spectrum is likely to be formed in a direction different from 35 °). Therefore, the sixth arrangement pattern (50ch) of the microphone is effective for searching for a sound source near a frequency of 270 Hz. It can be said that the estimation result of the sixth arrangement pattern of the microphone is better than the estimation result of the fifth arrangement pattern throughout each region.

FIG. 22 shows the estimation result of the seventh arrangement pattern (40ch) of the microphone. FIG. 22A shows the estimation result of the frequency 90Hz, and FIG. 22B shows the estimation result of the frequency 270Hz. As shown in FIG. 22 (b), in the region of 270 Hz, the estimation result is not blurred, and the peak of the MUSIC spectrum is in the direction close to the true sound source direction (horizontal angle φ = 233 °, elevation angle θ = 35 °). Is formed. On the other hand, as shown in FIG. 22A, in the low frequency range of 90 Hz, the estimation result tends to be blurred as a whole compared to the frequency of 270 Hz, and the true sound source direction (horizontal angle φ = 233 °, elevation angle θ =). The peak of the MUSIC spectrum is likely to be formed in a direction different from 35 °). Therefore, the seventh arrangement pattern (40ch) of the microphone is effective for searching for a sound source near a frequency of 270 Hz. It can be said that the estimation result of the 7th arrangement pattern of the microphone is better than the estimation result of the 6th arrangement pattern throughout each region.

FIG. 23 shows the estimation result of the eighth arrangement pattern (three-dimensional Y-shape) of the microphone. FIG. 23A shows the estimation result of the frequency 90Hz, and FIG. 23B shows the estimation result of the frequency 270Hz. As shown in FIGS. 23 (a) and 23 (b), the true sound source direction (horizontal angle φ = 233 °, elevation angle θ = 35 °) in the 90 Hz and 270 Hz regions as compared with the fifth, sixth, and seventh arrangement patterns. ), The peak of the MUSIC spectrum is formed. Therefore, it can be said that the eighth arrangement pattern (three-dimensional Y-shape) has a better resolution than the fifth, sixth, and seventh arrangement patterns. In the eighth arrangement pattern (three-dimensional Y-shape), the estimation direction in which the MUSIC spectrum is maximized is as follows.
・ In the case of "90Hz", the horizontal angle φ is "242 ° (+ 9 °)" and the elevation angle θ is "19 ° (-16 °)".
・ In the case of "270Hz", the horizontal angle φ is "240 ° (+ 7 °)" and the elevation angle θ is "18 ° (-17 °)".

FIG. 24 shows the estimation result of the ninth arrangement pattern (three-dimensional spiral) of the microphone. FIG. 24A shows the estimation result of the frequency 90Hz, and FIG. 24B shows the estimation result of the frequency 270Hz. As shown in FIGS. 24 (a) and 24 (b), the true sound source direction (horizontal angle φ = 233 °, elevation angle θ = 35 °) in the 90 Hz and 270 Hz regions as compared with the fifth, sixth, and seventh arrangement patterns. ), The peak of the MUSIC spectrum is formed. Therefore, it can be said that the 9th arrangement pattern (three-dimensional spiral) has a better resolution than the 5th, 6th, and 7th arrangement patterns. In the ninth arrangement pattern (stereoscopic spiral), the estimation direction in which the MUSIC spectrum is maximum is as follows.
・ In the case of "90Hz", the horizontal angle φ is "231 ° (-2 °)" and the elevation angle θ is "24 ° (-11 °)".
・ In the case of "270Hz", the horizontal angle φ is "227 ° (-5 °)" and the elevation angle θ is "40 ° (+ 5 °)".

As can be seen, the ninth arrangement pattern (three-dimensional spiral) has a smaller error angle at the same frequency than the eighth arrangement pattern (three-dimensional Y-shape), so that an estimation result closer to the true sound source direction can be obtained. Therefore, it can be said that the ninth arrangement pattern (three-dimensional spiral) of the microphone has a better resolution than the eighth arrangement pattern (three-dimensional Y-shape).

Next, the arrangement pattern of another microphone M based on the C80 fullerene molecular structure will be described with reference to FIGS. 25 to 28.
In the tenth arrangement pattern shown in FIG. 25, the microphone M _vw is arranged in a _{total of five nodes F vw in the first layer corresponding to the C80 fullerene molecular structure.} Specifically, it consists of first-level microphones M _{11 to} M ₁₅ .
In the eleventh arrangement pattern shown in FIG. 26, the microphone M _vw is arranged in a _{total of five nodes F vw in the second layer corresponding to the C80 fullerene molecular structure.} Specifically, it consists of microphones M _{21 to} M ₂₅ of the second layer.

In the twelfth arrangement pattern shown in FIG. 27, the microphone M _vw is arranged in a _{total of 10 nodes F vw in the third layer corresponding to the C80 fullerene molecular structure.} Specifically, it consists of microphones M _{31 to} M _{310 on} the third layer.
In the thirteenth arrangement pattern shown in FIG. 28, the microphone M _vw is arranged in a _{total of ten nodes F vw in the fourth layer corresponding to the C80 fullerene molecular structure.} Specifically, it consists of fourth-level microphones M _{41 to} M ₄₁₀ .

FIG. 29 shows the estimation result of the tenth arrangement pattern (first layer) of the microphone. FIG. 29A shows the estimation result of the frequency 90Hz, and FIG. 29B shows the estimation result of the frequency 270Hz. As shown in FIG. 29 (a), the estimation result is totally blurred in the low frequency range of 90 Hz, and the true sound source direction (horizontal angle φ) is shown through FIGS. 29 (a) and 29 (b). = 233 °, elevation angle θ = 35 °) The estimated value is also obtained in the opposite direction.

FIG. 30 shows the estimation result of the eleventh arrangement pattern (second layer) of the microphone. FIG. 30A shows the estimation result of the frequency 90Hz, and FIG. 30B shows the estimation result of the frequency 270Hz. The estimation result of the eleventh arrangement pattern (second layer) of the microphone is substantially the same as that of the tenth arrangement pattern (first layer). That is, as shown in FIG. 30 (a), the estimation result is totally blurred in the low frequency range of 90 Hz, and the true sound source direction (horizontal) is shown through FIGS. 30 (a) and 30 (b). Estimated values are also obtained in the direction opposite to the angle φ = 233 ° and elevation angle θ = 35 °). It should be noted that the resolution in the low frequency range of 90 Hz is better than that of the tenth arrangement pattern (first layer).

FIG. 31 shows the estimation result of the twelfth arrangement pattern (third layer) of the microphone. FIG. 31 (a) shows the estimation result of the frequency 90 Hz, and FIG. 31 (b) shows the estimation result of the frequency 270 Hz. The estimation result of the twelfth arrangement pattern (third layer) of the microphone is substantially the same as the tenth arrangement pattern (first layer) and the eleventh arrangement pattern (second layer). That is, as shown in FIG. 31 (a), the estimation result is totally blurred in the low frequency range of 90 Hz, and the true sound source direction (horizontal) is obtained through FIGS. 31 (a) and 31 (b). Estimated values are also obtained in the direction opposite to the angle φ = 233 ° and elevation angle θ = 35 °). It should be noted that the resolution in the low frequency range of 90 Hz is better than that of the 11th arrangement pattern (second layer).

FIG. 32 shows the estimation result of the thirteenth arrangement pattern (fourth layer) of the microphone. FIG. 32A shows the estimation result of the frequency 90Hz, and FIG. 32B shows the estimation result of the frequency 270Hz. The estimation result of the thirteenth arrangement pattern (fourth layer) of the microphone is substantially the same as the tenth arrangement pattern (first layer) to the twelfth arrangement pattern (third layer). That is, as shown in FIG. 32 (a), the estimation result is totally blurred in the low frequency range of 90 Hz, and the true sound source direction (horizontal) is obtained through FIGS. 32 (a) and 32 (b). Estimated values are also obtained in the direction opposite to the angle φ = 233 ° and elevation angle θ = 35 °). It should be noted that the resolution in the low frequency range of 90 Hz is better than that of the twelfth arrangement pattern (third layer).

_{In this way, even if} the microphone M is arranged in the section F vw of the similar model of the C80 fullerene molecular structure, the true sound source is the two-dimensional planar arrangement pattern (10th arrangement pattern to 13th arrangement pattern). The estimated value also increases in the direction opposite to the direction. This is because the position where the elevation angle θ is rotated by “90 °” is under the same conditions as before the rotation, so that the estimated value is also obtained in the opposite direction. _{Therefore, even if} the microphone M is arranged in the section F vw of the similar model of the C80 fullerene molecular structure, the two-dimensional planar arrangement pattern (10th arrangement pattern to 13th arrangement pattern) is not preferable.
Further, the resolution in the low frequency range of 90 Hz improves from the 10th arrangement pattern to the 13th arrangement pattern. Therefore, in estimating the arrival direction of the sound source in the low frequency range, it is particularly preferable to increase the radius of the reference circle (concentric circle) which is the reference for the arrangement of the microphone M.

As described above, according to the sound source exploration device 1A according to the second embodiment, it is possible to obtain substantially the same effect as the sound source exploration device 1 according to the first embodiment.
Further, in the sound source exploration device 1A according to the second embodiment, since the microphone M is arranged in the section F of the similarity model of the C80 fullerene molecular structure, the position where the microphone M is arranged can be easily determined.
Further, in the fifth to ninth arrangement patterns shown in FIGS. 13 to 17, all the microphones M are not located on the same plane, and the microphones M are arranged three-dimensionally. Therefore, it is extremely unlikely that the problem of not being able to estimate the direction of arrival of the sound source will occur.

[Modification example]
Although the embodiments of the present invention have been described above, the present invention is not limited to this, and can be carried out without changing the gist of the claims. A modification of the embodiment is shown below.

In the first embodiment, the microphone M is arranged on the outer periphery of the cut surface obtained by cutting a cone (see FIG. 4 (a)) or a hemisphere (see FIG. 4 (b)). However, it suffices if the microphone M can be arranged three-dimensionally, and the microphone M may be arranged on a concentric circle created from a figure other than a cone or a hemisphere.

In the second embodiment, the microphone M is arranged in the upper hemisphere of the C80 fullerene molecular structure. However, it suffices if the microphone M can be arranged three-dimensionally, and the microphone M may be arranged in the lower hemisphere or in the entire area. Further, the microphone M may be arranged based on another molecular structure (for example, C60 fullerene molecular structure).

1 Sound source search device 3 Microphone array 4 A / D converter 10 Signal processing unit M Microphone F section

Claims (2)

A sound source search device that searches for the position of a sound source using the MUSIC method.
With a microphone array consisting of multiple microphones,
It is equipped with a signal processing unit that receives sound source signals from the plurality of microphones and estimates the arrival direction of the sound source by the MUSIC algorithm.
By arranging the microphones in a plurality of layers in a similar model of the C80 fullerene molecular structure, the microphones are arranged on a plurality of concentric circles having different radii and shifting the center point in the direction of the central axis.
When the similarity model is numbered in ascending order from the upper layer with respect to the central axis of the concentric circles, and the nodes in the same layer are numbered in ascending order clockwise, the first layer is assigned. The microphone is located at the positions of the first, third, and fourth sections, the first, third, and fourth sections of the second layer, and the first, fifth, and seventh sections of the fifth layer. The microphone is placed, and the microphone is not placed at any other position.
A sound source exploration device characterized by this. A sound source search device that searches for the position of a sound source using the MUSIC method.
With a microphone array consisting of multiple microphones,
It is equipped with a signal processing unit that receives sound source signals from the plurality of microphones and estimates the arrival direction of the sound source by the MUSIC algorithm.
By arranging the microphones in a plurality of layers in a similar model of the C80 fullerene molecular structure, the microphones are arranged on a plurality of concentric circles having different radii and shifting the center point in the direction of the central axis.
When the similarity model is numbered in ascending order from the upper layer with respect to the central axis of the concentric circles, and the nodes in the same layer are numbered in ascending order clockwise, the first layer is assigned. The first, third, and fourth sections, the first, third, and fourth sections of the second layer, the first, fifth, and seventh sections of the third layer, and the second of the fourth layer. The microphone is placed at the position of the sixth, sixth, and eighth nodes, and the microphone is not placed at other positions.
A sound source exploration device characterized by this.

Images