JP6953287B2 - Sound source search device, sound source search method and its program - Google Patents

Description

The present disclosure relates to a sound source exploration device, a sound source exploration method, and a program thereof.

For example, Patent Document 1 proposes a sound source direction estimation device capable of accurately estimating the direction of a sound source from a plurality of acoustic signals obtained by a plurality of microphone units. In Patent Document 1, the direction of a sound source is accurately estimated from a plurality of acoustic signals by taking noise countermeasures using a correlation matrix of noise signals based on a plurality of acoustic signals.

Japanese Unexamined Patent Publication No. 2014-56181

However, in Patent Document 1, the correlation matrix of noise signals is calculated based on a plurality of acoustic signals obtained by a plurality of microphone units which are observation signals. Therefore, when the noise source and the sound source to be searched exist at the same time, or when the noise is at a higher level than the sound source to be searched, it is difficult to accurately obtain the correlation matrix of only the noise component. In other words, in the method of searching for a sound source based on the signal phase difference of a plurality of acoustic signals obtained by a plurality of microphone units, if there is noise with a sound pressure level higher than that of the sound source to be searched, the noise is used for the search. There is a problem that the target sound source cannot be detected, that is, explored.

The present disclosure has been made in view of the above circumstances, and an object of the present disclosure is to provide a sound source exploration device capable of more reliably exploring the direction of a sound source of an exploration target within the exploration target range.

The sound source exploration device according to one aspect of the present disclosure is a sound source exploration device that searches the direction of a sound source to be searched, and is picked up by a microphone array composed of two or more microphone units arranged apart from each other. A correlation matrix calculation unit that calculates a first correlation matrix that is a correlation matrix of an observation signal that is an acoustic signal, and a plurality of second correlation matrices that are stored in advance in the storage unit, from the array array of the microphone array. A learning unit that calculates the weight by learning so that the linear sum obtained by multiplying each of the plurality of second correlation matrices, which is the calculated correlation matrix for each direction, by a weight is equal to the first correlation matrix, and the learning unit. It is provided with a spatial spectrum calculation unit which is a spatial spectrum showing the sound pressure intensity for each direction and calculates the spatial spectrum of the observed signal by using the weight calculated by.

It should be noted that some specific embodiments of these may be realized by using a recording medium such as a system, a method, an integrated circuit, a computer program, or a computer-readable CD-ROM, and the system, the method, the method, and the like. It may be implemented using any combination of integrated circuits, computer programs and recording media.

According to the present disclosure, it is possible to realize a sound source exploration device or the like capable of more reliably exploring the direction of the sound source of the exploration target within the exploration target range.

FIG. 1 is a diagram showing an example of the configuration of the sound source exploration system according to the first embodiment. FIG. 2 is an explanatory diagram showing the positional relationship between the microphone array in the first embodiment and the sound source direction in which the sound source to be searched is located. FIG. 3 is a spatial spectrum diagram of the observation signal observed by the microphone array in the positional relationship shown in FIG. FIG. 4 is a diagram showing an example of a detailed configuration of the sound source exploration device shown in FIG. FIG. 5 is an explanatory diagram of a method of selecting a selection unit according to the first embodiment. FIG. 6 is a diagram showing an example of the configuration of the nonlinear function unit according to the first embodiment. FIG. 7 is a flowchart showing a sound source search process of the sound source search device according to the first embodiment. FIG. 8 is a flowchart showing details of the sound source exploration process shown in FIG. 7. FIG. 9 is a spatial spectrum diagram in a comparative example. FIG. 10 is a spatial spectrum diagram according to the first embodiment. FIG. 11 is a diagram showing an example of the configuration of the sound source exploration system according to the second embodiment.

The sound source exploration device according to one aspect of the present disclosure is a sound source exploration device that searches the direction of a sound source to be searched, and is picked up by a microphone array composed of two or more microphone units arranged apart from each other. A correlation matrix calculation unit that calculates a first correlation matrix that is a correlation matrix of an observation signal that is an acoustic signal, and a plurality of second correlation matrices that are stored in advance in the storage unit, from the array array of the microphone array. A learning unit that calculates the weight by learning so that the linear sum obtained by multiplying each of the plurality of second correlation matrices, which is the calculated correlation matrix for each direction, by a weight is equal to the first correlation matrix, and the learning unit. It is provided with a spatial spectrum calculation unit which is a spatial spectrum showing the sound pressure intensity for each direction and calculates the spatial spectrum of the observed signal by using the weight calculated by.

With this configuration, it is possible to more reliably search the direction of the sound source of the search target within the search target range. Further, since the spatial spectrum of the observation signal is calculated using the weight calculated by learning, it is possible to realize a sound source exploration device having excellent noise resistance and followability to changes in sound.

Here, for example, the sound source exploration device further includes the first element, which is one of the elements constituting the first correlation matrix, and the elements constituting each of the plurality of second correlation matrices. The learning unit includes a selection unit that selects a second element that is an element at a position corresponding to the first element and sequentially switches between the first element and the second element to be selected, and the learning unit is the second element. The first weight is updated to the second weight calculated by the learning so that the linear sum of the first elements obtained by multiplying the first weight by the first element is equal to the first element. The second weight was calculated by the learning so that the linear sum of the second elements multiplied by the second element selected by the selection unit would then be equal to the first element selected by the selection unit. The weight may be calculated by the learning by repeating the updating to the third weight sequentially.

As a result, the weights that are simultaneously equal for each corresponding matrix element of the first correlation matrix and the plurality of second correlation matrices can be calculated by learning, so that the acoustic signal picked up by the microphone array consisting of three or more microphone units can be obtained. Based on this, the direction of the sound source of the exploration target in the exploration target range can be searched more reliably.

Further, for example, the selection unit is a set of one of two sets of elements separated by the diagonal component among the elements excluding the diagonal components constituting the first correlation matrix and the second correlation matrix. The first element and the second element may be selected only from a plurality of elements.

As a result, the amount of calculation can be reduced, so that the direction of the sound source of the search target within the search target range can be searched at a higher speed.

Further, for example, the learning unit uses an LMS (Least Mean Square) algorithm or an ICA (Independent Component Analysis) to obtain an error that is the difference between the linear sum and the first correlation matrix and the second correlation matrix. Therefore, the weight may be calculated.

As a result, the strength for each direction can be calculated while canceling the influences between the directions, so that a sound source exploration device having better noise resistance can be realized.

Further, for example, the learning unit includes a holding unit that holds weights, a linear sum calculation unit that calculates a linear sum obtained by multiplying each of the plurality of second correlation matrices by the weights held by the holding unit, and the linearity. The weight update amount is calculated from the product of the error calculation unit that calculates the sum and the difference between the first correlation matrix and the error and the second correlation matrix, and the weight update amount is added to the weight held by the holding unit. May be provided with a weight updating unit which is a weight held by the holding unit by adding the above.

Here, for example, the weight update unit may calculate the weight update amount from the error and the second correlation matrix by using the LMS algorithm or the ICA.

Further, for example, the learning unit further includes a non-linear function unit that adds non-linearity to the error by using a predetermined non-linear function, and the weight updating unit is the non-linearity added by the non-linear function unit. The weight update amount may be calculated from the error and the second correlation matrix, and the weight update amount may be added to the weight held by the holding unit to obtain the weight held by the holding unit.

As a result, it is possible to give non-linearity to the calculated error and suppress mutual influence between directions, so that a sound source exploration device having better noise resistance can be realized.

Further, the sound source exploration method according to one aspect of the present disclosure is a sound source exploration method for exploring the direction of a sound source to be explored, and is collected by a microphone array composed of two or more microphone units arranged apart from each other. An array of the microphone array, which is a correlation matrix calculation step for calculating a first correlation matrix which is a correlation matrix of an observation signal which is a sounded acoustic signal, and a plurality of second correlation matrices stored in advance in a storage unit. The learning step of calculating the weight by learning so that the linear sum obtained by multiplying each of the plurality of second correlation matrices, which is the correlation matrix for each direction calculated from the array, by the weight is equal to the first correlation matrix. The spatial spectrum calculation step of calculating the spatial spectrum of the observed signal, which is a spatial spectrum showing the sound pressure intensity for each direction by using the weight calculated in the learning step, is included.

Further, the program according to one aspect of the present disclosure is a program for causing a computer to execute a sound source search method for searching the direction of a sound source to be searched, and is composed of two or more microphone units arranged apart from each other. A correlation matrix calculation step for calculating a first correlation matrix which is a correlation matrix of an observation signal which is an acoustic signal picked up by a computer array, and a plurality of second correlation matrices stored in advance in a storage unit. , The weight is calculated by learning so that the linear sum obtained by multiplying each of the plurality of second correlation matrices, which is the correlation matrix for each direction calculated from the array array of the computer array, by the weight is equal to the first correlation matrix. The computer is made to execute the learning step to be performed and the spatial spectrum calculation step of calculating the spatial spectrum of the observed signal, which is a spatial spectrum showing the sound pressure intensity for each direction by using the weight calculated in the learning step. ..

It should be noted that some specific embodiments of these may be realized by using a recording medium such as a system, a method, an integrated circuit, a computer program, or a computer-readable CD-ROM, and the system, the method, the method, and the like. It may be implemented using any combination of integrated circuits, computer programs or recording media.

Hereinafter, the sound source exploration device according to one aspect of the present disclosure will be specifically described with reference to the drawings. It should be noted that all of the embodiments described below show a specific example of the present disclosure. Numerical values, shapes, materials, components, arrangement positions of components, and the like shown in the following embodiments are examples, and are not intended to limit the present disclosure. Further, among the components in the following embodiments, the components not described in the independent claims indicating the highest level concept are described as arbitrary components. Moreover, in all the embodiments, each content can be combined.

(Embodiment 1)
FIG. 1 is a diagram showing an example of the configuration of the sound source exploration system 1000 according to the first embodiment.

The sound source exploration system 1000 is used to search the direction of the sound source to be searched. In the present embodiment, as shown in FIG. 1, the sound source exploration system 1000 includes a sound source exploration device 1, a microphone array 200, and a frequency analysis unit 300.

[Microphone Array 200]
The microphone array 200 is composed of two or more microphone units arranged apart from each other, and outputs an acoustic signal converted into an electric signal by observing, that is, collecting sound waves coming from all directions. In the present embodiment, the microphone array 200 will be described below assuming that the microphone array 200 is composed of three microphone units, that is, microphone units 201, 202, 203. The microphone unit 201, the microphone unit 202, and the microphone unit 203 are, for example, omnidirectional microphone elements having high sensitivity to sound pressure, and are arranged apart from each other (in other words, at different positions). Here, the microphone unit 201 outputs an acoustic signal m1 (n) which is a time domain signal obtained by converting the pickled sound wave into an electric signal. Similarly, the microphone unit 202 outputs an acoustic signal m2 (n) which is a signal in the time region obtained by converting the collected sound sound into an electric signal, and the microphone unit 203 outputs the time when the collected sound sound is converted into an electric signal. The acoustic signal m3 (n), which is a signal of the region, is output.

FIG. 2 is an explanatory diagram showing the positional relationship between the microphone array 200 in the first embodiment and the sound source direction in which the sound source S to be searched is located. FIG. 3 is a spatial spectrum diagram of the observation signal observed by the microphone array 200 in the positional relationship shown in FIG. FIG. 2 shows the configuration of a microphone array 200 including an array array in which the microphone unit 201, the microphone unit 202, and the microphone unit 203 are arranged in a row on the axis of θ = 0 degrees. Further, FIG. 2 shows a case where the sound source S to be searched exists in the direction of θ = θs with respect to the microphone array 200, and there is no sound source that becomes a disturbing sound. In this case, the spatial spectrum that is the exploration result of the sound source exploration device 1 is as shown in FIG. That is, the angle showing the highest intensity in the spatial spectrum shown in FIG. 3 which is the exploration result is θs.

[Frequency analysis unit 300]
The frequency analysis unit 300 converts the acoustic signal observed in each of the two or more microphone units into a signal in the frequency domain and outputs it as a frequency spectrum signal. More specifically, the frequency analysis unit 300 performs frequency analysis on the acoustic signal input from the microphone array 200, and outputs a frequency spectrum signal which is a signal in the frequency domain. For frequency analysis, one that transforms a time signal into amplitude information and phase information for each frequency component, such as Fast Fourier Transform (FFT) or Discrete Fourier Transform (DFT), may be used.

In the present embodiment, the frequency analysis unit 300 is composed of FFT301, FFT302, and FFT303 that perform a fast Fourier transform. The FFT 301 receives the acoustic signal m1 (n) output from the microphone unit 201 as an input, performs conversion from the time domain to the frequency domain using a fast Fourier transform, and outputs the frequency spectrum signal Sm1 (ω). The FFT 302 takes the acoustic signal m2 (n) output from the microphone unit 202 as an input, performs conversion from the time domain to the frequency domain using a fast Fourier transform, and outputs the frequency spectrum signal Sm2 (ω). The FFT 303 takes the acoustic signal m3 (n) output from the microphone unit 203 as an input, performs conversion from the time domain to the frequency domain using a fast Fourier transform, and outputs the frequency spectrum signal Sm3 (ω).

[Sound source exploration device 1]
FIG. 4 is a diagram showing an example of the detailed configuration of the sound source exploration device 1 shown in FIG.

The sound source search device 1 searches for the direction of the sound source to be searched. In the present embodiment, as shown in FIGS. 1 and 4, the sound source exploration device 1 includes a correlation matrix calculation unit 10, a storage unit 20, a selection unit 30, a learning unit 40, and a spatial spectrum calculation unit 100. , The output unit 110 is provided. The sound source search device 1 does not have to include the selection unit 30 as long as the number of microphone units constituting the microphone array 200 is 2. Further, the sound source search device 1 may include a microphone array 200 and a frequency analysis unit 300. Hereinafter, each component will be described.

<Correlation matrix calculation unit 10>
The correlation matrix calculation unit 10 calculates a first correlation matrix which is a correlation matrix of observation signals which are acoustic signals picked up by the microphone array 200.

In the present embodiment, the correlation matrix calculation unit 10 calculates the observation correlation matrix Rx (ω), which is the first correlation matrix, from the frequency spectrum output by the frequency analysis unit 300. More specifically, the correlation matrix calculation unit 10 uses the following (Equation 1) and (Equation 2) to obtain the frequency spectrum signal Sm1 (ω) from FFT301 and the frequency spectrum signal Sm2 (ω) from FFT302. And the frequency spectrum signal Sm3 (ω) from FFT303 are input to calculate the observation correlation matrix Rx (ω).

_{Here, each element X ij} (ω) constituting the observation correlation matrix Rx (ω) is a plurality of sound waves arriving at each microphone unit, and the phase difference with respect to a plurality of sound waves from a plurality of sound sources existing in the real environment. Information is stored. _{For example, the element X 12} (ω) shown in (Equation 1) indicates the phase difference information for the sound waves arriving at the microphone unit 201 and the microphone unit 202. Further, for example, the element X ₁₃ (ω) shown in (Equation 1) indicates the phase difference information for the sound waves arriving at the microphone unit 201 and the microphone unit 203. ^{(・) *} Shown in (Equation 2) indicates the complex conjugate.

In the present embodiment, when the sound pressure sensitivity characteristics of the microphone units shown as the microphone units 201 to 203 are substantially equal and uniform, each element X _ij (ω) of the observation correlation matrix Rx (ω) is expressed by (Equation). It can be shown by 3). _{Each element X ij} (ω) of the observation correlation matrix Rx (ω) shown in (Equation 3) corresponds to the one in which the normalization term of the denominator in (Equation 2) is omitted.

<Memory unit 20>
The storage unit 20 stores in advance a plurality of second correlation matrices, which are direction-specific correlation matrices calculated from the array array of the microphone array 200.

In the present embodiment, the storage unit 20 is composed of a memory or the like, and stores in advance the reference correlation matrix Rr (θ, ω) for each search direction θ, which is the second correlation matrix. In the example shown in FIG. 4, the storage unit 20 stores in advance the _{reference correlation matrices Rr (θ 1} , ω) to Rr (θ _N , ω) in the range of 0 ≦ θ ≦ 180, for example, the number of directions N = 180. ing.

Since the reference correlation matrix Rr (θ, ω) represents the phase difference between the microphone units for the sound wave in each direction θ, it is theoretical if the direction θ of the sound source and the array array which is the microphone unit array of the microphone array 200 are determined. Can be calculated. Hereinafter, a method of calculating the reference correlation matrix Rr (θ, ω) will be described by taking the array array of the microphone array 200 shown in FIG. 2 as an example.

As described above, FIG. 2 shows an example of an array arrangement in which the microphone units 201 to 203 constituting the microphone array 200 are linearly arranged. Further, FIG. 2 also shows a positional relationship in which the sound source S exists in the direction θs.

The arrival time of the sound wave from the sound source S to the microphone units 201 to 203 is earlier in time τ in the microphone unit 201 and later in the microphone unit 203 with reference to the central microphone unit 202. The time τ can be calculated using the following (Equation 4). In (Equation 4), L indicates the distance between microphone units, and c indicates the speed of sound.

Then, the direction vector showing the phase difference relationship of the microphone units 201 to 203 with respect to the sound wave from the sound source in the direction θ can be expressed by using (Equation 5) with reference to the position of the microphone unit 202 in the center.

Therefore, when the sound source is in the direction θ, in other words, the reference correlation matrix Rr (θ, ω) with respect to the direction θ is changed to (Equation 6) from the relations of (Equation 2), (Equation 3) and (Equation 5). Can be defined as shown. In (Equation 6), (.) ^H represents a complex conjugate transpose.

In this way, the _{reference correlation matrices Rr (θ 1} , ω) to Rr (θ _N , ω) for _{the directions θ 1 to θ N} (for example, N = 180) can be calculated.

<Selection unit 30>
The selection unit 30 is an element at a position corresponding to the first element among the first element which is one of the elements constituting the first correlation matrix and the elements constituting each of the plurality of second correlation matrices. A certain second element is selected, and the selected first element and the second element are sequentially switched. Here, the selection unit 30 is a plurality of elements of one set of two sets of elements separated by diagonal components among the elements excluding the diagonal components constituting the first correlation matrix and the second correlation matrix. Only one of them needs to select the first element and the second element.

In the present embodiment, the selection unit 30 receives the observation correlation matrix Rx (ω) from the correlation matrix calculation unit 10 and the reference correlation matrix Rr (θ, ω) from the storage unit 20 as inputs, and the observation correlation matrix Rx ( ω) and the elements of the corresponding correlation matrix in the plurality of reference correlation matrices Rr (θ, ω) are selected and output. As shown in FIG. 4, for example, the selection unit 30 includes a matrix element selection unit 31 and a matrix element selection unit 32-1 to a matrix element selection unit 32-N. In FIG. 4, the matrix element selection unit 32-1 into which the reference correlation matrix Rr (θ _{1 , ω) for the direction θ 1} is input, and the reference correlation matrix Rr (θ _N , ω) _{for the direction θ N are input.} An example is shown in the case where two matrix element selection units 32-N are provided, but the present invention is not limited to this. For direction number N = 180, the direction theta ₁ through? Reference correlation matrix for _{N Rr (θ 1, ω)} ~Rr (θ N, ω) N pieces that are input matrix element selecting unit 32-1～ matrix The element selection unit 32-N is provided.

Hereinafter, an example of the selection method of the selection unit 30 will be specifically described with reference to FIG.

FIG. 5 is an explanatory diagram of a selection method of the selection unit 30 in the first embodiment.

As shown in FIG. 5, the matrix element selection unit 31 selects one of the elements (also referred to as matrix elements) constituting the observation correlation matrix Rx (ω) input from the correlation matrix calculation unit 10. It is output as a phase difference signal x (ω). The matrix element selection unit 32-m (m is a natural number of 1 or more and N or less) is selected by the matrix element selection unit 31 among the elements constituting _{the reference correlation matrix Rr (θ m, ω) input from the storage unit 20.} The same row and column elements as the elements are selected and _{output as a phase difference signal r (θ m} , ω).

Normally, the diagonal element of the correlation matrix is 1, which has no meaning in signal processing. _{Further, x ij} and x _ji in which the row number and the column number are exchanged in the correlation matrix are the same as the information because the phase rotations are opposite to each other. In consideration of these, the selection unit 30 is separated by the diagonal component among the elements excluding the diagonal component constituting the correlation matrix of the reference correlation matrix Rr (θ, ω) and the observation correlation matrix Rx (ω). It suffices to select and output an element from a plurality of elements of one set of a plurality of elements of a set. That is, if the selection unit 30 selects and outputs the elements of the upper triangular matrix or the lower triangular matrix excluding the diagonal components of the correlation matrix of the reference correlation matrix Rr (θ, ω) and the observation correlation matrix Rx (ω). good. As a result, the sound source exploration device 1 can reduce the amount of calculation.

Further, the selection unit 30 may thin out the number of elements to be selected in the upper triangular matrix or the lower triangular matrix from the viewpoint of reducing the amount of calculation.

<Learning Department 40>
The learning unit 40 calculates the weights by learning so that the linear sum obtained by multiplying each of the plurality of second correlation matrices stored in advance in the storage unit 20 by the weights becomes equal to the first correlation matrix. Here, the learning unit 40 calculates the weight from the error which is the difference between the linear sum and the first correlation matrix and the second correlation matrix by using the LMS algorithm or ICA (Independent Component Analysis). More specifically, the learning unit 40 makes the first element linear sum obtained by multiplying the second element selected by the selection unit 30 by the first weight equal to the first element selected by the selection unit 30. , The first weight is updated to the second weight calculated by learning. Subsequently, in the learning unit 40, the second element linear sum obtained by multiplying the updated second weight by the second element selected by the selection unit 30 is then combined with the first element selected by the selection unit 30. The second weight is updated to the third weight calculated by learning so that they are equal. The learning unit 40 calculates the weight by learning by sequentially repeating these updates.

In the present embodiment, as shown in FIGS. 1 and 4, the learning unit 40 includes a holding unit 50, a linear sum calculation unit 60, an error calculation unit 70, a nonlinear function unit 80, and a weight update unit 90. To be equipped. The non-linear function unit 80 is not an indispensable configuration, and the learning unit 40 does not have to include the non-linear function unit 80.

≪Holding part 50≫
The holding unit 50 holds the weight updated by the weight updating unit 90. The holding unit 50 holds a weight to be multiplied for each reference correlation matrix Rr (θ, ω). In other words, the weight is common to each element constituting the correlation matrix of the reference correlation matrix Rr (θ ₁ , ω) to Rr (θ _{N, ω).}

The weight is a function with θ and ω as variables, but ω can be treated as a one-dimensional coefficient by treating it as a constant. Hereinafter, the weight will be described with reference to the weight coefficient a (θ, ω).

In the present embodiment, the weighting coefficient a (θ, ω) is a coefficient multiplied by the reference correlation matrix Rr (θ, ω) for each direction θ. As an example, FIG. 4 shows a weighting coefficient a (θ _{1 ) in the directions θ 1 to θ N} (N = 180) corresponding to the reference correlation matrix Rr (θ, ω) in the range of 180 0 ≦ θ ≦ 180. , Ω) to a (θ _N , ω) are shown.

The holding unit 50 holds the weighting coefficient a (θ, ω) updated by the weight updating unit 90. That is, the weight coefficient a (θ, ω) is a learning coefficient whose value is updated based on the weight update amount calculated by the weight update unit 90. Further, the holding unit 50 outputs the weighting coefficient a (θ, ω) to be held to the space spectrum calculation unit 100.

<< Linear sum calculation unit 60 >>
The linear sum calculation unit 60 calculates a linear sum obtained by multiplying each of the plurality of second correlation matrices by the weights held by the holding unit 50.

In the present embodiment, as shown in FIG. 4, the linear sum calculation unit 60 includes a signal multiplication unit 61-1 to a signal multiplication unit 61-N and a signal addition unit 62.

The signal multiplication unit 61-1 has a weighting coefficient a (θ ₁ ) in the direction θ _{1 on} the element r (θ _{1 , ω) of the reference correlation matrix Rr (θ 1} , ω) selected by the matrix element selection unit 32-1. , Ω) Multiply and output to the signal addition unit 62. Similarly, the signal multiplication unit 61-N has a weighting coefficient of the _{direction θ N on} the element r (θ _N , ω) of _{the reference correlation matrix Rr (θ N} , ω) selected by the matrix element selection unit 32-N. Multiply by a (θ _N , ω) and output to the signal addition unit 62. Thus, each signal multiplier unit 61-1～ signal multiplication unit 61-N, the direction theta ₁ through? _N reference correlation matrix in each Rr (θ, ω) the weighting coefficient a (θ, ω) multiplied by the signal Is output to the signal addition unit 62.

The signal addition unit 62 outputs an estimated phase difference signal xr (ω) obtained by adding the signals output from the signal multiplication units 61-1 to the signal multiplication units 61-N to the error calculation unit 70. More specifically, the signal addition unit 62 uses (Equation 7) to calculate the linear sum of the signals output from the signal multiplication units 61-1 to the signal multiplication units 61-N by estimating the phase difference signal xr (ω). ).

<< Error calculation unit 70 >>
The error calculation unit 70 calculates an error which is the difference between the linear sum calculated by the linear sum calculation unit 60 and the first correlation matrix. In the present embodiment, the error calculation unit 70 includes a signal subtraction unit 71 as shown in FIG.

The signal subtraction unit 71 calculates the error signal e (ω) by subtracting the estimated phase difference signal xr (ω) from the signal addition unit 62 from the phase difference signal x (ω) from the matrix element selection unit 31. .. More specifically, the signal subtraction unit 71 calculates the error signal e (ω) using (Equation 8).

<< Non-linear function part 80 >>
The non-linear function unit 80 adds non-linearity to the error by using a predetermined non-linear function. More specifically, the non-linear function unit 80 converts the error signal e (ω) input from the signal subtraction unit 71 into a signal to which non-linearity is added by a non-linear function which is a function having non-linear input / output characteristics. Non-linear functions are, for example, high public tangents, but are not limited to this. Any non-linear function having non-linear input / output characteristics that can limit the signal amplitude may be used. This is because even if the phase difference is disturbed by the disturbance and the error signal e (ω) temporarily increases, the influence on the weight update amount learned by the weight update unit 90, which will be described later, can be suppressed.

FIG. 6 is a diagram showing an example of the configuration of the nonlinear function unit 80 according to the first embodiment. As shown in FIG. 6, the non-linear function unit 80 includes a real part extraction unit 801, an imaginary part extraction unit 802, a non-linear addition unit 803, a non-linearity addition unit 804, an imaginary unit multiplication unit 805, and signal addition. A unit 806 is provided.

The real part extraction unit 801 extracts the real part of the input error signal e (ω) and outputs it to the non-linearity addition unit 803. The imaginary part extraction unit 802 extracts the imaginary part of the input error signal e (ω) and outputs it to the non-linearity addition unit 804.

The non-linearity addition unit 803 adds non-linearity to the signal amplitude of the real number part of the error signal e (ω) input from the real part extraction unit 801 by a non-linear function, and outputs it to the signal addition unit 806. The non-linearity addition unit 804 adds non-linearity to the signal amplitude of the imaginary part of the error signal e (ω) input from the imaginary part extraction unit 802 by a nonlinear function, and outputs it to the imaginary unit multiplication unit 805.

The imaginary unit multiplication unit 805 multiplies the imaginary unit j and outputs the signal to the signal addition unit 806 in order to return the signal input from the non-linearity addition unit 804 to an imaginary number. The signal addition unit 806 is a complex signal to which the non-linearity added by adding the signal input from the non-linearity addition unit 803 which is the real part signal and the signal input from the imaginary unit multiplication unit 805 which is the imaginary part signal is added. f (e (ω)) is output to the weight update unit 90.

An example of the complex signal f (e (ω)) to which the non-linearity is added is shown in (Equation 9). (Equation 9) is an example when the high public tangent tanh (・) is used for the nonlinear function. real (・) represents the real part, imag (・) represents the imaginary part, and j is the imaginary unit.

≪Weight update part 90≫
The weight update unit 90 calculates the weight update amount from the error and the second correlation matrix by using the LMS (Least Mean Square) algorithm or the ICA (Independent Component Analysis), and the weight held by the holding unit 50 is the weight. The update amount is added to obtain the weight held by the holding unit 50. When the sound source exploration device 1 includes the non-linear function unit 80, the weight update unit 90 calculates the weight update amount from the error to which the non-linearity is added by the non-linear function unit 80 and the second correlation matrix. The weight update amount is added to the weight held by the holding unit 50 to obtain the weight held by the holding unit 50.

In the present embodiment, the weight updating unit 90 includes the complex signal f (e (ω)) input from the nonlinear function unit 80 and the N phase difference signals r (θ ₁ , ω) input from the selection unit 30. ) To R (θ _N , ω) are input. Then, the weight updating unit 90 has weight coefficients a (θ ₁ , ω) to a (θ _{N , ω) multiplied by N phase difference signals r (θ 1} , ω) to r (θ N, ω). The weight update amount Δa (θ ₁ , ω) to Δa (θ _N , ω) with respect to is calculated.

For example, when the sound source search device 1 does not include the nonlinear function unit 80, the weight update unit 90 uses (Equation 10) to update the weights Δa (θ ₁ , ω) to Δa (θ _N , ω). Is calculated. On the other hand, when the sound source exploration device 1 includes the nonlinear function unit 80, the weight update unit 90 uses (Equation 11) to set the weight update amounts Δa (θ ₁ , ω) to Δa (θ _N , ω). calculate.

In addition, (Equation 10) and (Equation 11) show the case where the weight update amount is calculated by using the LMS algorithm. β is a parameter that controls the update speed. Further, in the correlation matrix, there is a phase inversion relationship between _{the elements r ij} (ω) and r _{ji (ω).} Therefore, in (Equation 10) and (Equation 11), the real (・) part is provided because the imaginary part is canceled due to the complex conjugate relationship.

Then, as shown in the following (Equation 12), the weight updating unit 90 updates the weight coefficient a (θ _k , ω) held by the holding unit 50 by using the calculated weight updating amount.

<< Spatial spectrum calculation unit 100 >>
The spatial spectrum calculation unit 100 uses the weights calculated by the learning unit 40 to calculate the spatial spectrum of the observed signal, which is a spatial spectrum indicating the sound pressure intensity for each direction.

_{In the present embodiment, the spatial spectrum calculation unit 100 inputs the weight coefficients a (θ 1} , ω) to a (θ _N , ω) held by the holding unit 50 and updated by learning by the weight updating unit 90. , The spatial spectrum p (θ) is calculated and output to the output unit 110.

More specifically, as shown in the following (Equation 13), the spatial spectrum calculation unit 100 calculates the sum or average of the weighting coefficients a (θ, ω) held by the holding unit 50 with respect to the frequency ω. The spatial spectrum p (θ) can be obtained. The principle will be described later, but the weighting coefficient a (θ, ω) can be treated as a function indicating the intensity of the sound wave for each direction θ and frequency ω.

[Operation of sound source search device 1]
The sound source exploration process performed by the sound source exploration device 1 configured as described above will be described.

FIG. 7 is a flowchart showing the sound source search process of the sound source search device 1 according to the first embodiment.

First, the sound source exploration device 1 performs a correlation matrix calculation process of the observed signal (S10). More specifically, the sound source exploration device 1 is an observation correlation matrix which is a correlation matrix of observation signals which are acoustic signals picked up by a microphone array 200 composed of two or more microphone units arranged apart from each other. Calculate Rx (ω).

Next, the sound source search device 1 performs a learning process of weights to be multiplied by each reference correlation matrix (S20). More specifically, the sound source search device 1 is a plurality of reference correlation matrices Rr (θ, ω) stored in advance in the storage unit 20, and is a correlation matrix for each direction calculated from the array array of the microphone array. The weights are calculated by learning so that the linear sum obtained by multiplying each of the plurality of reference correlation matrices Rr (θ, ω) is equal to the observed correlation matrix Rx (ω).

Next, the sound source exploration device 1 performs a spatial spectrum calculation process of the observation signal (S30). More specifically, the sound source exploration device 1 uses the weight calculated in step S20 to obtain a spatial spectrum p (θ) indicating the sound pressure intensity for each direction and a spatial spectrum p (θ) of the observed signal. calculate.

FIG. 8 is a flowchart showing details of the sound source exploration process shown in FIG. 7. The same elements as those in FIG. 7 are designated by the same reference numerals.

More specifically, first, in step S10, the microphone array 200 acquires an acoustic signal at time t (S101). Next, the frequency analysis unit 300 performs frequency analysis of the acoustic signal acquired in step S101 (S102) and converts it into a frequency spectrum signal which is a signal in the frequency domain. Then, the sound source exploration device 1 calculates the observation correlation matrix Rx (ω), which is the correlation matrix of the observation signals at time t, from the frequency spectrum signal converted in step S102 (S103).

Next, in step S20, first, a predetermined number of times Nt is set in the sound source exploration device 1 as the number of times the weight learning process is performed (S201). Next, the sound source exploration device 1 selects the corresponding matrix elements of the observation correlation matrix Rx (ω) and the reference correlation matrix Rr (θ, ω), and selects the phase difference signal x (ω) and the phase difference signal r (θ, ω). ) Is output (S202). Next, the sound source search device 1 calculates the error signal e (ω) from the phase difference signal x (ω), the phase difference signal r (θ, ω), and the weighting coefficient a (θ, ω) (S203). ). Next, the sound source search device 1 calculates a complex signal f (e (ω)) obtained by adding non-linearity to the error signal e (ω) (S204). Next, the sound source exploration device 1 has a weighting coefficient a (θ, ω) from the complex signal f (e (ω)) calculated in step S204 and the phase difference signal r (θ, ω) calculated in step S203. The weight update amount Δa (θ, ω) is calculated, and the weight coefficient a (θ, ω) is updated (S205). Then, the sound source exploration device 1 determines whether the matrix elements of the observation correlation matrix Rx (ω) and the reference correlation matrix Rr (θ, ω) selected in step S202 have made a round (S206). When one cycle is completed (YES in S206), it is determined whether the number of times of learning processing of the weighting coefficient a (θ, ω) has reached a predetermined number of times Nt (S207). When the predetermined number of times Nt is reached (YES in S207), the sound source search device 1 proceeds to the next step S30. In step S206 or step S207, if the cycle has not been completed (NO in S206) or the predetermined number of times Nt has not been reached (NO in S207), the process returns to step S202 and the process is repeated.

Next, in step S30, the sound source exploration device 1 calculates the spatial spectrum p (θ) of the observed signal from the weighting coefficients a (θ, ω) updated by the learning in step S20 (S301).

Next, in step S40, the sound source search device 1 updates the time t to, for example, time t + Δt, and determines in step S50 whether to end the sound source search process. If the sound source search process is not completed (NO in S50), the process returns to step S10, and the observation correlation matrix Rx (ω), which is the correlation matrix of the observation signals at time t + Δt, is calculated.

In this way, in the sound source exploration apparatus 1, the linear sum obtained by multiplying each of the plurality of reference correlation matrices Rr (θ, ω) by the weighting coefficient a (θ, ω) is equal to the observation correlation matrix Rx (ω). , The learning for each corresponding matrix element is repeated until the weighting coefficients a (θ, ω) for all the matrix elements are learned. Further, the sound source exploration device 1 may repeat learning Nt a predetermined number of times.

For example, if the reference correlation matrix Rr (θ, ω) and the observation correlation matrix Rx (ω) of 3 rows and 3 columns and the predetermined number of times Nt is 3 times, then for the three elements of the upper triangular matrix or the lower triangular matrix. Since the learning process is performed three times, the learning process is performed a total of nine times.

In this way, the linear sum obtained by multiplying each of the plurality of reference correlation matrices Rr (θ, ω) by the weighting coefficient a (θ, ω) is more equal to the observed correlation matrix Rx (ω). ) Can be learned.

[Principle of operation]
Next, the weighting coefficient a (θ, ω) at which the linear sum obtained by multiplying each of the plurality of reference correlation matrices Rr (θ, ω) by the weighting coefficient a (θ, ω) is equal to the observed correlation matrix Rx (ω) is obtained. The principle that can be calculated by learning will be explained. In addition, the principle that the spatial spectrum p (θ) can be calculated using the obtained weighting coefficients a (θ, ω) will also be described.

The observation correlation matrix Rx (ω) observed based on the signal from the microphone array 200, that is, the observation correlation matrix Rx (ω) which is the output of the correlation matrix calculation unit 10, is as shown in the following (Equation 14). It is known that it can be approximated by the linear sum of the correlation matrix Rs (θ, ω) and the intensity u (θ, ω) with respect to the sound source in the space existing in the direction θ. Rs (θ, ω) is the phase difference information between the microphone units depending on the arrival direction of the sound wave, and indicates the direction information. The intensity u (θ, ω) indicates the intensity of the sound wave. Then, the spatial spectrum p (θ) can be derived by obtaining the intensity u (θ, ω) for the sound wave in each direction θ.

In (Equation 14), the observed correlation matrix Rx (ω) is an observable correlation matrix and is a known number. On the other hand, the intensities u (θ, ω) and the correlation matrix Rs (θ, ω) are unknown. Here, the correlation matrix Rs (θ, ω) is a correlation matrix for each direction θ, and the matrix element is the phase difference between the microphone units when the sound wave arrival direction is the direction θ. Focusing on this, the correlation matrix Rs (θ, ω) can be replaced with a theoretical value by using the microphone unit array, the direction θ, and the sound velocity c in the microphone array, which are known information. In the above-mentioned (Equation 4), (Equation 5) and (Equation 6), the correlation matrix Rs (θ, ω) is calculated in advance using known information, that is, the reference correlation matrix Rr (θ, θ, which is a theoretical value). It is replaced with ω).

Further, by assuming that the unknown number obtained as the spatial spectrum in the sound source exploration device 1 is the weighting coefficient a (θ, ω), that is, the weighting coefficient a (θ, ω) is equal to the intensity u (θ, ω) of (Equation 14). , (Equation 14) can be rewritten as (Equation 15).

Therefore, in calculating (Equation 15), since the observed correlation matrix Rx (ω) is the observed value and the reference correlation matrix Rr (θ, ω) is a known theoretical value, the weighting coefficient a (θ, θ, It becomes a problem to find ω). It should be noted that such a problem is also referred to as a semi-blind identification problem.

Here, the difference from the identification of a normal acoustic signal is that the observation correlation matrix Rx (ω) and the reference correlation matrix Rr (θ, ω) are matrices, and the weighting coefficient a (θ, ω) is a one-dimensional coefficient. The point is that the signal corresponding to the observation signal and the reference signal is a rotor representing a phase difference, and is always a complex number having an amplitude of 1.

Since the observed correlation matrix Rx (ω) and the reference correlation matrix Rr (θ, ω) are matrices and the weighting coefficient a (θ, ω) is a one-dimensional coefficient, the weighting coefficient a (θ, ω) is , It can be seen that the values that are common to each of the corresponding matrix elements of the observed correlation matrix Rx (ω) and the reference correlation matrix Rr (θ, ω) are obtained. That is, the weighting coefficients a (θ, ω) satisfying (Equation 16) obtained by rewriting (Equation 15) with the matrix elements are obtained. In (Equation 16), x _ij (ω) represents a matrix element of the observation correlation matrix Rx (ω). r _ij (θ, ω) indicates the matrix elements of the reference correlation matrix Rr (θ, ω).

In this embodiment, (Equation 16) is rewritten as (Equation 17), and a learning method such as LMS or ICA (Independent Component Analysis) that minimizes the error signal e (ω), which is an estimation error, is used. Find (θ, ω). The learning method is not limited to these.

More specifically, in (Equation 17), in order to calculate the weighting coefficient a (θ, ω) that holds in common for the matrix elements of _{x ij} (ω) and r _{ij (θ, ω), the selection unit 30} The matrix elements are sequentially selected and the weighting coefficient is learned. The signal multiplication units 61-1, ..., 61-N are the multiplication of the second term on the right side of (Equation 17), the signal addition unit 62 is Σ on the right side of (Equation 17), and the signal subtraction unit 71 is (. Corresponds to the subtraction of the right side of equation 17).

Further, since the signal corresponding to the observation signal and the reference signal is a rotor representing a phase difference and is always a complex number having an amplitude of 1, the error signal e (ω) is given non-linearity to cause mutual influence between directions. Add the effect of Independent Component Analysis (ICA) to suppress.

In the present embodiment, as shown in FIG. 6, after decomposing into a real part and an imaginary part, a non-linear function as described in (Equation 9) described above is applied. By doing so, it is possible to learn the difference in the direction θ as the sound wave arrival direction as an independent component, so that it is possible to obtain a convergence operation that is not easily affected by the different directions.

Based on the above idea, the weighting coefficient is updated using (Equation 11) and (Equation 12). Then, by using the weighting coefficient a (θ, ω) after learning and using (Equation 13), the spatial spectrum p (θ) which is the output of the sound source exploration device 1 can be calculated.

[effect]
As described above, according to the sound source exploration device 1 of the present embodiment, the spatial spectrum p (θ) is based on the observation correlation matrix Rx (ω) of the acoustic signal observed by the microphone unit which is a plurality of elements of the microphone array 200. Can be sought. More specifically, a reference correlation matrix Rr (θ, ω) for each direction that can be calculated as a theoretical value from the array array of the microphone array 200 is prepared in advance, and the reference correlation matrix Rr (θ, ω) indicating each direction is used. The weighting coefficient a (θ, ω) is calculated by learning so that the linear sum obtained by multiplying each weighting coefficient a (θ, ω) is equal to the observed correlation matrix Rx (ω). Then, the spatial spectrum p (θ) is calculated using the obtained weighting coefficients a (θ, ω). As a result, the intensity corresponding to the direction of the sound source that becomes the disturbing sound and the sound source to be searched is set as the weighting coefficient a (θ, ω) without performing the calculation of deriving the spatial spectrum from the correlation matrix and the direction vector having a large amount of calculation. Since it can be estimated sequentially, the spatial spectrum p (θ) should be obtained based on the observation correlation matrix Rx (ω) of the acoustic signal observed by the microphone unit at intervals of millisecond to second order in frequency analysis frame units. Can be done. That is, according to the present embodiment, it can be seen that the sound source exploration device 1 having excellent followability to changes in sound can be realized.

Further, according to the sound source exploration device 1 of the present embodiment, it is possible to calculate the intensity for each direction while canceling the influences between the directions. For example, the direction of theta ₁ through? Search range to be detected an angular range of _m, there is θ _{m + 1} ~θ _N angular range interference sound of a is the direction of the non-search range. Then, (Equation 15) is transformed as in (Equation 18) so that the exploration range to be detected is on the left side and the non-exploration range in which the disturbing sound exists is on the right side.

Then, it can be seen that the left side of (Equation 18) corresponds to the correlation matrix corresponding to the spatial spectrum obtained as the sound source search result. The first term on the right side of (Equation 18) corresponds to the observation correlation matrix in which sound waves in all directions are mixed, and the second term on the right side of (Equation 18) corresponds to the correlation matrix showing the disturbing sound component. I understand. Further, it can be seen that by subtracting the right side of (Equation 18), the correlation matrix of the disturbing sound component is subtracted from the observed correlation matrix Rx (ω), and a canceling effect is obtained. It can be seen that this is an operation of canceling interference with each other for each component of each direction θ, so that the noise resistance performance is improved. In addition, since the solutions are obtained for the weighting coefficients a (θ, ω) in all directions at the same time, it can be seen that the followability to changes in sound is also excellent.

Therefore, the sound source exploration device 1 of the present embodiment is excellent in noise resistance performance and followability to changes in sound by calculating the spatial spectrum p (θ) from the weighting coefficients a (θ, ω) in the exploration range. Sound source exploration can be realized.

As described above, according to the sound source exploration device 1 of the present embodiment, the direction of the sound source of the exploration target in the exploration target range can be searched more reliably. Further, the sound source exploration device 1 of the present embodiment has excellent noise resistance and followability to changes in sound by calculating the spatial spectrum p (θ) using the weighting coefficients a (θ, ω). Can be demonstrated.

Here, the effect of the sound source exploration device 1 in the present embodiment will be described with reference to FIGS. 9 and 10.

FIG. 9 is a spatial spectrum diagram in a comparative example. In FIG. 9, when the sound source S to be searched and the sound source N1 and the sound source N2 that are interfering sounds of the sound source S exist in the vicinity of the sound source S, the spatial spectrum is calculated by using the technique of Patent Document 1. The figure of is shown as a comparative example.

In the spatial spectrum shown in FIG. 9, the intensity of the sound source N1 which is an interfering sound appears to be attenuated not only in the direction in which the sound source N1 exists but also as the distance (angle) from the direction of the sound source N1. The intensity of the sound source N2, which is a disturbing sound, also appears with the same behavior as the sound source N1. Therefore, as shown in FIG. 9, when the sound pressure level of the sound source N1 and the sound source N2 is higher than the sound pressure level of the sound source S to be searched, the peak of the intensity of the sound source S is the two sound sources N1 which are interfering sounds. And it becomes a state of being buried in the peak of the intensity of the sound source N2. Therefore, in the comparative example, the existence of the sound source S to be searched, that is, the peak of the intensity of the sound source S cannot be detected, so that the direction of the sound source S cannot be searched.

On the other hand, FIG. 10 is a spatial spectrum diagram according to the first embodiment. Also in FIG. 10, when the sound source S to be searched and the sound source N1 and the sound source N2 which are the interfering sounds of the sound source S exist in the vicinity of the sound source S, the sound source search device 1 of the present embodiment obtains a spatial spectrum. The figure when calculated is shown. Since the sound source exploration device 1 calculates the spatial spectrum p (θ) using the weighting coefficients a (θ, ω), it is possible to cancel the interference with each other for each component of each direction θ. Therefore, as shown in FIG. 10, regardless of whether the sound pressure levels of the sound source N1 and the sound source N2 are higher or lower than the sound pressure level of the sound source S to be searched, the intensity peak of the sound source S and the disturbing sound are obtained. The two sound source N1 and the intensity peaks of the sound source N2 appear independently. That is, the peaks of the intensities of the sound source N1 and the sound source N2 which are the disturbing sounds and the sound source S to be searched can be searched independently at the same time.

Therefore, according to the sound source exploration device 1 of the present embodiment, it can be seen that the direction of the sound source of the exploration target in the exploration target range can be searched more reliably.

The observation correlation matrix Rx (ω) calculated by the correlation matrix calculation unit 10 and the reference correlation matrix Rr (θ, ω) for each search direction θ stored in the storage unit 20 are the upper triangular of the correlation matrix used in the calculation. The elements of the matrix or arbitrarily selected elements may be realized in the form of a vector. In that case, the selection unit 30 may sequentially select and output the vector elements.

Further, in the present embodiment, it has been described that the number of directions N is 180 in the reference correlation matrix Rr (θ, ω) and the weighting coefficient a (θ, ω) for each direction θ, but the present invention is not limited to this. .. The number of directions N may be large or small depending on the application of the sound source exploration device 1, the scale of the microphone array, or the calculation scale, and is not particularly limited. Further, the angle interval to be set may be uniform or may have a bias.

Further, in the present embodiment, the range of the frequency ω is not particularly limited in the observation correlation matrix Rx (ω) for each frequency ω, the reference correlation matrix Rr (θ, ω), and the weighting coefficient a (θ, ω). However, the range of the frequency ω may be limited according to the frequency component included in the sound source to be searched.

(Embodiment 2)
In the first embodiment, the case where the spatial spectrum p (θ) is calculated using the learned weighting coefficients a (θ, ω) has been described, but the present invention is not limited to this. The learned weighting coefficient a (θ, ω) may be used to calculate the acoustic signal waveform arriving from the specified direction. Hereinafter, this case will be described as the second embodiment.

FIG. 11 is a diagram showing an example of the configuration of the sound source exploration system 1000A according to the second embodiment. The sound source exploration system 1000A corresponds to a microphone device using the sound source exploration device. The same elements as those in FIGS. 1 and 4 are designated by the same reference numerals, and detailed description thereof will be omitted.

The sound source exploration system 1000A shown in FIG. 11 has different configurations of the acoustic signal spectrum calculation unit 100A, the output unit 110A, and the IFFT 120 from the sound source exploration system 1000 according to the first embodiment.

[Acoustic signal spectrum calculation unit 100A]
The acoustic signal spectrum calculation unit 100A obtains the weight coefficient a (θ, ω) held by the holding unit 50, the frequency spectrum signal Sm1 (ω) with respect to the acoustic signal m1 (n) from the microphone unit 201, and the signal acquisition. The output acoustic signal spectrum Y (ω) is calculated _{with the direction θ 0} , which is the direction specified in 1., as the input.

More specifically, the acoustic signal spectrum calculation unit 100A calculates the acoustic signal spectrum Y (ω) using (Equation 19).

From the viewpoint of the angular resolution of sound source exploration, depending on the size of the microphone array 200 or the number of microphone units _{, the weighting coefficients of adjacent angles in the designated direction θ 0} may be added and used as in (Equation 20). ..

Since the weighting coefficients a (θ, ω) in (Equation 19) and (Equation 20) represent the intensities for sound waves in each direction θ as described in the principle of the above operation, the spectrum in the θ direction with respect to the spectrum in all directions. Represents the strength ratio of. Therefore, by multiplying the frequency spectrum Sm1 (ω) in all directions, the acoustic signal spectrum Y (ω) for the sound wave arriving from the _{designated direction θ 0 can be calculated.}

[IFFT120]
The IFFT (Inverse Fast Fourier Transform) 120 calculates an acoustic signal waveform y (n) obtained by performing a fast inverse Fourier transform on the acoustic signal spectrum Y (ω) calculated by the acoustic signal spectrum calculation unit 100A, and outputs the acoustic signal waveform y (n) to the output unit 110A. do.

[effect]
As described above, according to the sound source exploration system 1000A of the present embodiment, only the specified specific direction is used by using the weighting coefficient a (θ, ω) calculated by learning in the sound source exploration device having excellent noise resistance. The acoustic signal waveform y (n) can be output. This makes it possible to realize the function of a microphone device that extracts sound only in a specific direction.

The sound source exploration device and the like according to one or more aspects of the present disclosure have been described above based on the embodiments and modifications, but the present disclosure is not limited to these embodiments and the like. As long as it does not deviate from the gist of the present disclosure, one or more of the present embodiments may be modified by those skilled in the art, or may be constructed by combining components in different embodiments. It may be included within the scope of the embodiment. For example, the following cases are also included in the present disclosure.

(1) Specifically, the above-mentioned sound source search device or the like may be a computer system composed of a microprocessor, ROM, RAM, hard disk unit, display unit, keyboard, mouse and the like. A computer program is stored in the RAM or the hard disk unit. As the microprocessor operates according to the computer program, each component achieves its function. Here, a computer program is configured by combining a plurality of instruction codes indicating commands to a computer in order to achieve a predetermined function.

(2) A part or all of the components constituting the sound source search device or the like may be composed of one system LSI (Large Scale Integration). A system LSI is an ultra-multifunctional LSI manufactured by integrating a plurality of components on a single chip, and specifically, is a computer system including a microprocessor, a ROM, a RAM, and the like. .. A computer program is stored in the RAM. When the microprocessor operates according to the computer program, the system LSI achieves its function.

(3) A part or all of the components constituting the sound source exploration device or the like may be composed of an IC card or a single module that can be attached to and detached from each device. The IC card or the module is a computer system composed of a microprocessor, a ROM, a RAM, and the like. The IC card or the module may include the above-mentioned super multifunctional LSI. When the microprocessor operates according to a computer program, the IC card or the module achieves its function. This IC card or this module may have tamper resistance.

The present disclosure can be used for a sound source exploration device using a plurality of microphone units, and in particular, the sound reaching the microphone unit such as a radio-controlled helicopter or a drone located relatively far from the sound source exploration device is smaller than the ambient sound. It can be used for a sound source exploration device that can more reliably explore the direction of a sound source.

1 Sound source search device 10 Correlation matrix calculation unit 20 Storage unit 30 Selection unit 31, 32-1, 32-m, 32-N Matrix element selection unit 40 Learning unit 50 Holding unit 60 Linear sum calculation unit 61-1, 61-N Signal multiplication unit 62, 806 Signal addition unit 70 Error calculation unit 71 Signal subtraction unit 80 Non-linear function unit 90 Weight update unit 100 Spatial spectrum calculation unit 100A Acoustic signal spectrum calculation unit 110, 110A Output unit 120 IFFT
200 Microphone Array 201, 202, 203 Microphone Unit 300 Frequency Analyzer 301, 302, 303 FFT
801 Real part extraction part 802 Imaginary part extraction part 803, 804 Non-linearity addition part 805 Imaginary unit multiplication part 1000, 1000A Sound source search system

Claims (12)

A sound source exploration device that explores the direction of the sound source to be explored.
A correlation matrix calculation unit that calculates a first correlation matrix that is a correlation matrix of observation signals that are acoustic signals picked up by a microphone array composed of two or more microphone units arranged apart from each other.
A linear sum obtained by multiplying each of a plurality of second correlation matrices stored in advance in the storage unit, which are direction-specific correlation matrices calculated from the array array of the microphone array, by a weight. A learning unit that calculates the weights by learning so that they are equal to the first correlation matrix.
A spatial spectrum calculation unit for calculating the spatial spectrum of the observation signal, which is a spatial spectrum showing the sound pressure intensity for each direction by using the weight calculated by the learning unit, is provided.
The sound source exploration device further
The first element, which is one of the elements constituting the first correlation matrix, and the element constituting each of the plurality of second correlation matrices, which is an element at a position corresponding to the first element. A selection unit that selects two elements and sequentially switches between the first element and the second element to be selected is provided.
The learning unit updates and updates the first weight to the second weight calculated by the learning so that the linear sum of the first elements obtained by multiplying the second element by the first weight becomes equal to the first element. The second element linear sum obtained by multiplying the second element selected by the selection unit by the second weight is equal to the first element selected by the selection unit. By sequentially repeating updating the second weight to the third weight calculated by the learning, the weight is calculated by the learning.
Sound source exploration equipment. The selection unit is a plurality of elements of one set of two sets of elements separated by the diagonal components among the elements excluding the diagonal components constituting the first correlation matrix and the second correlation matrix. Select the first element and the second element only from among
The sound source exploration device according to claim 1. A sound source exploration device that explores the direction of the sound source to be explored.
A correlation matrix calculation unit that calculates a first correlation matrix that is a correlation matrix of observation signals that are acoustic signals picked up by a microphone array composed of two or more microphone units arranged apart from each other.
A linear sum obtained by multiplying each of a plurality of second correlation matrices stored in advance in the storage unit, which are direction-specific correlation matrices calculated from the array array of the microphone array, by a weight. A learning unit that calculates the weights by learning so that they are equal to the first correlation matrix.
A spatial spectrum calculation unit for calculating the spatial spectrum of the observation signal, which is a spatial spectrum showing the sound pressure intensity for each direction by using the weight calculated by the learning unit, is provided.
The learning unit
By using the LMS (Least Mean Square) algorithm or ICA (Independent Component Analysis), the weight is calculated from the error which is the difference between the linear sum and the first correlation matrix and the second correlation matrix.
Sound source exploration equipment. A sound source exploration device that explores the direction of the sound source to be explored.
A correlation matrix calculation unit that calculates a first correlation matrix that is a correlation matrix of observation signals that are acoustic signals picked up by a microphone array composed of two or more microphone units arranged apart from each other.
A linear sum obtained by multiplying each of a plurality of second correlation matrices stored in advance in the storage unit, which are direction-specific correlation matrices calculated from the array array of the microphone array, by a weight. A learning unit that calculates the weights by learning so that they are equal to the first correlation matrix.
A spatial spectrum calculation unit for calculating the spatial spectrum of the observation signal, which is a spatial spectrum showing the sound pressure intensity for each direction by using the weight calculated by the learning unit, is provided.
The learning unit
A holding part that holds the weight and
A linear sum calculation unit that calculates a linear sum obtained by multiplying each of the plurality of second correlation matrices by the weights held by the holding unit.
An error calculation unit that calculates an error that is the difference between the linear sum and the first correlation matrix, and
The weight update amount is calculated from the product of the error and the second correlation matrix, and the weight update amount is added to the weight held by the holding unit to obtain the weight held by the holding unit.
Sound source exploration equipment. The weight update unit
By using the LMS algorithm or ICA, the weight update amount is calculated from the error and the second correlation matrix.
The sound source exploration device according to claim 4. The learning unit further
A non-linear function unit that adds non-linearity to the error by using a predetermined non-linear function is provided.
The weight update unit calculates the weight update amount from the error to which the non-linearity is added by the non-linear function unit and the second correlation matrix, and adds the weight update amount to the weight held by the holding unit. Is the weight held by the holding unit.
The sound source exploration apparatus according to claim 4 or 5. A sound source exploration method that explores the direction of the sound source to be explored.
A correlation matrix calculation step for calculating a first correlation matrix, which is a correlation matrix of observation signals, which are acoustic signals picked up by a microphone array composed of two or more microphone units arranged apart from each other.
A linear sum obtained by multiplying each of a plurality of second correlation matrices stored in advance in the storage unit, which are direction-specific correlation matrices calculated from the array array of the microphone array, by a weight. A learning step in which the weights are calculated by learning so as to be equal to the first correlation matrix.
Using the weight calculated in the learning step, a spatial spectrum calculation step of calculating the spatial spectrum of the observation signal, which is a spatial spectrum showing the sound pressure intensity for each direction ,
The first element, which is one of the elements constituting the first correlation matrix, and the element constituting each of the plurality of second correlation matrices, which is an element at a position corresponding to the first element. Includes a selection step of selecting two elements and sequentially switching between the first element and the second element to be selected.
In the learning step, the first weight is updated to the second weight calculated by the learning so that the linear sum of the first elements obtained by multiplying the second element by the first weight becomes equal to the first element. The second element linear sum obtained by multiplying the second element selected in the selection step by the second weight is equal to the first element selected in the selection step. The weight is calculated by the learning by sequentially repeating updating the second weight to the third weight calculated by the learning.
Sound source exploration method. It is a program for making a computer execute a sound source search method for searching the direction of a sound source to be searched.
A correlation matrix calculation step for calculating a first correlation matrix, which is a correlation matrix of observation signals, which are acoustic signals picked up by a microphone array composed of two or more microphone units arranged apart from each other.
A linear sum obtained by multiplying each of a plurality of second correlation matrices stored in advance in the storage unit, which are direction-specific correlation matrices calculated from the array array of the microphone array, by a weight. A learning step in which the weights are calculated by learning so as to be equal to the first correlation matrix.
Using the weight calculated in the learning step, a spatial spectrum calculation step of calculating the spatial spectrum of the observation signal, which is a spatial spectrum showing the sound pressure intensity for each direction,
The first element, which is one of the elements constituting the first correlation matrix, and the element constituting each of the plurality of second correlation matrices, which is an element at a position corresponding to the first element. A computer is made to select two elements and execute a selection step of sequentially switching between the first element to be selected and the second element to be selected.
In the learning step, the first weight is updated to the second weight calculated by the learning so that the linear sum of the first elements obtained by multiplying the second element by the first weight becomes equal to the first element. The second element linear sum obtained by multiplying the second element selected in the selection step by the second weight is equal to the first element selected in the selection step. The weight is calculated by the learning by sequentially repeating updating the second weight to the third weight calculated by the learning.
program. A sound source exploration method that explores the direction of the sound source to be explored.
A correlation matrix calculation step for calculating a first correlation matrix, which is a correlation matrix of observation signals, which are acoustic signals picked up by a microphone array composed of two or more microphone units arranged apart from each other.
A linear sum obtained by multiplying each of a plurality of second correlation matrices stored in advance in the storage unit, which are direction-specific correlation matrices calculated from the array array of the microphone array, by a weight. A learning step in which the weights are calculated by learning so as to be equal to the first correlation matrix.
The spatial spectrum calculation step of calculating the spatial spectrum of the observation signal, which is a spatial spectrum showing the sound pressure intensity for each direction by using the weight calculated in the learning step, is included.
In the learning step
By using the LMS (Least Mean Square) algorithm or ICA (Independent Component Analysis), the weight is calculated from the error which is the difference between the linear sum and the first correlation matrix and the second correlation matrix.
Sound source exploration method . A sound source exploration method that explores the direction of the sound source to be explored.
A correlation matrix calculation step for calculating a first correlation matrix, which is a correlation matrix of observation signals, which are acoustic signals picked up by a microphone array composed of two or more microphone units arranged apart from each other.
A linear sum obtained by multiplying each of a plurality of second correlation matrices stored in advance in the storage unit, which are direction-specific correlation matrices calculated from the array array of the microphone array, by a weight. A learning step in which the weights are calculated by learning so as to be equal to the first correlation matrix.
The spatial spectrum calculation step of calculating the spatial spectrum of the observation signal, which is a spatial spectrum showing the sound pressure intensity for each direction by using the weight calculated in the learning step, is included.
In the learning step
A retention step that retains weights and
A linear sum calculation step for calculating a linear sum obtained by multiplying each of the plurality of second correlation matrices by the weights held in the holding step.
An error calculation step for calculating an error that is the difference between the linear sum and the first correlation matrix, and
The weight update amount is calculated from the product of the error and the second correlation matrix, and the weight update amount is added to the weight held in the holding step to obtain the weight held in the holding step. include,
Sound source exploration method. It is a program for making a computer execute a sound source search method for searching the direction of a sound source to be searched.
A correlation matrix calculation step for calculating a first correlation matrix, which is a correlation matrix of observation signals, which are acoustic signals picked up by a microphone array composed of two or more microphone units arranged apart from each other.
A linear sum obtained by multiplying each of a plurality of second correlation matrices stored in advance in the storage unit, which are direction-specific correlation matrices calculated from the array array of the microphone array, by a weight. A learning step in which the weights are calculated by learning so as to be equal to the first correlation matrix.
Using the weight calculated in the learning step, a computer is made to execute a spatial spectrum calculation step of calculating the spatial spectrum of the observed signal, which is a spatial spectrum indicating the sound pressure intensity for each direction.
In the learning step
By using the LMS (Least Mean Square) algorithm or ICA (Independent Component Analysis), the weight is calculated from the error which is the difference between the linear sum and the first correlation matrix and the second correlation matrix.
program. It is a program for making a computer execute a sound source search method for searching the direction of a sound source to be searched.
A correlation matrix calculation step for calculating a first correlation matrix, which is a correlation matrix of observation signals, which are acoustic signals picked up by a microphone array composed of two or more microphone units arranged apart from each other.
A linear sum obtained by multiplying each of a plurality of second correlation matrices stored in advance in the storage unit, which are direction-specific correlation matrices calculated from the array array of the microphone array, by a weight. A learning step in which the weights are calculated by learning so as to be equal to the first correlation matrix.
Using the weight calculated in the learning step, a computer is made to execute a spatial spectrum calculation step of calculating the spatial spectrum of the observed signal, which is a spatial spectrum indicating the sound pressure intensity for each direction.
In the learning step
A retention step that retains weights and
A linear sum calculation step for calculating a linear sum obtained by multiplying each of the plurality of second correlation matrices by the weights held in the holding step.
An error calculation step for calculating an error that is the difference between the linear sum and the first correlation matrix, and
The weight update amount is calculated from the product of the error and the second correlation matrix, and the weight update amount is added to the weight held in the holding step to obtain the weight held in the holding step. Let the computer do it,
program.

Images