JP4724054B2 - Specific direction sound collection device, specific direction sound collection program, recording medium - Google Patents

Description

  The present invention relates to a sound collection device that acquires a voice in a hands-free manner such as a voice call or operation of a device, and particularly emphasizes only sound from a sound source existing in a specific direction when viewed from the sound collection device. The present invention relates to a specific direction sound pickup device that is suitable for use when desired.

In the conventional technique, as shown in FIG. 18, microphones mic. Arranged at _M different positions (p ₁ , q ₁ ) to (p _M , q _M ) on the xy plane are arranged. 1-mic. When using M as a sound generated from a sound source in the direction of an arbitrary angle θ _S as a signal and a sound generated in other directions as noise, only the signal is emphasized and a high SNR (signal noise) Ratio). FIG. 19 is a block diagram showing the configuration of a conventional enhanced sound collection method. The signal x _m (n) (m = 1... M) received by the microphone m arranged at the position (x _m , y _m ) is added by adding a delay D _m as shown in the equation (1). Obtain y _m (n).

y _m (n) = x _m (n−D _m ) (1)
At this time, the delay amount D _m can be derived from the desired sound source direction θ _S given in advance by the equation (2).
D _m = (d _m / c) sin θ _S (2)
Where c is the speed of sound, d _m is when viewed from the sound wave arriving from theta _S direction in FIG. 18, the distance between the microphone m and a reference point o, represented by the formula (3).
d _m = p _m sin θ + q _m cos θ (3)
Next, y _m (n) obtained now is added as shown in Expression (4) to obtain a signal z (n) that emphasizes the sound emitted from the desired position.

以上が従来の強調収音法（非特許文献１）である。この従来技術で形成される指向特性には図２０に示すように、主ビームＢＭに近接して比較的大きい利得を持つサイドローブＳＢとよばれる領域が生じるため、雑音を十分に抑圧することができない。またこのサイドローブＳＢの利得を極力小さく抑えるためには、マイクロホン数を増やし、またマイクロホンアレーを大型にする必要がある（非特許文献１）。
大賀寿郎、山崎芳男、金田豊共著、音響システムとディジタル処理、電子情報通信学会 Ｐ．１８１〜Ｐ．１８６ ７．１．２．

The above is the conventional enhanced sound collection method (Non-Patent Document 1). As shown in FIG. 20, the directivity formed by this conventional technique has a region called a side lobe SB having a relatively large gain close to the main beam BM, so that noise can be sufficiently suppressed. Can not. In order to keep the gain of the side lobe SB as small as possible, it is necessary to increase the number of microphones and increase the size of the microphone array (Non-Patent Document 1).
Toshiro Oga, Yoshio Yamazaki, Yutaka Kaneda, Acoustic systems and digital processing, IEICE 181-P. 186 7.1.2.

When the sound collecting device is directed to a specific direction using conventional technology, the sound emitted in that direction is emphasized, and the sound emitted in other directions is suppressed and collected. Since the formed directional characteristic has side lobes, there is a problem in that sound emitted from the direction in which it is desired to be suppressed is collected without being sufficiently suppressed.
For this reason, when there is a noise source that emits a very loud sound other than the direction of the sound source to be emphasized, the sound collecting device of the prior art cannot obtain a sufficient enhancement effect for the desired sound source.
Further, in the prior art, in order to reduce the side lobes, the number of microphones must be increased and the microphone array must be increased in size, which is difficult to install and transport in practical use. Furthermore, since the directivity characteristics of the sound collecting device according to the prior art change depending on the frequency, there is a problem that a sufficient emphasis effect cannot be obtained depending on the frequency structure of the desired sound or noise.

  The present invention has been made to solve the above-described problems, and realizes an apparatus for enhancing and collecting sound from a desired sound source with an SNR higher than that of the prior art without increasing the scale of a microphone array. There is.

In the present invention, as shown in FIG. 3, the sound collecting area is divided into angular areas Θ _{1 to} Θ _Q in a plurality of directions, and the directivity of the microphone array is controlled so as to be directed to the angular areas in the respective directions. Use the sound signal. At this time, the power of the signal processed by the microphone array changes in accordance with the direction in which the sound source exists, as compared to before the processing. In the present invention, the amount of change in power is used to estimate the power of a signal arriving from each direction area. Then, from the estimated power, a gain coefficient for emphasizing a signal coming from an angle region in a given direction is calculated, and a signal obtained by multiplying the signal arriving from a specific direction by the gain coefficient is finally output. Get as a signal.

As a specific configuration of the sound collecting device in a specific direction according to the present invention, the output signals of a microphone array configured by mounting a plurality of microphones are used to emphasize and collect sounds coming from angular regions in different directions. A plurality of beamformer units for sound, a frequency domain conversion unit for converting each of the angle domain signals collected by the plurality of beamformer units into a frequency domain signal divided into a plurality of band components, and each frequency domain conversion unit A specific direction selection unit that selects a specific direction frequency domain signal belonging to an angular domain signal of an angular domain in a desired direction from among the frequency domain signals output by the signal, and a sum of signals included in each angular domain is a signal of the angular domain Each frequency domain with respect to the inverse matrix of the gain matrix whose elements are parameters obtained from the directivity characteristics for each angular domain of multiple beamformer units The signal amount estimation unit that estimates the signal amount of each angle region by multiplying the signal amount of the frequency region signal output by the conversion unit, the signal amount in the angle region of the desired direction, and the sum of the signal amount in all angle regions A gain coefficient calculation unit that calculates a gain coefficient for each frequency band according to the ratio, and a multiplication unit that multiplies the signal amount of each corresponding frequency band of the specific direction frequency domain signal by the gain coefficient calculated by the gain coefficient calculation unit. It is characterized by that.

The specific direction sound collecting apparatus according to the present invention further uses a plurality of microphones configured to mount a plurality of microphones to emphasize and collect a plurality of sounds that are collected from angle regions in different directions. A beamformer unit, a frequency domain conversion unit that converts each of the angle domain signals collected by a plurality of beamformer units into a frequency domain signal divided into a plurality of band components, and a frequency domain output by the frequency domain conversion unit Provided on the output side of the same band component collection unit, the band division unit that divides the signal into frequency band components, the same band component collection unit that collects the band components output by the band division unit for each same band component, and the same A band synthesizing unit that synthesizes the band components output from the band component collecting unit, and the same band component collecting unit is an angle in a desired direction in which any of the beam former units collects sound. A specific direction selection unit for selecting a specific direction band component belonging to the area signal, the sum of the signals included in each angular region as the signal of the angle regions, for each angle and frequency bands of a plurality of beamformer unit A signal amount estimation unit that estimates the signal amount of each angle region by multiplying the inverse matrix of the gain matrix whose elements are parameters obtained from directivity characteristics by the signal amount of the frequency band component output by each band dividing unit ; A gain coefficient calculation unit that calculates a gain coefficient based on a ratio of a signal amount in a direction angle region and a sum of signal amounts in all angle regions, and the gain coefficient calculated by the gain factor calculation unit as the specific direction band component And a multiplication unit for multiplying.

Furthermore, in the specific direction sound pickup device according to the present invention, the signal amount of the frequency domain signal is a signal power value, or the signal amount of the frequency band component is a signal power value. characterized in that there.
Further, the specific direction sound pickup device according to the present invention is characterized in that, in the above-described specific direction sound pickup device, the signal amount of the frequency domain signal is an absolute value of the signal, or the signal amount of the frequency band component is The absolute value of the signal .
Moreover, the point which is characterized in a particular direction and collection device according to the present invention, in the specific direction and collecting apparatus described above, the gain factor calculating characteristics of the gain coefficient calculation unit for the signal amount in the desired direction of angular regions, other When the signal amount in the angle region is negligibly small, the gain coefficient value is given as a predetermined maximum value, and the signal amount in the angle region in the desired direction is specified relative to the signal amount in other angle regions . When the amount of the area signal is a negligible value, the gain coefficient value is calculated with a characteristic that is almost close to zero.

Further, the specific direction sound pickup device according to the present invention is characterized in that, in the above-described specific direction sound pickup device, the gain coefficient calculation characteristic of the gain coefficient calculation unit is different from the signal amount in the angle region of the desired direction . the maximum value or the value of the signal amount is small bandwidth the gain coefficient of the angular region is kept at a value close to the maximum value, against the amount of signal in the desired direction of angular region, the gain in the region signal amount is large and other angular regions The coefficient is calculated with a characteristic that maintains the value of the coefficient at a value close to 0 to nearly 0.

According to the specific direction sound collecting apparatus of the present invention, in order to improve the enhancement effect when sound is collected by enhancing the sound emitted from a sound source in a desired direction, a plurality of beam forks are used using signals received by the microphone array. The power of the sound signal emitted from each sound source is estimated from the result of the image processing, and the desired sound signal is emphasized by calculating a gain coefficient corresponding to the SN ratio. For this reason, in order to obtain the same effect with the prior art, it was necessary to increase the number of microphones and increase the size of the microphone array. It has the effect of improving the emphasis effect with an easy small system.
Further, in the specific direction sound pickup device according to the present invention, the power value or absolute value of the signal received by the microphone array is estimated for each frequency component and the enhancement process is performed. Therefore, regardless of the frequency structure of the desired sound or noise. A high emphasis effect can be obtained.

When implementing the specific direction sound pickup apparatus according to the present invention, it is possible to configure all by hardware, but in the simplest implementation, the specific direction sound pickup apparatus according to the present invention is installed in a computer, and the computer The best mode is to let the device function as a specific direction sound pickup device.
In order for a computer to function as a specific direction sound pickup device according to the present invention, sound that arrives from angular regions in different directions by using a microphone array output signal in the computer by a specific direction sound pickup program installed in the computer. A plurality of beamformer sections for emphasizing and collecting sound, a frequency domain conversion section for converting each of the angle domain signals collected by the plurality of beamformer sections into frequency domain signals divided into a plurality of band components, and A specific direction selection unit that selects a specific direction frequency domain signal belonging to an angular domain signal of an angular domain in a desired direction from among the frequency domain signals output by the frequency domain conversion unit, a signal amount of the specific direction frequency domain signal, and the like A signal amount estimation unit for estimating the sum of frequency domain signals converted from frequency domain signals into angle domain signals in the direction of A signal amount estimation unit that estimates a signal amount of a specific direction frequency domain signal, a total amount of frequency domain signals converted from a frequency domain signal from an angle domain signal in another direction, and a signal amount of a specific direction frequency domain signal, A gain coefficient calculation unit that calculates a gain coefficient for each frequency domain based on a ratio with the total amount of other frequency domain signals including the signal quantity of the specific direction frequency domain signal, and the gain coefficient calculated by the gain coefficient calculation unit in a specific direction A multiplication unit that multiplies the signal amount of each corresponding frequency band of the region signal is constructed and functions as a specific direction sound collecting device.

First, an overall outline of the present invention will be described. FIG. 1 shows the overall configuration of a sound collecting device in a specific direction according to the present invention. Signals x _m (n) (m = 1, 2,..., M) received by the microphone array 11 composed of M (≧ 2) microphones are transmitted from the first beam former unit 12-1. The signals are input to Q beam former units 12-1 to 12-Q up to the Q beam former unit 12-Q. Here, n represents the sample number of the discrete time signal.
In the beam former units 12-1 to 12-Q, for example, a directional beam BM as shown in FIG. 2 is directed to any one of the _Q direction regions Θ _{1 to} Θ Q given in advance in FIG. A process of collecting sound by emphasizing the sound emitted in the corresponding direction area is performed, and the result is output. The output signals y ₁ (n), y ₂ (n),..., Y _Q (n) of the beam former units 12-1 to 12-Q are respectively input to the frequency domain transform units 13-1 to 13-Q. The output signals Y ₁ (ω, l), Y ₂ (ω, l),... Y _Q (ω, l) of the frequency domain transforming units 13-1 to 13-Q are the signal amount estimating unit 14 and the specific direction selecting unit. 15 respectively.

The signal amount estimation unit 14 obtains a power component of the sum of sound signals emitted from the sound sources in the respective direction regions Θ _{1 to} Θ _Q from the output signal power of the input beam former units 12-1 to 12 _-Q . The signal power vector X _est * (ω, l) in which the symbols are combined into one vector (in the following sentence, the English character with * represents a vector, and the bold character in the formula represents a vector).
The specific direction selection unit 15 selects the output of the beam former unit that directs the directional beam to the direction region to be emphasized, and outputs it as Y _S (ω, l).

The gain coefficient calculation unit 16 calculates the gain coefficient R (ω, l) from the input signal power vector X _est * (ω, l) and outputs it. The gain coefficient R (ω, l) is input to the multiplication unit 17. The multiplier 17 outputs the result of multiplying the input gain coefficient R (ω, l) and the output Y _S (ω, l) of the specific direction selector 15 for each component of the same frequency. The output signal Y _SR (ω, l) of the multiplication unit 17 is input to the inverse frequency domain transform unit 18, and a signal y (n) restored to a time signal by performing inverse discrete Fourier transform is output. This signal y (n) is a signal picked up by enhancing the desired sound by the apparatus of the present invention.
The details of the beam former units 12-1 to 12-Q, the signal amount estimation unit 14, the gain coefficient calculation unit 16, and the specific direction selection unit 15 will be described in order below with reference to another drawing.

各フィルタ処理部ＦＣ１〜ＦＣＭの出力信号は加算部ＡＤＤに入力される。加算部ＡＤＤでは入力信号を式（６）のように加算し、ビームフォーマー部の出力信号ｙ_ｑ（ｎ）（ｑ＝１…Ｑ）を得る。

ここでフィルタ係数Ｗ_ｑｍ（ｎ）は、それぞれのビームフォーマー部１２−１〜１２−Ｑの指向特性Ｄ_ｑ（ω，θ）が、図３に示すあらかじめ与えられた第Ｑ方向領域Θ_Ｑで発せられる音を強調して受音し、それ以外の方向で発せられる音を抑圧するように構成される。 (Beam former part)
FIG. 4 shows the configuration of one of the beam formers 12-1 to 12-Q, for example, the beam former 12-q. Similar processing is performed in all beam former units. The input signal x _m (n) (m = 1, 2,..., M) is input to the filter processing units FC1 to FCM. The filter processing units FC1 to FCM output a signal x ′ _qm (n) obtained by substituting a filter coefficient W _qm (n) given in advance (determination method will be described later) into the convolution operation shown in Expression (5). To do.

The output signals of the filter processing units FC1 to FCM are input to the adding unit ADD. The adder ADD adds the input signals as shown in Expression (6) to obtain the output signal y _q (n) (q = 1... Q) of the beam former.

Here, the filter coefficient W _qm (n) indicates that the directivity characteristics D _q (ω, θ) of the respective beam former units 12-1 to 12-Q are given in the Q-direction region Θ _Q given in advance shown in FIG. It is configured to emphasize and receive the sound emitted from and to suppress the sound emitted in other directions.

ビームフォーマー部出力パワーベクトルＹ※（ω，ｌ）は乗算部１４Ｂに入力される。乗算部１４Ｂのもう一方の入力であるパワー推定行列Ｔ^−１※は、逆行列演算部１４Ｃの出力信号である。逆行列演算部１４Ｃには式（８）により定義されるゲイン行列Ｔ※が入力され、その逆行列Ｔ^−１※を出力する。 (Signal amount estimation unit)
FIG. 5 shows the configuration of the signal amount estimation unit 14. The frequency components Y ₁ (ω, l), Y ₂ (ω, l),..., Y _Q (ω, l) input to the signal amount estimation unit 14 are input to the power calculation units PW-1 to PW-Q, respectively. The signal power values | Y ₁ (ω, l) | ² , | Y ₂ (ω, l) | ² ,..., | Y _Q (ω, l) | ² are output and input to the vectorization unit 14A. Is done. The vectorization unit 14A outputs a beamformer unit output power vector Y * (ω, l) in which input power values are collected in a vector format as shown in Expression (7).

The beamformer unit output power vector Y * (ω, l) is input to the multiplication unit 14B. The power estimation matrix T ⁻¹ *, which is the other input of the multiplier 14B, is an output signal of the inverse matrix calculator 14C. A gain matrix T * defined by Expression (8) is input to the inverse matrix calculation unit 14C, and the inverse matrix T ⁻¹ * is output.

ゲイン行列Ｔ※の各要素は、図６に示すように各ビームフォーマー部１２−１〜１２−Ｑの各方向領域に対する指向特性から求められるパラメータであり、例えば式（９）に示すような指向特性の周波数および方向に関する平均値を用いる。なお指向特性は、例えば前述の非特許文献１に記載されている周知の技術を用いてフィルタ係数Ｗ_ｍ（ｎ）より求めることができる。

α_ｐｑは第ｐビームフォーマー部の第ｑ方向領域に対する指向特性の平均値である。乗算部１４Ｂは式（１６）に示すように入力されたビームフォーマー部出力パワーベクトルとパワー推定行列の乗算を周波数成分ごとに行い、推定信号パワーベクトルＸ_ｅｓｔ※（ω，ｌ）を出力する。

Each element of the gain matrix T * is a parameter obtained from directivity characteristics with respect to each direction region of each beamformer unit 12-1 to 12-Q as shown in FIG. 6, for example, as shown in Expression (9). The average value for the frequency and direction of the directivity is used. The directivity can be obtained from the filter coefficient W _m (n) using a known technique described in Non-Patent Document 1, for example.

α _pq is an average value of directivity with respect to the q-th direction region of the p-th beam former part. The multiplication unit 14B performs multiplication of the input beamformer unit output power vector and the power estimation matrix for each frequency component as shown in Expression (16), and outputs an estimated signal power vector X _est * (ω, l). .

(Specific direction selector)
FIG. 7 shows the configuration of the specific direction selection unit 15. In the specific direction selection unit 15, the frequency components Y ₁ (ω, l) input from the beam former units 12-1 to 12 -Q through the frequency domain conversion units 13-1 to 13 -Q to Y _Q (ω, Among the l), the one corresponding to the qs direction region to be emphasized is selected and output as Y _S (ω, l).
Y _S (ω, l) = Y _qs (ω, l) (11)

ＳＮ比推定部１６Ｂでは式（１３）を用いて所望方向領域の信号を強調する利得係数Ｒ（ω，ｌ）を計算し出力する。

利得係数Ｒ（ω，ｌ）は図９に示すように、特定方向信号｜Ｓ_ｑｓ（ω，ｌ）｜^２と角度領域の信号の総和Σ_ｑ｜Ｓ_ｑ（ω，ｌ）｜^２の関係が｜Ｓ_ｑｓ（ω，ｌ）｜^２≒Σ_ｑ｜Ｓ_ｑ（ω，ｌ）｜^２であった場合は利得係数Ｒ（ω，ｌ）は最大値である例えば「１」が与えられ、｜Ｓ_ｑｓ（ω，ｌ）｜^２＜＜Σ_ｑ｜Ｓ_ｑ（ω，ｌ）｜^２であった場合は利得係数Ｒ（ω，ｌ）は≒０を与える。 (Gain coefficient calculator)
FIG. 8 shows the flow of processing in the gain coefficient calculation unit 16. The estimated signal power vector X _est * (ω, l) input from the signal amount estimation unit 14 is input to the vector element extraction unit 16A. In the vector element extraction unit 16A, as shown in Expression (12), the first component of the input estimated signal power vector is the first direction area signal estimated power | S ₁ (ω, l) | ² , and the second component is the second component. The q-th component is output as the signal estimation power from the q-th direction area as in the two-direction area signal estimation power | S ₂ (ω, l) | ² , and these are input to the SN ratio estimation unit 16B.

The signal-to-noise ratio estimation unit 16B calculates and outputs a gain coefficient R (ω, l) that enhances the signal in the desired direction region using Expression (13).

Gain factor R (omega, l), as shown in FIG. 9, a specific direction signal _{| S qs (ω, l)} | sum of signals ² and angular region _{Σ q | S q (ω,} l) | 2 of the relationship Is | S _qs (ω, l) | ² ≈Σ _q | S _q (ω, l) | ² , the gain coefficient R (ω, l) is a maximum value, for example, “1”, When | S _qs (ω, l) | ² << Σ _q | S _q (ω, l) | ² , the gain coefficient R (ω, l) gives ≈0.

According to the characteristics of the gain coefficient shown in FIG. 9, the gain coefficient R (ω, l) increases as the specific direction signal component S _qs (ω, l) is larger than the signal component S _q (ω, l) in the other angle region. Approaches “1”, and the signal in that frequency band is output without being attenuated. As the specific direction signal component S _qs (ω, l) is smaller than the other angle region signal component S _q (ω, l), the gain coefficient R (ω, l) becomes closer to “0”. The signal component is attenuated by multiplication of the gain factor R (ω, l). As a result, the specific direction signal component is emphasized and output by multiplying the specific direction signal extracted by the specific direction selection unit 15 by the gain coefficient R (ω, l).

Here, the principle that enables the sound collection by selectively emphasizing the desired sound according to the present invention will be described. The output power of each beamformer section, which is each element of the beamformer section output power vector Y * (ω, l), is a signal X _θ * (*) received by the microphone array as shown in Expression (14). It can be approximated by multiplying the power of ω, l) by the directivity characteristics based on the sound source direction and frequency of the signal. However, here, it is assumed that the sounds emitted by the sound sources are uncorrelated with each other, and the sound is received at the same level in all microphones.

いま信号Ｘ_Θｑ※（ω，ｌ）は図３で示したいずれかの方向領域Θ_ｑに含まれる信号の総和を表すとする。このときビームフォーマー部の指向特性は方向領域内で一様であると仮定すると、Ｙ※（ω，ｌ）は式（１５）により表される。本実施例では各領域に対する指向特性の代表値として式（９）により求めた指向特性の平均値を用いている。

Now it signals _X Θq ※ (ω, l) is a representative of the sum of the signal included in either direction region theta _q shown in FIG. At this time, assuming that the directivity characteristic of the beam former is uniform in the direction region, Y * (ω, l) is expressed by Expression (15). In this embodiment, the average value of the directivity obtained by the equation (9) is used as the representative value of the directivity for each region.

以上の関係より、あらかじめ与えられている行列Ｔ※の逆行列を左側からビームフォーマー部出力ベクトルＹ※（ω，ｌ）にかけることで、Ｘ＾※（ω，ｌ）の推定値である推定信号パワーベクトルＸ_ｅｓｔ※（ω，ｌ）が求められる。

From the above relationship, the estimated value of X ^ * (ω, l) is obtained by applying the inverse matrix of the matrix T * given in advance to the beamformer unit output vector Y * (ω, l) from the left side. An estimated signal power vector X _est * (ω, l) is obtained.

In the second embodiment, the procedure in the gain coefficient calculation unit 16 of the first embodiment is modified. FIG. 10 shows a processing procedure of the gain coefficient calculation unit 16 used in the second embodiment. The difference from the gain coefficient calculation unit 16 in the first embodiment is that a nonlinear processing unit 16C is added. In the nonlinear processing unit 16C, the original gain coefficient of the characteristic shown in FIG. 9 output from the SN ratio estimation unit 16B is expressed as R ′ (ω, l), and this original gain coefficient R ′ (ω, l) A gain coefficient R (ω, l), which is a calculation result obtained by multiplying a nonlinear function that varies between 0 and 1, is output. Here, the nonlinear function is given in advance, and the region where the original gain coefficient R ′ (ω, l) is large, that is, Σ _q | S _q (ω, l) | ² ≈ | S _qs (ω, l) | ² In the region, the value is maintained at 1 or close to 1, and the original gain coefficient R ′ (ω, l) is small, that is, Σ _q | S _q (ω, l) | ² >> | S _qs (ω, l) | ² is a function that maintains 0 or a value close to 0. For example, the hypothetical tangent shown in the equation (17) (for example, Z1: R (ω, l) = 1/2 + 1 / 2tanh (ρR ′ (ω, l) -Ν)) and a logarithmic function (for example, Z2: R (ω, l) = 1/2 + 1 / 2tanh (ρ (10 log ₁₀ ) R ′ (ω, l) −ν)) shown in equation (18) Things are used. FIG. 11 shows an example of the nonlinear function.

ここでρ、νは非線形関数の特定を変化させるパラメータで任意に設定される。これ以外の部分に関しては実施例１と同様であるので説明を省略する。図１１に示した非線形特性によれば、所望音声が優勢な周波数領域ではその周波数成分を強調し、背景雑音が優勢な周波数領域ではその周波数成分を抑圧することができ、本発明による特定方向強調量を向上させる効果がある。

Here, ρ and ν are arbitrarily set as parameters for changing the specification of the nonlinear function. Since other parts are the same as those in the first embodiment, the description thereof is omitted. According to the non-linear characteristic shown in FIG. 11, the frequency component can be emphasized in the frequency region where the desired speech is dominant, and the frequency component can be suppressed in the frequency region where the background noise is dominant. There is an effect of improving the amount.

In the third embodiment, the processing procedure in the signal amount estimation unit 14, the gain coefficient calculation unit 16, and the multiplication unit 17 in the first embodiment is changed. FIG. 12 shows the overall configuration of the present invention in the third embodiment. The difference from the first embodiment is that band division units 19-1 to 19-Q are provided in the subsequent stages of the frequency domain conversion units 13-1 to 13-Q, and the band division units 19-1 to 19-Q. It is the point which set it as the structure which provided the same band component collection part 20-1-20- (omega | ohm) which collects each band component divided | segmented by (5) for every same frequency band.
Each of the same band component collection units 20-1 to 20-Ω includes a signal amount estimation unit 14, a specific direction selection unit 15, a gain coefficient calculation unit 16, and a multiplication unit 17. Therefore, the feature of the third embodiment is that the signal amount estimating unit 14, the specific direction selecting unit 15, the gain coefficient calculating unit 16, and the multiplying unit 17 are Ω different same band component collecting units 20-1, 20-... 20-Ω separately, and the output of the multiplier 17 provided in the same band component collection unit 20-1 to 20 -Ω is synthesized by the band synthesizing unit 21.

FIGS. 13 and 14 show processing procedures of the band dividing units 19-1 to 19-Q and the band synthesizing unit 21, respectively. In the band dividing units 19-1 to 19-Ω, the input frequency components Y ₁ (ω, l) to Y _Q (ω, l) are divided into bands from Ω ₁ to Ω _Ω and output respectively. On the other hand, the band synthesizing unit 21 outputs the frequency components of the input bands from Ω ₁ to Ω _Ω together as one signal Y (ω, l). In this embodiment, each element of the gain matrix T * used in the power estimation unit uses an average value within a given band as shown in, for example, Expression (19) and Expression (20).

以上の部分以外については実施例１と同様であるので説明を省略する。一般にビームフォーマー部の指向特性は周波数に応じて変化するが、実施例１ではすべての周波数においてゲインは均一であると仮定し、ゲイン行列の要素を代入する値を決定していた。本実施例ではゲイン行列を周波数帯域ごとに用意して処理を行うことができるため、信号量推定部１４においてより正しい指向特性に近い処理を行うことができ、推定信号パワーベクトルの推定精度を向上させる効果がある。

Since the other parts are the same as those in the first embodiment, description thereof is omitted. In general, the directivity characteristic of the beamformer unit varies depending on the frequency. In the first embodiment, however, the gain is assumed to be uniform at all frequencies, and the value for substituting the elements of the gain matrix is determined. In this embodiment, the gain matrix can be prepared for each frequency band and processing can be performed, so that the signal amount estimation unit 14 can perform processing closer to the correct directivity and improve the estimation accuracy of the estimated signal power vector. There is an effect to make.

In the fourth example, the procedure in the signal amount estimation unit 14 and the gain coefficient calculation unit 16 in the first example is modified. FIG. 15 shows the configuration of the signal amount estimation unit 14 used in the fourth embodiment, and FIG. 16 shows the configuration of the gain coefficient calculation unit 16. The frequency components Y ₁ (ω, l) to Y _Q (ω, l) input to the signal amount estimation unit 14 are input to the absolute value calculation units 14D-1 to 14D-Q, respectively, and the absolute value | Y _{1 of the} signal (Ω, l) |, | Y ₂ (ω, l) |, | Y ₃ (ω, l) |,... | Y _Q (ω, l) | are output and input to the vectorization unit 14A. The vectorization unit 14A outputs an absolute value vector Y * (ω, l) shown in Expression (21) from the input signal.

絶対値ベクトルは乗算部１４Ｂに入力される。乗算部１４Ｂのもう一方の入力である絶対値推定行列Ｔ^−１※は、逆行列演算部１４Ｃの出力信号である。逆行列演算部１４Ｃは入力されたゲイン行列Ｔ※の逆行列Ｔ^−１※を出力する。
ゲイン行列Ｔ※の各要素は式（８）と同様にビームフォーマー部の各方向領域に対する指向特性が求められるパラメータであり、フィルタ係数から計算される指向特性のゲイン量から式（２２）により定義され、事前に与えられる。

The absolute value vector is input to the multiplication unit 14B. The absolute value estimation matrix T ⁻¹ *, which is the other input of the multiplication unit 14B, is an output signal of the inverse matrix calculation unit 14C. The inverse matrix calculator 14C outputs an inverse matrix T ⁻¹ * of the input gain matrix T *.
Each element of the gain matrix T * is a parameter for which the directivity characteristic with respect to each direction region of the beamformer unit is obtained in the same manner as Expression (8). From the gain amount of the directivity characteristic calculated from the filter coefficient, Expression (22) Defined and given in advance.

乗算部１４Ｂに入力されたビームフォーマー部出力パワーベクトルとパワー推定行列の乗算を周波数成分ごとに行い、推定信号絶対値ベクトルＸ_ｅｓｔ※（ω，ｌ）を出力する。
次に図１６に示す利得係数算出部１６ではベクトル要素抽出部１６Ａは式（２３）に示すように入力された推定信号絶対値ベクトルの第１成分を｜Ｓ_１（ω，ｌ）｜、第２成分を｜Ｓ_２（ω，ｌ）｜、第３成分を｜Ｓ_３（ω，ｌ）｜、…としてそれぞれ出力し、それらはＳＮ比推定部１６Ｂに入力される。

The beamformer output power vector input to the multiplier 14B is multiplied by the power estimation matrix for each frequency component, and an estimated signal absolute value vector X _est * (ω, l) is output.
Next, in the gain coefficient calculation unit 16 shown in FIG. 16, the vector element extraction unit 16A sets the first component of the input estimated signal absolute value vector as | S ₁ (ω, l) | The two components are output as | S ₂ (ω, l) |, the third component is output as | S ₃ (ω, l) |,..., And these are input to the SN ratio estimation unit 16B.

これら以外の部分に関しては第１の実施例と同じであるのでこれ以上の説明を省略する。この第４の実施例によれば実施例１に比べて２乗計算をする必要がないことから演算量を削減することができる。
なおこの実施例４は実施例２の信号量推定部１４と利得係数算出部１６に対しても適用することができる。図１７は実施例２に本実施例４の変更を加えた場合の利得係数算出部１６の構成を示す。

Since other parts are the same as those of the first embodiment, further explanation is omitted. According to the fourth embodiment, it is not necessary to perform the square calculation as compared with the first embodiment, so that the calculation amount can be reduced.
The fourth embodiment can also be applied to the signal amount estimating unit 14 and the gain coefficient calculating unit 16 of the second embodiment. FIG. 17 shows the configuration of the gain coefficient calculation unit 16 when the modification of the fourth embodiment is added to the second embodiment.

  The sound collecting device according to the present invention described above can be entirely configured by hardware. However, in order to achieve the simplest implementation, the above-described procedures are recorded by a program language readable by a computer. An embodiment in which a sound program is created, this sound collection program is installed in a computer, the computer executes the sound collection program, and the computer functions as a sound collection device is the best. The sound collection program according to the present invention is recorded on a computer-readable recording medium such as a magnetic medium, a CD-ROM, or a semiconductor memory, and is installed in the computer from the recording medium or through a communication line. The installed sound collection program is decoded by a CPU provided in the computer, and causes the computer to function as a sound collection device.

  The sound pickup device according to the present invention is utilized in the field of hands-free call devices such as a telephone conference system.

The block diagram for demonstrating the 1st Example of this invention. The figure for demonstrating the directivity of the beam former part used for this invention. The figure for demonstrating the example of a division | segmentation of the directivity direction area | region set to the beam former part used for this invention. The block diagram for demonstrating an example of a structure of the beam former part used for this invention. The block diagram for demonstrating an example of a structure of the signal amount estimation part used for this invention. The figure for demonstrating an example of the directional characteristic of the beam former part used for this invention. The block diagram for demonstrating an example of a structure of the specific direction selection part used for this invention. The block diagram for demonstrating an example of a structure of the gain coefficient calculation part used for this invention. The graph for demonstrating an example of the characteristic of the gain coefficient used for this invention. The block diagram for demonstrating the 2nd Example of this invention. The graph for demonstrating the characteristic of the gain coefficient which the gain coefficient calculation part shown in FIG. 10 calculates. The block diagram for demonstrating the 3rd Example of this invention. The block diagram for demonstrating an example of a structure of the band division part used for the 3rd Example of this invention. The block diagram for demonstrating the structure of the zone | band synthetic | combination part used for the 3rd Example of this invention. The block diagram for demonstrating the 4th Example of this invention. The block diagram for demonstrating the structure of the gain coefficient calculation part used for the 4th Example of this invention. The block diagram for demonstrating the other structure of the gain coefficient calculation part used for the 4th Example of this invention. The figure for demonstrating the sound collection method of the microphone array by a prior art. The block diagram for demonstrating the conventional specific direction sound-collecting apparatus. The figure for demonstrating the sound collection characteristic of the conventional specific direction sound collection apparatus.

Explanation of symbols

DESCRIPTION OF SYMBOLS 11 Microphone array 16 Gain coefficient calculation part 12-1 to 12-Q Beam former part 17 Multiplication part 13-1 to 13-Q Frequency domain conversion part 18 Inverse frequency domain conversion part
14 Signal amount estimation unit 19-1 to 19-Q Band division unit
15 Specific direction selector 20-1 to 20-Ω Same band component collector
21 Band synthesis unit

Claims (10)

A plurality of beamformer sections that emphasize and collect sound coming from angular regions in different directions using output signals of a microphone array configured with a plurality of microphones;
A frequency domain conversion unit that converts each of the angle domain signals collected by the plurality of beamformer units into a frequency domain signal divided into a plurality of band components;
A specific direction selection unit that selects a specific direction frequency domain signal belonging to an angular domain signal of an angular domain in a desired direction among the frequency domain signals output by each frequency domain transform unit;
For each inverse of the gain matrix whose element is a parameter obtained from the directivity characteristics of each of the plurality of beamformer units for each angle region, with the sum of the signals included in each angle region as the signal amount of that angle region. A signal amount estimation unit that multiplies the signal amount of the frequency domain signal output by the region conversion unit and estimates the signal amount of each angle region;
A gain coefficient calculation unit that calculates a gain coefficient for each frequency band by a ratio of a signal amount in an angle region of a desired direction and a sum of signal amounts in all angle regions;
A multiplier that multiplies the signal amount in each corresponding frequency band of the specific direction frequency domain signal by the gain coefficient calculated by the gain coefficient calculator;
A specific direction sound pickup device comprising: A plurality of beamformer units that use the output signals of a microphone array configured with a plurality of microphones to emphasize and collect sound coming from angular regions in different directions; and
A frequency domain conversion unit that converts each of the angle domain signals collected by the plurality of beamformer units into a frequency domain signal divided into a plurality of band components;
A band dividing unit that divides the frequency domain signal output by the frequency domain converting unit into frequency band components;
The same band component collecting unit that collects the band components output by the band dividing unit for each same band component;
A band synthesizing unit that is provided on the output side of the same band component collecting unit and synthesizes band components output from each same band component collecting unit;
With
The same band component collecting unit
A specific direction selection unit that selects a specific direction band component belonging to an angle region signal of a desired direction in which any of the beam former units collects sound;
For the inverse matrix of the gain matrix whose elements are the parameters obtained from the directivity characteristics for each angle region and frequency band of the plurality of beamformer units, with the sum of signals included in each angle region as the signal amount of that angle region. A signal amount estimation unit that multiplies the signal amount of the frequency band component output by each band dividing unit to estimate the signal amount of each angle region;
A gain coefficient calculation unit that calculates a gain coefficient by a ratio of a signal amount in an angle region of a desired direction and a sum of signal amounts in all angle regions;
A multiplier for multiplying the specific direction band component by the gain coefficient calculated by the gain coefficient calculator;
A specific direction sound pickup device comprising: In certain direction pickup device according to Claim 1, the signal amount of the frequency domain signal, a specific direction sound collection device which is a power value of the signal. 2. The specific direction sound pickup apparatus according to claim 1, wherein a signal amount of the frequency domain signal is an absolute value of the signal.   The specific direction sound pickup apparatus according to claim 2, wherein the signal amount of the frequency band component is a signal power value.   The specific direction sound pickup apparatus according to claim 2, wherein the signal amount of the frequency band component is an absolute value of the signal. In certain direction sound collection device according to any one of claims 1 to 6, with respect to the signal amount gain factor calculating characteristics of said gain coefficient calculation unit in the desired direction of angular regions, signal in the other angles regions When the amount is negligibly small, the gain coefficient value is given as a predetermined maximum value, so that the signal amount in the angle region of the desired direction is negligible with respect to the signal amount in other angle regions. The specific direction sound collecting device, wherein the gain coefficient value is calculated with a characteristic that is a value close to 0 when the value is very small. In certain direction sound collection device according to any one of claims 1 to 6, the gain factor calculating characteristics of said gain coefficient calculation unit, signals of the other angular region for the signal amount in the desired direction of angular regions in a small band is maintained at a value close to the maximum value to the maximum value the value of the gain factor, with respect to the signal amount in the desired direction of angular regions, wherein the gain factor in the region signal amount is large and other angular regions The specific direction sound collecting device is characterized in that it is calculated with a characteristic that maintains the value of 0 at a value close to approximately 0 to 0. A specific direction sound collecting program which is written in a computer readable program language and causes the computer to function as the specific direction sound collecting device according to any one of claims 1 to 8 . A recording medium comprising a recording medium readable by a computer, wherein the sound recording program for a specific direction according to claim 9 is recorded on the recording medium.

Images