JP5454375B2 - Sound processor - Google Patents

Description

  The present invention relates to a technique for generating a plurality of acoustic signals in which incoming sounds from different directions are emphasized by signal processing on a plurality of acoustic signals generated by a non-directional sound collection device.

  Conventionally, a technique for generating a plurality of acoustic signals in which incoming sounds from different directions are emphasized has been proposed. For example, Patent Document 1 discloses a technique for generating stereo-format acoustic signals using four omnidirectional sound pickup devices. Patent Document 2 discloses a technique for generating stereo-type acoustic signals using three unidirectional sound collecting devices. It is also possible to generate multiple sound signals by sound source separation using known techniques such as independent component analysis (ICA) and computational auditory scene analysis (CASA). .

Japanese Patent Laid-Open No. 7-075195 JP 7-298387 A

  However, the techniques of Patent Document 1 and Patent Document 2 have a problem that the apparatus becomes large in scale because three or more sound collecting devices are required to generate two systems of acoustic signals. Furthermore, since the unidirectional sound collecting device is essential in the technique of Patent Document 2, it is difficult to mount the unidirectional sound collecting device (for example, for mounting in a portable electronic device). The technique of Patent Document 2 cannot be applied to a case where a omnidirectional small sound collecting device must be adopted. In addition, in sound source separation using independent component analysis or computational auditory scene analysis, there is a problem that processing time (latency) and power consumption increase due to the complexity of arithmetic processing. In view of the above circumstances, an object of the present invention is to generate a plurality of acoustic signals by a simple process using a small number of omnidirectional sound collecting devices.

  Means employed by the present invention to solve the above problems will be described. In order to facilitate understanding of the present invention, in the following description, the correspondence between each element of the present invention and the element of each of the embodiments described later is indicated in parentheses, but the scope of the present invention is not limited to the embodiment. It is not intended to limit the configuration.

  The acoustic processing apparatus of the present invention generates a sum component (for example, sum component M) by adding the spectra of the first acoustic signal and the second acoustic signal generated by the sound collecting devices at each position across the reference line. Means (for example, sum component generation unit 34) and difference component generation means (for example, difference component generation unit 36) for generating a difference component (for example, difference component S) by subtracting the spectrum of the first acoustic signal and the second acoustic signal. , A first combining unit (for example, the first combining unit 382) that generates a first component (for example, the right component R) by adding the sum component and the difference component, and a second component (for example, by subtracting the sum component and the difference component) A second synthesizing unit (for example, the second synthesizing unit 384) that generates the left component L), and a first processing coefficient sequence (for example, for emphasizing the incoming sound from the first side (for example, the left side) across the reference line First coefficient sequence generation that generates a processing coefficient sequence GL) by subtraction of the sum component and the first component Means (for example, coefficient sequence generation unit 42L) and a second processing coefficient sequence (for example, processing coefficient sequence) for emphasizing the incoming sound from the second side (for example, the right side) opposite to the first side across the reference line GR) is generated by subtraction of the sum component and the second component (for example, a coefficient sequence generator 42R), and a third processing coefficient sequence for emphasizing the incoming sound from the direction of the reference line ( For example, a third coefficient sequence generation unit (for example, a coefficient sequence generation unit 42C) that generates a processing coefficient sequence GC) by subtraction of a sum component and a difference component is provided.

  In the above configuration, the directivity is directed in the direction intersecting the reference line (that is, the dead angle of sound collection is formed in the direction of the reference line) by subtracting the spectrum of the first acoustic signal and the second acoustic signal. Since the difference component is generated, the first component emphasizing the incoming sound from the second side of the reference line is generated by adding the sum component and the difference component, and the incoming sound from the first side of the reference line is emphasized. The second component is generated by subtraction of the sum component and the difference component. Therefore, the first processing coefficient sequence generated by subtraction of the sum component and the first component acts to enhance the incoming sound from the first side across the reference line, and the sum of the sum component and the second component The second processing coefficient sequence generated by the subtraction acts to enhance the incoming sound from the second side across the reference line. As described above, in the sound processing device of the present invention, the first processing coefficient sequence that emphasizes the incoming sound from the first side of the reference line and the second processing coefficient sequence that emphasizes the incoming sound from the second side. And a third processing sequence for enhancing the incoming sound from the direction of the reference line can be generated from the first acoustic signal and the second acoustic signal generated by the two non-directional sound pickup devices. is there. That is, there is an advantage that a plurality of acoustic signals can be generated by a simple process using a small number of omnidirectional sound collecting devices. As described above, in the present invention, the directivity of the sound collecting device is not necessary in principle, but an acoustic signal generated by a sound collecting device having directivity (for example, unidirectionality) may be processed. Is possible.

  The sound processing apparatus according to a preferred aspect of the present invention is a signal obtained by adding (including a weighted sum or average) one or both of the first sound signal and the second sound signal, or the first sound signal and the second sound signal. First directivity imparting means (for example, directivity imparting section 52L) that applies a first processing coefficient sequence to the processed signal as a processed signal (for example, frequency spectrum X), and a second processing coefficient sequence to the processed signal Second directivity imparting means (for example, directivity imparting section 52R) to be applied, and third directivity imparting means (for example, directivity imparting section 52C) for applying a third processing coefficient sequence to the signal to be processed. In the above aspect, each processing coefficient sequence is applied with the addition of at least one or both of the first acoustic signal and the second acoustic signal as the signal to be processed. Therefore, the sum component generation unit, the difference component generation unit, and the first synthesis are performed. Compared with a configuration in which each processing coefficient sequence is applied to a signal after processing by the means or the second synthesizing means, it is possible to reduce the influence of noise components that are manifested in each processing. For example, from the viewpoint of compensating for the intensity reduction of the low frequency component due to subtraction by the difference component generation means, a configuration in which correction means (for example, a correction unit 364) that amplifies the low frequency component of the difference component is added is preferable. The noise component is also manifested by the amplification of the low frequency component by the correcting means. Therefore, in the configuration including the correction unit, the form in which each processing coefficient sequence is applied to the signal to be processed that has not been processed by the correction unit is particularly preferable from the viewpoint of reducing the influence of the noise component. Of course, if the noise component is not a problem (for example, the noise component is reduced by other elements), a configuration in which each processing coefficient sequence is made to operate using the acoustic signal corrected by the correcting means as a signal to be processed (for example, a first described later). 2 embodiment) may also be employed.

  The sound processing apparatus according to a preferred aspect of the present invention is any one of a signal generated by the first directivity providing unit, a signal generated by the second directivity providing unit, and a signal generated by the third directivity providing unit. Selection means for selecting (for example, the selection unit 54) is provided. In the above aspect, the incoming sound from the direction of the reference line, the incoming sound from the first side of the reference line, and the incoming sound from the second side of the reference line are separated from each other and selectively used (for example, reproduction) ) Is possible. In addition, the specific example of the above aspect is later mentioned, for example as 1st Embodiment or 3rd Embodiment.

  The sound processing apparatus according to a preferred aspect of the present invention includes a plurality of combinations of a signal generated by the first directivity providing unit, a signal generated by the second directivity providing unit, and a signal generated by the third directivity providing unit. Synthesizer (for example, synthesizer 64) is provided. In the above aspect, a plurality of acoustic signals are generated by synthesizing the signals after processing by each directivity imparting unit. Therefore, for example, by giving priority to the output signal of the third directivity imparting means, an acoustic signal having high directivity in the direction of the reference line is generated, or the output of the first directivity imparting means or the second directivity imparting means Various processes such as generating an acoustic signal in which the sound (stereo feeling) from both sides of the reference line is emphasized by giving priority to the signal are possible. In addition, the specific example of the above aspect is later mentioned, for example as 5th Embodiment.

  The sound processing apparatus according to a preferred aspect of the present invention includes two sound collecting devices according to the combination for each of a plurality of combinations in which two are selected from three or more sound collecting devices arranged at different positions. A synthesis means for synthesizing signals obtained by processing the acoustic signals generated by the processing coefficient sequences generated as the first acoustic signal and the second acoustic signal between different combinations of sound collection devices (for example, the synthesis unit 62) ). In the above aspect, since the signals generated by the processing coefficient sequences corresponding to the different combinations of the sound collecting devices are synthesized, it is possible to change the sound collecting characteristics in various ways. In addition, the specific example of the above aspect is later mentioned as 4th Embodiment, for example.

  The acoustic processing device according to each of the above aspects is realized by hardware (electronic circuit) such as a DSP (Digital Signal Processor) dedicated to processing of an acoustic signal, or a general-purpose calculation such as a CPU (Central Processing Unit). It is also realized by cooperation between the processing device and a program (software). The program of the present invention includes a sum component generation process for generating a sum component by adding the spectra of the first acoustic signal and the second acoustic signal generated by the sound collecting devices at each position across the reference line, the first acoustic signal, Difference component generation processing for generating a difference component by subtracting the spectrum from the second acoustic signal, first synthesis processing for generating a first component by adding the sum component and the difference component, and subtraction of the sum component and the difference component And a second synthesis process for generating the second component, and a first processing coefficient sequence for enhancing the incoming sound from the first side across the reference line by first subtracting the sum component and the first component A second coefficient coefficient generation process and a second process coefficient string for enhancing the incoming sound from the second side opposite to the first side across the reference line are generated by subtraction of the sum component and the second component Coefficient sequence generation processing and third processing coefficient sequence for enhancing incoming sound from the direction of the reference line are summed and difference components To execute a third coefficient sequence generation process which generates at the subtraction to the computer. According to the above program, the same operation and effect as the sound processing apparatus of the present invention are realized. The program of the present invention is provided to a user in a form stored in a computer-readable recording medium and installed in the computer, or provided from a server device in a form of distribution via a communication network and installed in the computer. Is done.

1 is a block diagram of a sound processing apparatus according to a first embodiment. It is explanatory drawing of the position of each sound collection apparatus. It is explanatory drawing of the directivity of a sum component and a difference component. It is explanatory drawing of the phase difference of a sum component and a difference component. It is a block diagram of the sound processing apparatus which concerns on 2nd Embodiment. It is a block diagram of the sound processing apparatus which concerns on 3rd Embodiment. It is a block diagram of the sound processing apparatus which concerns on 4th Embodiment. It is explanatory drawing of the position of each sound collection apparatus in 4th Embodiment. It is explanatory drawing of the sound collection characteristic according to the combination of sound collection apparatuses. It is explanatory drawing of the synthesis | combination of the directional component by a synthetic | combination part. It is explanatory drawing of the directivity characteristic of the sound collection in 4th Embodiment. It is explanatory drawing of the other process by a synthetic | combination part. It is a block diagram of the sound processing apparatus which concerns on 5th Embodiment.

<A: First Embodiment>
FIG. 1 is a block diagram of a sound processing apparatus 100A according to the first embodiment of the present invention. As shown in FIG. 1, the sound processing apparatus 100A is connected to two sound collecting devices MIC (MIC1, MIC2) and two sound emitting devices 16 (16R, 16L). Each sound collecting device MIC is an omnidirectional microphone in which sound collecting sensitivity is substantially uniform in all directions. The sound collecting device MIC1 generates a time domain acoustic signal x1 representing the waveform of surrounding sounds (speech and music). Similarly, the sound collection device MIC2 generates an acoustic signal x2. The A / D converter for converting the acoustic signal x1 and the acoustic signal x2 into a digital signal is not shown for convenience.

  FIG. 2 is a schematic diagram for explaining the arrangement of the sound collection device MIC1 and the sound collection device MIC2. The sound collecting device MIC1 and the sound collecting device MIC2 are arranged at intervals between the respective positions sandwiching the predetermined reference line DC (specifically, positions symmetrical with respect to the reference line DC). The interval d is about 1 cm, for example. The reference point Q on the reference line DC corresponds to the midpoint between the sound collection device MIC1 and the sound collection device MIC2. One side along the reference line DC as viewed from the reference point Q is denoted as “front”, and the opposite side is denoted as “rear”.

  The acoustic processing device 100A in FIG. 1 generates an acoustic signal z1 and an acoustic signal z2 by signal processing on the acoustic signal x1 and the acoustic signal x2. The acoustic signal z1 and the acoustic signal z2 are signals in which the incoming sound from a specific direction is selectively emphasized (signals in which the incoming sound from other directions is suppressed). Specifically, in the acoustic signal z1 and the acoustic signal z2, as shown in FIG. 2, the incoming sound from the range AC (center) ahead and behind the reference point Q and the range AR to the right of the reference line DC are shown. Either the incoming sound from (right) or the incoming sound from the range AL (left) to the left of the reference line DC is selectively emphasized.

  The sound emitting device 16R in FIG. 1 is disposed on the right side (for example, right front) of the listener, and the sound emitting device 16L is disposed on the left side (for example, left front) of the listener. The sound emitting device 16R reproduces sound waves according to the acoustic signal z1, and the sound emitting device 16L reproduces sound waves according to the acoustic signal z2. In addition, illustration of the D / A converter which converts the acoustic signal z1 and the acoustic signal z2 into an analog signal is omitted for convenience.

  As shown in FIG. 1, the sound processing device 100 </ b> A includes an arithmetic processing device 22 and a storage device 24. The storage device 24 stores a program PGM executed by the arithmetic processing device 22 and various types of information used by the arithmetic processing device 22. A known recording medium such as a semiconductor recording medium or a magnetic recording medium or a combination of a plurality of types of recording media is arbitrarily employed as the storage device 24. The arithmetic processing unit 22 executes a program PGM stored in the storage device 24 to thereby generate a plurality of functions (frequency analysis units 32, 32) for generating the acoustic signal z1 and the acoustic signal z2 from the acoustic signal x1 and the acoustic signal x2. A sum component generation unit 34, a difference component generation unit 36, a first synthesis unit 382, a second synthesis unit 384, coefficient sequence generation units 42C to 42L, a signal processing unit 50, a selection unit 54, and a waveform synthesis unit 56) are realized. A configuration in which each function of the arithmetic processing unit 22 is distributed over a plurality of integrated circuits, or a configuration in which a dedicated electronic circuit (DSP) realizes each function may be employed.

  The frequency analysis unit 32 sequentially generates a frequency spectrum (complex spectrum) X1 of the acoustic signal x1 and a frequency spectrum (complex spectrum) X2 of the acoustic signal x2 for each unit period (frame) on the time axis. For the generation of the frequency spectrum X1 and the frequency spectrum X2, a known frequency analysis such as short-time Fourier transform is arbitrarily employed. The frequency spectrum X1 is a series of K component values X1 [1] to X1 [K] corresponding to different frequencies. Similarly, the frequency spectrum X2 is expressed by K component values X2 [1] to X2 [K].

  The sum component generator 34 sequentially generates a sum component M for each unit period by an operation including addition of the frequency spectrum X1 of the acoustic signal x1 and the frequency spectrum X2 of the acoustic signal x2. The sum component M is a series of K component values M [1] to M [K] corresponding to different frequencies. As shown in FIG. 1, the sum component generation unit 34 includes an addition unit 342 and a correction unit 344. The adder 342 generates a sum component M0 by adding the frequency spectrum X1 and the frequency spectrum X2. The sum component M0 is a series of K component values M0 [1] to M0 [K] (M0 [k] = X1 [k] + X2 [k]).

数式(1)の記号ｊは虚数単位を意味し、記号ωは角周波数を意味する。また、記号τは、図１に示すように、基準線ＤCに対して角度θをなす方向からの到来音（速度ｃ）が収音機器ＭIC1および収音機器ＭIC2の各々に到来する時間差（τ＝ｄ・sinθ／ｃ）を意味する。角度θは、例えば、収音機器ＭIC1と収音機器ＭIC2とを通過する直線（エンドファイア方向）が基準線ＤCに対してなす角度（θ＝±π/２）に設定される。数式(1)から理解されるように、周波数が低いほど関数ＨM(ω)は小さい数値に設定される。したがって、加算部３４２による加算に起因した強度の不均衡が補償される。なお、角度θの選定の方法は任意である。例えば利用者からの指示に応じて角度θを可変に設定する構成も採用され得る。 By the way, the incoming sound for the sound collecting device MIC1 and the sound collecting device MIC2 reaches each of the sound collecting device MIC1 and the sound collecting device MIC2 with a time difference (phase difference) according to the arrival direction. Therefore, there is a phase difference between the component of the incoming sound from the specific direction in the acoustic signal x1 and the component in the acoustic signal x2. Since the phase difference between the acoustic signal x1 and the acoustic signal x2 is smaller as the frequency of the arriving sound is lower (low frequency component), the intensity is increased by the addition by the adding unit 342 as the low frequency component of the arriving sound. The correcting unit 344 generates the sum component M by correcting the sum component M0 so that the intensity imbalance described above (the tendency that the intensity increases as the low frequency component) is suppressed. Specifically, the correction unit 344 multiplies the sum component M0 by a function HM (ω) defined by the following formula (1).

The symbol j in the formula (1) means an imaginary unit, and the symbol ω means an angular frequency. Further, as shown in FIG. 1, the symbol τ is the time difference (τ) when the incoming sound (velocity c) from the direction forming the angle θ with respect to the reference line DC arrives at each of the sound collecting devices MIC1 and MIC2. = D · sin θ / c). The angle θ is set to, for example, an angle (θ = ± π / 2) formed by a straight line (endfire direction) passing through the sound collecting device MIC1 and the sound collecting device MIC2 with respect to the reference line DC. As can be understood from Equation (1), the lower the frequency, the smaller the function HM (ω) is set. Therefore, the intensity imbalance caused by the addition by the adding unit 342 is compensated. The method for selecting the angle θ is arbitrary. For example, a configuration in which the angle θ is variably set according to an instruction from the user may be employed.

  As described above, since the sum component M is generated by adding the frequency spectrum X1 and the frequency spectrum X2, as shown in FIG. 3, a virtual omnidirectional (substantially omnidirectionality) located at the reference point Q is obtained. ), The frequency spectrum of the acoustic signal generated by the sound collection device.

  The difference component generation unit 36 in FIG. 1 sequentially generates the difference component S for each unit period by calculation including subtraction of the frequency spectrum X1 of the acoustic signal x1 and the frequency spectrum X2 of the acoustic signal x2. The difference component S is a series of K component values S [1] to S [K] corresponding to different frequencies. As shown in FIG. 1, the difference component generation unit 36 includes a subtraction unit 362 and a correction unit 364. The subtractor 362 generates a difference component S0 (S0 [1] to S0 [K]) by subtracting the frequency spectrum X2 from the frequency spectrum X1 (S0 [k] = X1 [k] −X2 [k]). .

By the way, as understood from FIG. 4, the sum component M0 (M) corresponding to the addition of the frequency spectrum X1 and the frequency spectrum X2 and the difference component S0 generated by subtraction between them are defined in a complex plane. Vector to be orthogonal to each other. Further, as described above, the lower the phase component of the incoming sound, the smaller the phase difference between the acoustic signal x1 and the acoustic signal x2. The correcting unit 364 corrects the difference component S0 by correcting the difference component S0 so that the phase difference from the sum component M0 and the intensity imbalance described above (the tendency that the intensity decreases as the low frequency component) is suppressed. Is generated. Specifically, the correction unit 364 multiplies the difference component S0 by a function HS (ω) defined by the following equation (2).

The first half (e ^{−jπ / 2} ) of Equation (2) means correction of the phase of the difference component S0 (processing for matching the phase to the sum component M0). Further, as can be understood from the latter half (| 1 / (1-e− ^jωτ ) |) of the formula (2), the function HS (ω) is set to a larger numerical value as the frequency is lower. Therefore, the phase difference from the sum component M 0 and the intensity imbalance caused by the subtraction by the subtraction unit 362 are compensated. A configuration that corrects the phase of the sum component M0 (matches the phase to the difference component S0) instead of correcting the phase of the difference component S0 may be employed.

  Since the incoming sound from the direction of the reference line DC (forward or backward) reaches the sound collecting device MIC1 and the sound collecting device MIC2 in substantially the same phase, in the difference component S obtained by subtracting the frequency spectrum X2 from the frequency spectrum X1, The incoming sound from the direction of the reference line DC is suppressed compared to the incoming sound from the other direction. That is, a dead angle of sound collection is formed in the direction of the reference line DC. Therefore, as shown in FIG. 3, the difference component S generated by the difference component generation unit 36 has a bidirectional characteristic in which incoming sounds from the right side (range AR) and the left side (range AL) of the reference line DC are emphasized. Signal. In the example of the present embodiment, since the frequency spectrum X2 is subtracted with the frequency spectrum X1 as a divisor, the component corresponding to the incoming sound from the right side (range AR) of the reference line DC is the sum component in the difference component S. The component corresponding to the incoming sound from the left side (range AL) of the reference line DC has the opposite polarity (negative polarity) to that of the sum component M.

  1 generates a right component R by adding the sum component M generated by the sum component generation unit 34 and the difference component S generated by the difference component generation unit 36. The right component R is expressed by K component values R [1] to R [K] (R [k] = M [k] + S [k]). On the other hand, the second synthesis unit 384 generates the left component L by subtracting the difference component S from the sum component M. The left component L is expressed by K component values L [1] to L [K] (L [k] = M [k] −S [k]).

  Of the difference component S, the component corresponding to the incoming sound from the right side of the reference line DC has the same polarity as the sum component M, and the component corresponding to the incoming sound from the left side of the reference line DC is opposite to the sum component M. Polarity. Therefore, in the right component R generated by the addition of the sum component M and the difference component S, the incoming sound from the right side (range AR) with respect to the reference line DC is emphasized compared to the incoming sound from the other direction, In the left component L generated by subtraction of the sum component M and the difference component S, the incoming sound from the left (range AL) with respect to the reference line DC is emphasized as compared with the incoming sound from the other direction. As can be understood from the above description, the sum component M is used as an M microphone (unidirectional) sound collection signal in the MS (Mid-Side) sound collection technology, and the difference component S is an S microphone. It is used as a (bidirectional) sound pickup signal.

  The coefficient sequence generation unit 42C in FIG. 1 generates a processing coefficient sequence GC for enhancing the incoming sound from the direction of the reference line DC (range AC). The processing coefficient sequence GC is a series of K coefficient values gC [1] to gC [K] corresponding to different frequencies. Similarly, the coefficient sequence generator 42R generates a processing coefficient sequence GR (sequence of coefficient values gR [1] to gR [K]) for emphasizing the incoming sound from the right side (range AR) of the reference line DC. The coefficient sequence generation unit 42L generates a processing coefficient sequence GL (sequence of coefficient values gL [1] to gL [K]) for emphasizing the incoming sound from the left side (range AL) of the reference line DC. . In the following description, the code (C, L, R) for identifying the direction (range) to be emphasized may be generalized with the symbol V (V = C, R, L). For example, the description related to the coefficient sequence generation unit 42V is similarly applied to each of the coefficient sequence generation unit 42C, the coefficient sequence generation unit 42R, and the coefficient sequence generation unit 42L.

Each coefficient sequence generation unit 42V calculates each coefficient value gV [k] (gC [k], gR [k], gL [k] of the processing coefficient sequence GV (GC, GR, GL) by the following equation (3). ]). The reason why the expression (3) includes the division of the sum component M by the amplitude | M [k] | is to normalize the coefficient value gV [k] to a numerical value of 1 or less (0 ≦ gV [k] ≦ 1). It is.

数式(4C_a)から理解されるように、数式(3)の強度ＰC[k]の数値は、差成分Ｓの強度（パワー）|Ｓ[k]|²と所定の係数αCとの乗算値を和成分Ｍの強度|Ｍ[k]|²から減算した数値に設定される（スペクトル減算）。ただし、数式(4C_a)の数値がゼロ以下となる周波数での強度ＰC[k]は所定値（具体的には和成分Ｍの強度|Ｍ[k]|²と所定の係数βCとの乗算値）に設定される（フロアリング処理）。 The coefficient sequence generation unit 42C calculates the strength (power) PC [k] of the equation (3) by the calculation of the following equations (4C_a) and (4C_b).

As understood from the equation (4C_a), the numerical value of the intensity PC [k] in the equation (3) is a product of the intensity (power) | S [k] | ² of the difference component S and a predetermined coefficient αC. It is set to a value obtained by subtracting from the intensity | M [k] | ^{2 of the} sum component M (spectral subtraction). However, the intensity PC [k] at the frequency at which the numerical value of the equation (4C_a) is zero or less is a predetermined value (specifically, the product of the intensity | M [k] | ² of the sum component M and the predetermined coefficient βC). ) (Flooring process).

  Since the difference component S is a component that emphasizes the incoming sound from the right side and the left side with respect to the reference line DC, the series of intensities PC [1] to PC [K] calculated by the equations (4C_a) and (4C_b) Corresponds to a power spectrum in which incoming sounds from the right side and the left side of the sum component M are suppressed. Accordingly, the coefficient values gC [1] to gC [K] of the processing coefficient sequence GC are the coefficient values gC [k] of the frequency where the intensity of the incoming sound from the direction of the reference line DC (front or rear of the reference point Q) is large. As the coefficient value gC [k] of the frequency at which the intensity of the incoming sound from the direction of the reference line DC is smaller, the value is closer to zero.

すなわち、係数列生成部４２Rは、左成分Ｌの強度|Ｌ[k]|²と係数αRとの乗算値を和成分Ｍの強度|Ｍ[k]|²から減算することで強度ＰR[k]を算定するスペクトル減算（数式(４R_a)）と、数式(4R_a)の数値がゼロ以下となる周波数の強度ＰR[k]を所定値（βR|Ｍ[k]|²）に設定するフロアリング処理とを実行する。左成分Ｌは基準線ＤCの左方からの到来音を強調した成分であるから、数式(4R_a)および数式(4R_b)で算定される強度ＰR[1]〜ＰR[K]の系列は、和成分Ｍのうち左方からの到来音を抑圧（右方からの到来音を強調）したパワースペクトルに相当する。したがって、係数列生成部４２Rが数式(3)で生成する処理係数列ＧRの係数値ｇR[1]〜ｇR[K]は、基準線ＤCの右方からの到来音の強度（振幅）が大きい周波数の係数値ｇR[k]ほど１に近い数値となり、基準線ＤCの右方からの到来音の強度が小さい周波数の係数値ｇR[k]ほどゼロに近い数値となる。 The coefficient sequence generation unit 42R calculates the intensity PR [k] of the equation (3) by the calculation of the following equations (4R_a) and (4R_b).

That is, the coefficient sequence generation unit 42R subtracts the product of the intensity | L [k] | ² of the left component L and the coefficient αR from the intensity | M [k] | ² of the sum component M, thereby calculating the intensity PR [k. ] Subtracting the spectrum (Formula (4R_a)) and the flooring to set the intensity PR [k] at which the numerical value of Formula (4R_a) is less than or equal to zero to a predetermined value (βR | M [k] | ² ) Process. Since the left component L is a component that emphasizes the incoming sound from the left side of the reference line DC, the series of intensities PR [1] to PR [K] calculated by the equations (4R_a) and (4R_b) The component M corresponds to a power spectrum in which the incoming sound from the left side is suppressed (the incoming sound from the right side is emphasized). Therefore, the coefficient values gR [1] to gR [K] of the processing coefficient sequence GR generated by the coefficient sequence generation unit 42R using Equation (3) have a large intensity (amplitude) of the incoming sound from the right side of the reference line DC. The frequency coefficient value gR [k] is closer to 1, and the frequency coefficient value gR [k] having a smaller intensity of incoming sound from the right side of the reference line DC is closer to zero.

Similar to the coefficient sequence generation unit 42R, the coefficient sequence generation unit 42L calculates the intensity PL [k] of the expression (3) by the following expressions (4L_a) and (4L_b).

That is, the coefficient sequence generation unit 42L subtracts the product of the intensity | R [k] | ² of the right component R and the coefficient αL from the intensity | M [k] | ² of the sum component M to obtain the intensity PL [k. ] Subtracting the spectrum (formula (4L_a)) and flooring to set the intensity PL [k] of the frequency at which the numeric value of formula (4L_a) is zero or less to a predetermined value (βL | M [k] | ² ) Process. Since the right component R is a component in which the incoming sound from the right side of the reference line DC is emphasized, the series of intensities PL [1] to PL [K] calculated by the equations (4L_a) and (4L_b) is the sum. Of the component M, this corresponds to a power spectrum in which the incoming sound from the right side is suppressed (the incoming sound from the left side is emphasized). Accordingly, the coefficient values gL [1] to gL [K] of the processing coefficient sequence GL generated by the coefficient sequence generation unit 42L using the formula (3) have a large intensity (amplitude) of the incoming sound from the left side of the reference line DC. The frequency coefficient value gL [k] is closer to 1, and the frequency coefficient value gL [k] having a smaller intensity of incoming sound from the left side of the reference line DC is closer to zero.

The coefficient αV (αC, αR, αL) and the coefficient βV (βC, βR, βL) applied to the calculation of the intensity PV [k] by the coefficient sequence generator 42V are set to appropriate numerical values. Further, the numerical value of the intensity PV [k] set in the flooring process is not limited to the above-described example (βV | M [k] | ² ), for example, a predetermined value (for example, zero) that is unrelated to the sum component M Can be set to

  The signal processing unit 50 in FIG. 1 causes each processing coefficient sequence GV (GC, GR, GL) to act on the frequency spectrum X1 of the acoustic signal x1 and the frequency spectrum X2 of the acoustic signal x2. That is, the signal processing unit 50 imparts directivity to the acoustic signal x1 and the acoustic signal x2 (emphasizes incoming sound from a specific direction). As shown in FIG. 1, the signal processing unit 50 includes a directivity imparting unit 52C, a directivity imparting unit 52R, and a directivity imparting unit 52L.

  Each directivity imparting unit 52V (52C, 52R, 52L) applies the processing coefficient sequence GV (coefficient values gV [1] to gV [K]) generated by the coefficient sequence generating unit 42V to the frequency spectrum X1 and the frequency spectrum X2, respectively. The directional component Y1V and the directional component Y2V are sequentially generated for each unit period by acting on (specifically multiplying). The directional component Y1V is a frequency spectrum (complex spectrum) expressed by a sequence of K component values Y1V [1] to Y1V [K]. Similarly, the directional component Y2V is a frequency spectrum expressed by a sequence of K component values Y2V [1] to Y2V [K].

前述のように、処理係数列ＧCの係数値ｇC[1]〜ｇC[K]は、基準線ＤCの方向（前方または後方）からの到来音の強度が大きい周波数の係数値ｇC[k]ほど１に近い数値に設定される。したがって、音響信号ｘ1のうち基準線ＤCの方向からの到来音を強調した指向成分Ｙ1Cと音響信号ｘ2のうち基準線ＤCの方向からの到来音を強調した指向成分Ｙ2Cとが数式(5C)の演算で生成される。 Specifically, the directivity imparting unit 52C calculates the directional component Y1C (component value Y1C [k]) and the directional component Y2C (component value Y2C [k]) by the calculation of the following formula (5C).

As described above, the coefficient values gC [1] to gC [K] of the processing coefficient sequence GC are approximately equal to the coefficient value gC [k] of the frequency at which the intensity of the incoming sound from the direction of the reference line DC (forward or backward) is large. A value close to 1 is set. Therefore, the directional component Y1C that emphasizes the incoming sound from the direction of the reference line DC in the acoustic signal x1 and the directional component Y2C that emphasizes the incoming sound from the direction of the reference line DC in the acoustic signal x2 are expressed by the equation (5C). Generated by calculation.

In addition, the directivity imparting unit 52R calculates the directional component Y1R and the directional component Y2R by the calculation of the following formula (5R). The coefficient values gR [1] to gR [K] of the processing coefficient sequence GR are set to values closer to 1 as the coefficient value gR [k] of the frequency where the intensity of the incoming sound from the right side of the reference line DC is larger. The directional component Y1R generated by Equation (5R) is a frequency spectrum that emphasizes the incoming sound from the right side of the acoustic signal x1. On the other hand, the component values Y2R [1] to Y2R [K] of the directional component Y2R are set to zero (silence).

Similar to the directivity imparting unit 52R, the directivity imparting unit 52L calculates the directivity component Y1L and the directivity component Y2L by the calculation of the following formula (5L). The coefficient values gL [1] to gL [K] of the processing coefficient sequence GL are set to values closer to 1 as the coefficient value gL [k] of the frequency where the intensity of the incoming sound from the left side of the reference line DC is larger. The directional component Y2L generated by Equation (5L) is a frequency spectrum that emphasizes the incoming sound from the left side of the acoustic signal x2. On the other hand, the component values Y1L [1] to Y1L [K] of the directional component Y1L are set to zero.

  1 includes a set of directivity component Y1C and directivity component Y2C generated by the directivity assignment unit 52C, a set of directivity component Y1R and directivity component Y2R generated by the directivity assignment unit 52R, and a directivity assignment unit. One of the set of the directional component Y1L and the directional component Y2L generated by 52L is selected. A selection target by the selection unit 54 is variably set in accordance with, for example, an operation from a user with respect to an operation element (not shown). The directivity component Y1V (Y1C, Y1R, Y1L) selected by the selection unit 54 is supplied to the waveform synthesis unit 56 as the frequency spectrum Z1, and the directivity component Y2V (Y2C, Y2R, Y2L) selected by the selection unit 54 is used as the frequency spectrum Z2. It is supplied to the waveform synthesis unit 56.

  The waveform synthesizer 56 in FIG. 1 generates the time domain acoustic signal z1 and the acoustic signal z2 from the frequency spectrum Z1 and the frequency spectrum Z2 supplied from the selector 54. Specifically, the waveform synthesizer 56 generates the acoustic signal z1 by converting the frequency spectrum Z1 for each unit period into a signal in the time domain by inverse Fourier transform and connecting the unit periods before and after. Similarly, the waveform synthesizer 56 generates an acoustic signal z2 from each frequency spectrum Z2. The acoustic signal z1 is reproduced from the sound emitting device 16R, and the acoustic signal z2 is reproduced from the sound emitting device 16R.

When the selection unit 54 selects the directional component Y1C and the directional component Y2C, the reproduced sound that emphasizes the incoming sound from the direction of the reference line DC (range AC) is emitted from both the sound emitting device 16R and the sound emitting device 16L. The On the other hand, when the selection unit 54 selects the directional component Y1R and the directional component Y2R, the reproduced sound that emphasizes the incoming sound from the right side (range AR) of the reference line DC is emitted from the sound emitting device 16R, and the sound emitting device 16L is set to silence (Y2R = 0). In addition, when the selection unit 54 selects the directional component Y1L and the directional component Y2L, the reproduced sound that emphasizes the incoming sound from the left side (range AL) of the reference line DC is emitted from the sound emitting device 16L, and the sound emitting device 16R is set to silence (Y1L = 0). That is, each sound obtained by separating the incoming sound (sound source separation) according to the direction with respect to the reference point Q is selectively reproduced. The expression (5C) includes the division by the coefficient (1/2) ^1/2 when the incoming sound from the direction of the reference line DC is emphasized (both the sound emitting device 16R and the sound emitting device 16L are The difference in volume perceived by the listener between when the sound is emitted) and when the incoming sound from one side of the reference line DC is emphasized (when only one of the sound emission devices 16R and 16L emits sound). This is to reduce the above.

  According to the first embodiment described above, sound source separation in three directions is realized by signal processing on the acoustic signal x1 and the acoustic signal x2 generated by the two sound collection devices MIC (MIC1, MIC2). The acoustic processing device 100A can be downsized as compared with the technique of Patent Document 1 that requires four sound collecting devices and the technique of Patent Document 2 that requires three sound collecting devices. In addition, since the omnidirectional sound collecting device MIC (MIC1, MIC2) can be used, compared with the technique of Patent Document 2 that requires a unidirectional sound collecting device, the cost required to realize the acoustic processing apparatus 100A. In the case where it is difficult to mount a unidirectional sound collecting device (for example, a small omnidirectional sound collecting device MIC has to be adopted for mounting in a portable electronic device) Case) has an advantage that it can be applied effectively. In addition, the sound signal z1 and the sound signal z2 after sound source separation are generated by a very simple process as compared with sound source separation using independent component analysis (ICA) or computational auditory scene analysis (CASA). Therefore, there is an advantage that processing time (latency) and power consumption are reduced. In the computational auditory scene analysis using the harmonic structure of the sound, the target of sound source separation is restricted to an acoustic signal having the harmonic structure, but the first embodiment does not assume the acoustic harmonic structure. There is also an advantage that the acoustic signal x1 and the acoustic signal x2 can be processed regardless of the presence or absence of the harmonic structure.

<B: Second Embodiment>
A second embodiment of the present invention will be described. In addition, about the element which an effect | action and function are equivalent to 1st Embodiment in each following illustration, the code | symbol referred by the above description is diverted and each detailed description is abbreviate | omitted suitably.

  FIG. 5 is a block diagram of the sound processing apparatus 100B of the second embodiment. In the acoustic processing apparatus 100A of the first embodiment, the frequency spectrum X1 of the acoustic signal x1 and the frequency spectrum X2 of the acoustic signal x2 are supplied to the signal processing unit 50. On the other hand, in the sound processing apparatus 100B of the second embodiment, as shown in FIG. 5, the right component R generated by the first combining unit 382 and the left component L generated by the second combining unit 384 are the signal processing unit 50. It is processed. In other words, the directivity part assigning unit 52V (52C, 52R, 52L) generates the directional component Y1V (Y1C, Y1R, Y1L) by applying the processing coefficient sequence GV to the right component R, and the processing coefficient sequence GV to the left. By acting on the component L, the directional component Y2V (Y2C, Y2R, Y2L) is generated. That is, the component value X1 [k] in the equations (5C) to (5L) is replaced with the component value R [k] in the right direction R, and the component value X2 [k] in the equations (5C) to (5L) is replaced with the left component. Replaced with the component value L [k] of L. In the second embodiment, the same effect as in the first embodiment is realized.

  Note that when the correction unit 364 amplifies the low frequency component of the difference component S0 by correction using the function HS (ω) (Equation (2)) of the equation (2), noise components (for example, each of the difference components S0) Noise generated when the vibration of the diaphragm is changed to an electric signal by the sound collecting device MIC is also amplified. Since the right component R and the left component L are generated using the difference component S, the right component R and the left component L include the noise component of the difference component S amplified by the correction by the function HS (ω). Therefore, in the second embodiment in which the acoustic signal z1 and the acoustic signal z2 are generated by the signal processing on the right component R and the left component L, the noise component amplified by the correction by the correction unit 364 is converted into the acoustic signal z1 and the acoustic signal z2. May appear. In the first embodiment, the acoustic signal z1 and the acoustic signal z2 are generated by signal processing on the acoustic signal x1 and the acoustic signal x2 before the correction by the correcting unit 364 (before the noise component is amplified). There is an advantage that it is possible to generate a high-quality reproduced sound with less noise components than the form.

<C: Third Embodiment>
FIG. 6 is a block diagram of a sound processing apparatus 100C according to the third embodiment. In the acoustic processing apparatus 100A of the first embodiment, two systems of the frequency spectrum X1 and the frequency spectrum X2 are processed by the signal processing unit 50, whereas the signal processing unit 50 of the acoustic processing apparatus 100C of the third embodiment is As shown in FIG. 6, one frequency spectrum X (a sequence of K component values X [1] to X [K]) is processed. The frequency spectrum X is one of the frequency spectrum X1 and the frequency spectrum X2 generated by the frequency analysis unit 32, for example. However, a configuration in which the frequency spectrum X is an addition or average of the frequency spectrum X1 and the frequency spectrum X2 (or an addition or average frequency spectrum of the acoustic signal x1 and the acoustic signal x2) may be employed.

  Each directivity imparting unit 52V (52C, 52R, 52L) of the signal processing unit 50 causes the processing coefficient sequence GV generated by the coefficient sequence generating unit 42V to act on the frequency spectrum X so that one system of directional components YV (YC, YR, YL). The directional component YC (YC [k] = gC [k] · X [k]) is a frequency spectrum that emphasizes the incoming sound from the direction of the reference line DC. In the directional component YR (YR [k] = gR [k] · X [k]), the incoming sound from the right side (range AR) of the reference line DC is emphasized, and the directional component YL (YL [k] = In gL [k] · X [k]), the incoming sound from the left side (range AL) of the reference line DC is emphasized.

  The selection unit 54 selects, as the frequency spectrum Z, the directional component YC generated by the directivity providing unit 52C, the directional component YR generated by the directivity providing unit 52R, or the directional component YL generated by the directivity providing unit 52L. To do. The time domain acoustic signal z generated by the waveform synthesizer 56 from the frequency spectrum Z is reproduced from one sound emitting device 16. Also in the third embodiment, since the three sound source separations are performed using two non-directional sound pickup devices MIC (MIC1, MIC2), the same effect as in the first embodiment is realized. .

<D: Fourth Embodiment>
FIG. 7 is a block diagram of a sound processing apparatus 100D according to the fourth embodiment. As shown in FIG. 7, three sound collection devices MIC (MIC1, MIC2, MIC3) are connected to the sound processing apparatus 100D of the fourth embodiment. As shown in FIG. 8, the three sound collection devices MIC (MIC1, MIC2, and MIC3) are arranged at the positions of the vertices of an equilateral triangle with each side having a length d. The sound collecting device MIC1 generates an acoustic signal x1, the sound collecting device MIC2 generates an acoustic signal x2, and the sound collecting device MIC3 generates an acoustic signal x3.

  In the fourth embodiment, for each of all combinations (three types) of selecting two from the three sound collection devices MIC (MIC1, MIC2, MIC3), each sound collection device MIC of the combination generates 2 The same signal processing as that in the third embodiment is executed on the system acoustic signal x. In the following description, a combination of a sound collection device MICi (i = 1, 2, 3) and a sound collection device MICj (j = 1, 2, 3, i ≠ j) (a combination of an acoustic signal xi and an acoustic signal xj) Subscript “ij” is added to the elements related to).

  As shown in FIG. 8, the sound collection device MICi and the sound collection device MICj are installed at positions symmetrical with respect to the reference line DCij. In the following description, for convenience, the combination of the sound collection device MICi and the sound collection device MICj is one side along the reference line DCij as viewed from the reference point (middle point) Qij between the two (what is the remaining sound collection device MIC? The opposite side is denoted as “front” and the opposite side as “rear”. That is, the notation of the direction (front, back, right, left) with respect to the reference point Qij is individually defined for each combination of the sound collection device MICi and the sound collection device MICj.

  The acoustic processing device 100D of FIG. 7 includes three signal separation units Uij (U12, U23, U31), a synthesis unit 62, and a waveform synthesis unit 56. Each signal separation unit Uij has a configuration in which the selection unit 54 and the waveform synthesis unit 56 are omitted from the acoustic processing apparatus 100C (FIG. 6) of the third embodiment, and the frequency analysis unit 32, the sum component generation unit 34, and the difference component. A generation unit 36, three coefficient string generation units 42C to 42L, and a signal processing unit 50 are provided. Each element of the signal separation unit Uij executes the same processing as in the third embodiment for the acoustic signal xi and the acoustic signal xj.

  Similarly to the third embodiment, each directivity providing unit 52V (52C, 52R, 52L) of the signal processing unit 50 converts the processing coefficient sequence GV generated by the coefficient sequence generation unit 42V into the frequency spectrum X (for example, the acoustic signal xi or The directivity component YVij (YCij, YRij, YLij) is generated by acting on the acoustic signal xj or the sum or average of both. Specifically, the signal separation unit U12 generates three systems of directivity components YV12 (YC12, YR12, YL12) to which the directivity characteristics of the portion (A) in FIG. 9 are given. Similarly, the signal separation unit U23 generates the directivity component YV23 (YC23, YR23, YL23) of the part (B) in FIG. 9, and the signal separation part U31 generates the directivity component YV31 (YC31) of the part (C) in FIG. , YR31, YL31). As can be understood from FIG. 9, each directional component YCij is a signal that emphasizes the incoming sound from the direction of the reference line DCij (front and rear). In the directional component YRij, the incoming sound from the right side of the reference line DCij is emphasized, and in the directional component YLij, the incoming sound from the left side of the reference line DCij is emphasized.

  7 generates a frequency spectrum Z (Za to Zf) by selecting and synthesizing nine directivity components YVij generated in parallel by the three signal separation units Uij. Similarly to the third embodiment, the time domain acoustic signal z generated from the frequency spectrum Z by the waveform synthesis unit 56 is reproduced from one sound emitting device 16.

すなわち、周波数スペクトルＺaの振幅|Ｚa[k]|は、指向成分ＹC12の振幅|ＹC12[k]|と指向成分ＹR23の振幅|ＹR23[k]|との差分値に設定される。ただし、振幅|ＹC12[k]|と振幅|ＹR23[k]|との差分値が負数となる周波数の振幅|Ｚa[k]|はゼロに設定される。 The specific operation of the combining unit 62 will be described below. The synthesizer 62 generates the frequency spectrum Za (amplitude | Za [k] |) by the calculation of the following formula (6a). The operator max (a, b) in the equation (6a) means the maximum value of the numerical value a and the numerical value b.

That is, the amplitude | Za [k] | of the frequency spectrum Za is set to a difference value between the amplitude | YC12 [k] | of the directional component YC12 and the amplitude | YR23 [k] | of the directional component YR23. However, the amplitude | Za [k] | of the frequency at which the difference value between the amplitude | YC12 [k] | and the amplitude | YR23 [k] | is a negative number is set to zero.

  FIG. 10 is a schematic diagram in which the directional component YC12 and the directional component YR23 are extracted. As shown in FIG. 10, since the incoming sound from behind in the directional component YC12 substantially overlaps with the component of the directional component YR23, subtracting the directional component YR23 from the directional component YC12 follows the reference line DC12 in the directional component YC12. The incoming sound from behind is suppressed. Therefore, as shown in FIG. 11, the frequency spectrum Za generated by the equation (6a) is a signal that emphasizes only the incoming sound from the front along the reference line DC12 in the directional component YC12.

Similar to Equation (6a), by calculating the difference between the two directional components YVij selected from the nine directional components YVij, the synthesizer 62 generates incoming sounds from different directions as shown in FIG. Emphasized frequency spectra Zb to Zf are generated. That is, the direction of sound collection in the fourth embodiment is variably controlled in each of six types of directions (front and rear of each of the three types of reference lines DCij) according to the processing by the synthesis unit 62. For calculation of the frequency spectra Zb to Zf, for example, the same equations (6b) to (6f) as the equation (6a) are used.

  In the fourth embodiment, the same effect as in the first embodiment is realized. In the fourth embodiment, the sound collection direction can be variously changed by appropriately combining the directional components YVij generated for each combination of the sound collection devices MIC. Therefore, for example, if the acoustic processing device 100D of the fourth embodiment is mounted on a portable audio recording device (IC recorder), it is uttered for each speaker (for example, a conference participant) located in different directions (six directions). It is possible to separate voice sounds. For example, if the sound processing device 100D of the fourth embodiment is mounted on an imaging device (video camera) that records moving images and sounds, a plurality of incoming sound collections (surround sound collections) from different directions can be obtained. Is possible. Since the interval d between the sound collecting devices MIC of the fourth embodiment is about 1 cm, it can be easily mounted on a small audio recording apparatus or imaging apparatus.

  Note that the synthesis method by the synthesis unit 62 is not limited to the above examples. For example, as shown in part (A) of FIG. 12, by adding the directional component YL12 and the directional component YR23, the collected sound that emphasizes the incoming sound from the left rear (shaded portion) with respect to the reference line DC12 is obtained. Realized. Also, as shown in part (B) of FIG. 12, by adding the directional component YL23 and the directional component YR12, the sound collection that emphasizes the incoming sound from the right front (shaded portion) with respect to the reference line DC12 is obtained. Realized.

<E: Fifth Embodiment>
FIG. 13 is a block diagram of a sound processing apparatus 100E according to the fifth embodiment. As illustrated in FIG. 13, the sound processing device 100E of the fifth embodiment has a configuration in which the selection unit 54 of the sound processing device 100A of the first embodiment is replaced with a synthesis unit 64. The synthesizer 64 generates directivity components Y1V (Y1) generated by the directivity imparting units 52V.
C, Y1R, Y1L) and the directivity component Y2V (Y2C, Y2R, Y2L) are combined to generate the frequency spectrum Z1 and the frequency spectrum Z2.

The combining unit 64 calculates a weighted sum of the directional component Y1C generated by the directivity imparting unit 52C and the directional component Y1R generated by the directivity imparting unit 52R as the frequency spectrum Z1, as shown in the following formula (7R). . Further, as shown in the following formula (7L), the combining unit 64 uses the weighted sum of the directional component Y2C generated by the directivity providing unit 52C and the directional component Y2L generated by the directivity providing unit 52 as the frequency spectrum Z2. Calculate. Symbols γC and γRL in Equation (7R) and Equation (7L) are weight values. The weight value γC and the weight value γRL are set variably in accordance with, for example, an operation from a user with respect to an operator (not shown). The acoustic signal z1 converted from the frequency spectrum Z1 is reproduced from the sound emitting device 16R, and the acoustic signal z2 converted from the frequency spectrum Z2 is reproduced from the sound emitting device 16L.

  As described above in the first embodiment, the incoming sound from the direction of the reference line DC (range AC) is emphasized in the directional component Y1C and the directional component Y2C. In the directional component Y1R, the incoming sound from the right side of the reference line DC is emphasized, and in the directional component Y2L, the incoming sound from the left side of the reference line DC is emphasized. Accordingly, as the weighted value γC in the formulas (7R) and (7L) is set to a larger value than the weighted value γRL, the incoming sound from the direction of the reference line DC ( As the weighted value γRL is set to a larger value than the weighted value γC, the incoming sound from the left and right direction (stereo feeling) is emphasized in the reproduced sound from each sound emitting device 16. The

  In the fifth embodiment, the stereo sound signal z1 and the sound signal z2 are generated by the signal processing of the sound signal x1 and the sound signal x2 generated by the two omnidirectional sound pickup devices MIC (MIC1, MIC2). The Therefore, as in the first embodiment, it is possible to reduce the number of sound collection devices MIC while using non-directional sound collection devices MIC. In addition, the sound signal z1 and the sound signal z2 after the sound source separation are generated by a very simple process as compared with the sound source separation using independent component analysis (ICA) and computational auditory scene analysis (CASA). There is also.

  Note that the fifth embodiment can be suitably employed for audio zoom processing that variably controls sound collection characteristics of an imaging device (video camera) that records moving images and audio, for example. That is, the composition unit 64 variably sets the weight value γC and the weight value γRL according to the focal length (imaging magnification) of the zoom lens of the image pickup apparatus. Specifically, the combining unit 64 sets the weight value γC to the weight value γRL so that the incoming sound from the direction of the reference line DC is emphasized as the focal distance approaches the telephoto end (that is, the shooting range is narrow). Increase it. The synthesizing unit 64 sets the weight value γRL to the weight value γC so that the incoming sound from the left and right directions is emphasized (stereo feeling increases) as the focal length approaches the wide-angle end (that is, the shooting range is wide). Increase against.

<F: Modification>
Each of the above forms can be variously modified. Specific modifications are exemplified below. Two or more aspects arbitrarily selected from the following examples can be appropriately combined.

(1) Modification 1
In the first to third embodiments, the selection target (that is, the sound collection direction) by the selection unit 54 is variably controlled according to an instruction from the user, but according to the intensity of each of the plurality of directional components Y. Thus, a configuration in which the selection unit 54 selects the directional component Y may be employed. For example, if the selection unit 54 selects a directional component Y having a high intensity among a plurality of directional components Y, the blind sound source separation that extracts the incoming sound from the sound source without realizing the direction of the sound source in advance is realized. Is done. When a plurality of sound sources are located in the same direction, incoming sound from each sound source is mixed in the acoustic signal z, but can be separated according to differences in frequency and volume.

(2) Modification 2
In the first to third embodiments, the waveform synthesizing unit 56 is installed at the subsequent stage of the selecting unit 54. However, the configuration in which the selection unit 54 and the waveform synthesizing unit 56 are reversed in reverse (that is, the waveform synthesizing unit 56 has A configuration in which the selection unit 54 selects a plurality of time domain acoustic signals z generated from each directional component Y may also be employed. Similarly, in the fourth and fifth embodiments, the combination of the synthesis units (62, 64) and the waveform synthesis unit 56 is reversed (that is, the waveform synthesis unit 56 generates a plurality of directional components Y from each directional component Y). A configuration in which the synthesizing unit 62 or the synthesizing unit 64 synthesizes the acoustic signal z in the time domain may be employed.

(3) Modification 3
The first to fifth embodiments can be appropriately merged. For example, in the third to fifth embodiments, the right component R generated by the first combining unit 382 and the left component L generated by the second combining unit 384 are supplied to the signal processing unit 50 in the second embodiment. It is possible to adopt a configuration.

(4) Modification 4
In the fourth embodiment, the case where three sound collecting devices MIC are used is exemplified, but the number of sound collecting devices MIC is arbitrary. For example, a configuration using four or more sound collecting devices MIC may be employed. The sound processing apparatus 100D generates a directional component YVij (YCij, YRij, YLij) for each combination of selecting two of the four sound collecting devices MIC. In each of the above embodiments, a case where a plurality of sound collecting devices MIC are arranged linearly or planarly is exemplified, but a configuration in which four or more sound collecting devices MIC are arranged three-dimensionally can also be adopted. For each directional component YVij selected by the synthesizer 62, the coefficient αV applied to the generation of the processing coefficient sequence GV corresponding to the directional component YVij is variably controlled, so that the range of the sound collection directivity is variable. It is also possible to control it.

(5) Modification 5
A configuration in which the processing described in the above embodiments is applied only to components within a predetermined frequency band may be employed. A configuration may also be employed in which each acoustic signal x is divided into a plurality of frequency bands, and the processing in each of the above forms is performed in parallel for each frequency band and then combined.

(6) Modification 6
In each of the above embodiments, the non-directional sound collecting device MIC is used. However, the present invention can also be applied to a configuration using a directivity (for example, unidirectional or bi-directional) sound collecting device MIC. For example, a configuration in which the frequency spectrum of an acoustic signal generated by each of two unidirectional sound pickup devices MIC is a sum component M and a difference component S (therefore, the sum component generation unit 34 and the difference component generation unit 36 are omitted). Is adopted. Further, in the configuration using three or more sound collecting devices MIC as in the fourth embodiment, all sound collecting devices MIC are set to have directivity, and non-directional sound collecting devices MIC and directivity (for example, single It is also possible to use a unidirectional sound collecting device MIC together.

100A, 100B, 100C, 100D, 100E …… Sound processing device, MIC (MIC1, MIC2, MIC3) …… Sound collecting device, 16R, 16L …… Sound emitting device, 22 …… Calculation processing device, 24 …… Storage device , 32... Frequency analysis unit, 34... Sum component generation unit, 36... Difference component generation unit, 382... First synthesis unit, 384 ... Second synthesis unit, 42V (42C, 42R, 42L). Coefficient sequence generation unit, 50... Signal processing unit, 52V (52C, 52R, 52L) .. directivity imparting unit, 54 .. selection unit, 56 .. waveform synthesis unit, 62, 64. U12, U23, U31) …… Signal separator.
.

Claims (5)

An apparatus for processing a first acoustic signal and a second acoustic signal generated by a sound collecting device at each position across a reference line,
Sum component generating means for generating a sum component by addition of spectra of the first acoustic signal and the second acoustic signal;
Difference component generating means for generating a difference component by subtracting the spectrum of the first acoustic signal and the second acoustic signal;
First combining means for generating a first component by adding the sum component and the difference component;
Second combining means for generating a second component by subtraction of the sum component and the difference component;
First coefficient sequence generation means for generating a first processing coefficient sequence for enhancing incoming sound from the first side across the reference line by subtraction of the sum component and the first component;
A second coefficient sequence for generating a second processing coefficient sequence for emphasizing the incoming sound from the second side opposite to the first side across the reference line by subtraction of the sum component and the second component Generating means;
A sound processing apparatus comprising: a third coefficient sequence generation unit configured to generate a third processing coefficient sequence for enhancing incoming sound from the direction of the reference line by subtraction of the sum component and the difference component. One or both of the first acoustic signal and the second acoustic signal or a signal obtained by adding the first acoustic signal and the second acoustic signal is used as a processed signal, and the first processing coefficient sequence is applied to the processed signal. First directivity imparting means for causing
Second directivity imparting means for causing the second processing coefficient sequence to act on the signal to be processed;
The sound processing apparatus according to claim 1, further comprising: a third directivity providing unit that causes the third processing coefficient sequence to act on the signal to be processed. The selection means which selects any of the signal which the said 1st directivity provision means produces | generates, the signal which the said 2nd directivity provision means produces | generates, and the signal which the said 3rd directivity provision means produces | generates. Sound processing equipment. Combining means for generating a plurality of acoustic signals by combining the signal generated by the first directivity imparting means, the signal generated by the second directivity imparting means, and the signal generated by the third directivity imparting means The sound processing apparatus according to claim 2. For each of a plurality of combinations for selecting two of three or more sound collecting devices arranged at different positions, the sound signal generated by each of the two sound collecting devices according to the combination is the first sound signal. The sound processing apparatus according to claim 1, further comprising: a combining unit configured to combine a signal processed by the processing coefficient sequence generated as the second sound signal between different combinations of the sound collection devices.

Images