JP2011221087A - Active noise controller - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an active noise controller capable of reducing noise even if noise from multiple noise sources include noise in the same frequency band by applying active noise control to the noise from noise source(s) other than arbitrarily selected noise source(s).SOLUTION: The active noise controller Y1 includes a sound source separation part 141 that separates residual waves from each of noise sources A1-A3 based on residual waves detected by residual detection microphones 12a-12c. Based on noise waves detected by noise detection microphones 11a-11c and residual waves separated by the sound source separation part 141, interfering waves are emitted from noise control speakers 13a-13c to cancel the noise waves from noise sources other than one or plural noise source(s) selected by a sound source selection part 142 from among the noise sources A1-A3.

Description

  The present invention relates to a feedforward type active noise control device that attenuates a noise wave by interfering with a noise wave having the same amplitude and opposite phase, and in particular, only arbitrary noise among noises from a plurality of noise sources. It is related with the technique for attenuating.

In general, there are two methods for suppressing noise generated by air conditioning fans and engines, such as passive noise control (PNC), which uses a sound absorbing material, and noise waves with the same amplitude and opposite phase. Active noise control (ANC) for attenuating noise waves is known. PNC is effective for high-frequency noise, but the low-frequency noise has a long wavelength. Therefore, the equipment becomes large and costs are required.
On the other hand, ANC is effective against low frequency noise. The ANC is classified into two types, feedforward type and feedback type, according to its structure, and the present invention relates to an active noise control apparatus for executing the former feedforward type ANC.

One of the typical feed-forward ANC algorithms is the Filtered-X LMS method. In this Filtered-X LMS method, a transfer function of a portion called a secondary system from a noise control speaker that outputs silenced sound to a residual detection microphone provided at a noise control target point is measured or measured in advance before the operation of the ANC. It is necessary to estimate. For this reason, when an environmental change occurs in a system in which an ANC device is installed, an error may occur between the value of the secondary transfer function prepared in advance and the value after the change, and the operation of the ANC may become unstable. .
On the other hand, the simultaneous equation method is known as a technique that does not require measurement or estimation of the transfer function of the second-order system in advance and can perform stable ANC operation even if the system environment changes (for example, Non-patent documents 1 and 2).
This simultaneous equation method introduces an auxiliary filter that identifies the overall system characteristics by adding the secondary system from the noise control speaker to the residual detection microphone to the primary system from the noise detection microphone to the residual detection microphone. By solving simultaneous equations composed of the same number of equations, the filter coefficients of the second-order system are calculated as needed. As a result, it is possible to automatically follow changes in the transmission system caused by changes in the environment such as temperature and humidity without requiring preparations for the secondary system. Non-Patent Document 2 discloses that the simultaneous equation method can be adopted even when there are a plurality of noise sources, noise detection microphones, noise control speakers, and noise control target points.

By the way, in noise measurement experiments (such as noise contribution analysis experiments) conducted during the development of machinery and equipment, it is necessary to analyze the noise individually for each part that generates noise when the equipment is assembled. It is desirable to create a state where there is no noise in the part.
In addition, as an environment in which an active noise control device that executes feed-forward ANC using the simultaneous equation method described above is actually used, reduction of occupant's ear noise in a vehicle (CAB) can be considered. At this time, for example, engine noise may be considered as the main noise in the vehicle. However, since the engine sound of a sports car is comfortable to the driver, it may be desired not to perform ANC for the engine sound.
Here, for example, Patent Document 1 discloses a technique for executing ANC so as not to reduce noise in a specific frequency band. Thereby, for example, it is possible not to reduce only noise in the frequency band of engine sound.
On the other hand, for example, Patent Documents 2 and 3 disclose sound source separation processing that can separate a sound signal for each sound source from a mixed sound signal on which sound signals generated from a plurality of sound sources are superimposed.

JP 2006-213297 A JP 2006-154314 A JP 2008-295010 A

Kensaku Fujii, Kotaro Yamaguchi, Shigeyuki Hashimoto, Yusuke Fujita, Mitsuji Muneyasu, “Verification of simultaneous equations method by an experimental active noise control system”, Acoustical science and technology 27 (5) The Acoustical Society of Japan, September 2006, p. .270-277 Shinji Muneyasu, Takashi Asai, Kensaku Fujii, Takao Hinamoto, “Extension of simultaneous equation method to multichannel active noise control system”, IEICE Journal A Vol.J83-A No.11, November 2000, p.1235-1243

However, in the configuration in which noise in a specific frequency band is excluded from the target of ANC as in Patent Document 1, when noise including the same frequency band is generated from a plurality of noise sources, the noise from the specific noise source Cannot be excluded from ANC. That is, not only the noise of a specific noise source that is not desired to be reduced, but also noise in the same frequency band as the noise of the specific noise source among the noise sources that are desired to be reduced.
Therefore, the present invention has been made in view of the above circumstances, and an object of the present invention is to provide a plurality of noise sources even when the noise generated by the plurality of noise sources includes noise in the same frequency band. It is an object of the present invention to provide an active noise control apparatus capable of obtaining a noise reduction effect by ANC by excluding only noises from arbitrarily selected noise sources.

In order to achieve the above object, the present invention provides a plurality of noise detection means for individually detecting noise waves generated from a plurality of noise sources, and a noise wave of the noise source provided individually for each of the noise sources. A silencer output means for outputting a silencer for reducing noise, a residual detection means having the same number as the noise sources for detecting residual waves at a noise control target point at which the noise wave is desired to be reduced, and the residual detection means From the residual wave separating means for separating the residual wave corresponding to each of the noise sources based on the residual wave detected by each, and from the noise wave reduction target by the silencer output means among the plurality of noise sources Exclusion selection means for selecting one or a plurality of noise sources to be excluded, and the plurality of noise sources based on the noise wave detected by each of the noise detection means and the residual wave separated by the residual wave separation means Of the exclusion selection means And a silencer generating means for generating a silencer output from each of the silencer output means in order to cancel noise waves from other noise sources other than the selected one or a plurality of noise sources. Configured as an active noise control device.
According to the present invention, the residual wave separation means separates the residual wave for each noise source, not for each frequency band, from the residual waves detected by each of the residual detection means. Then, based on the residual wave corresponding to each of the noise sources after being separated by the residual wave separating means, the silencer generating means generates the silenced sound. Therefore, even if the noise wave from each noise source includes the same frequency band, only the noise wave from any noise source is excluded from the noise waves generated from each noise source. The noise reduction effect by ANC can be obtained for the noise waves from the noise sources.
Further, each of the noise detection means includes a plurality of separated noise detection means for detecting a noise wave from the corresponding noise source, and the noise source is detected based on the noise wave detected by each of the separated noise detection units. It is conceivable that the noise wave from the sound is separated and detected. Thereby, only the noise wave from the noise source can be detected by excluding the noise wave generated separately from the noise source corresponding to the noise detection means at the installation position of each of the noise detection means. Therefore, the accuracy of the generation of the silencer by the silencer generation means is increased, and the noise reduction effect can be enhanced.
Here, the exclusion selection means is selected from noise wave reduction targets by the silencer output means in response to an operation input to a predetermined operation input means such as a sound source selection switch for switching whether or not to select one or a plurality of noise sources. It is conceivable that one or more noise sources to be excluded are selected. Thus, the user can obtain a noise reduction effect for noise from other noise sources by leaving only a desired noise source by operating the predetermined operation input means.
In addition, the sound source separation unit and the noise separation unit may be configured to execute sound source separation processing by a blind sound source separation method based on an independent component analysis method. Thereby, even if the noise wave from each noise source includes the same frequency band, the residual wave and noise wave corresponding to each noise source can be detected separately.

  According to the present invention, since the silencer is generated by the silencer generator based on the residual wave corresponding to each of the noise sources after being separated by the residual wave separator, each noise source is Even if the noise wave from the same frequency band is included, only noise waves from any noise source out of the noise waves generated by each noise source are excluded and noise from other noise sources is excluded. Noise reduction effect by ANC can be obtained for waves.

The block diagram which shows schematic structure of the active noise control apparatus which concerns on embodiment of this invention. The flowchart explaining an example of the procedure of the active noise reduction process performed with an active noise control apparatus. The block diagram which shows the acoustic characteristic in an ANC process part. The block diagram which shows the structure of the comprehensive system in an ANC process part. The flowchart explaining an example of the procedure of the update process of the optimal filter coefficient performed in an ANC process part. The block diagram which shows schematic structure of the active noise control apparatus which concerns on the Example of this invention. The flowchart explaining an example of the procedure of the active noise reduction process performed with the active noise control apparatus which concerns on the Example of this invention.

Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings so that the present invention can be understood. The following embodiment is an example embodying the present invention, and does not limit the technical scope of the present invention.
First, a schematic configuration of the active noise control device Y1 according to the embodiment of the present invention will be described with reference to FIG.
As shown in FIG. 1, the active noise control apparatus Y1 according to the embodiment of the present invention includes noise detection microphones 11a to 11c (an example of noise detection means) and residual detection microphones 12a to 12c (an example of residual detection means). ), Noise control speakers 13a to 13c (an example of a silencer output unit), a DSP 14, a sound source selection switch 15 (an example of a sound source selection unit), and the like.

As shown in FIG. 1, the active noise control device Y1 includes noise detection microphones 11a to 11c and residual detection microphones corresponding to the noise sources A1 to A3 in acoustic spaces where the noise sources A1 to A3 exist in different directions. Reference signals X (z) input from the noise sources A1 to A3 through the noise detection microphones 11a to 11c in the DSP 14 in a state where the noise control speakers 12a to 12c and the noise control speakers 13a to 13c are provided, respectively. Is passed through the ANC processing unit 143 to generate a silenced sound having the same amplitude and opposite phase as the reference signal X (z), and the silenced sound is output from each of the noise control speakers 13a to 13c.
As a result, the silenced sound is made to interfere with noise waves from the noise sources A1 to A3, and the noise control target point B1 (hereinafter referred to as “control point B1”) where the residual detection microphones 12a to 12c are installed. It is possible to realize a feed-forward type ANC that reduces noise in).
As described below, the active noise control device Y1 according to the embodiment of the present invention is characterized in that it can arbitrarily select a noise source to be subjected to ANC among the noise sources A1 to A3. have. In the present embodiment, the case where there are three noise sources A1 to A3 will be described as an example, but the number is not limited as long as there are at least two noise sources. When the number of noise sources is different, the same number of noise detection microphones, residual detection microphones, and noise control speakers are provided.

The noise detection microphones 11a to 11c are provided corresponding to the noise sources A1 to A3, individually detect noise waves from the corresponding noise sources A1 to A3, and serve as reference signals X (z). Input to the DSP 14. Note that a vibration sensor or the like (an example of noise detection means) may be used instead of the noise detection microphones 11a to 11c.
The noise control speakers 13a to 13c are individually provided corresponding to the noise sources A1 to A3 between the noise sources A1 to A3 and the control point B1 that is the target of noise reduction by the active noise control device Y1. ing. Then, the noise control speakers 13a to 13c output silenced waves generated by an ANC processing unit 143, which will be described later, in order to reduce the noise waves of the noise sources A1 to A3, and the noise from the noise sources A1 to A3. It interferes with (combines) waves. Therefore, if the silencer has the same amplitude and opposite phase as the noise wave from the noise sources A1 to A3, the noise wave can be attenuated to cancel the noise at the control point B1.
On the other hand, the residual detection microphones 12a to 12c are provided at the control point B1, detect residual waves that reach the control point B1 from the noise sources A1 to A3, and perform a residual signal E (z). To the DSP 14. Here, the residual wave is noise remaining after interference between the noise wave from the noise sources A1 to A3 and the silenced sound from the noise control speakers 13a to 13c. The residual detection microphones 12a to 12c and the noise control speakers 13a to 13c are provided in the same number corresponding to the noise sources A1 to A3.

The DSP 14 is generally configured to include a sound source separation unit 141 (an example of a residual wave separation unit), a sound source selection unit 142 (an example of an exclusion selection unit), an ANC processing unit 143 (an example of a silencer generation unit), and the like. It is a digital signal processor specialized in digital signal processing.
The sound source separation unit 141 separates residual signals E (z) corresponding to the noise sources A1 to A3 based on the residual signals E (z) input from the residual detection microphones 12a to 12c. The sound source separation processing detected in this way is executed. Each residual signal E (z) corresponding to each of the noise sources A1 to A3 after being separated by the sound source separation unit 141 is individually input to the sound source selection unit 142.

Specifically, the sound source separation unit 141 is considered to execute a sound source separation process by a blind sound source separation method (BSS method) based on an independent component analysis method (ICA method). In the BSS method sound source separation processing based on the ICA method, even if the residual wave corresponding to each of the noise sources A1 to A3 includes the same frequency band, the residual wave corresponding to each of the noise sources A1 to A3. Can be separated. Of course, even if the residual wave corresponding to each of the noise sources A1 to A3 includes the same frequency band, if the residual wave corresponding to each of the noise sources A1 to A3 can be separated, the sound source separation process is performed. The method is not limited to this. Note that the BSS sound source separation processing based on the ICA method is disclosed in the above-mentioned Patent Documents 2 and 3 and the like, and therefore will be briefly described here and detailed description thereof will be omitted.
The BSS sound source separation processing based on the ICA method uses a predetermined separation matrix (inverse mixing) by utilizing the fact that noises from a plurality of noise sources are statistically independent in an acoustic signal input through a plurality of microphones. This is a processing method for separating (identifying) the sound signal for each sound source by optimizing the matrix) and applying the filter processing with the separation matrix to each of the input sound signals.
Here, in the sound source separation processing, a time domain independent component analysis method (TDICA) calculated in the time domain and a frequency domain independent component analysis method (FDICA) calculated in the frequency domain are employed. In Y1, FDICA calculated in the frequency domain is employed in the same manner as the ANC processing in the ANC processing unit 143 described later.

In addition, the sound source selection unit 142 excludes one or more of the noise sources A1 to A3 from the noise wave reduction target by the ANC processing unit 143 according to the contents of the operation input of the sound source selection switch 15 by the user. The noise source is selected. For example, the sound source selection switch 15 is a plurality of hard switches for switching whether or not each of the noise sources A1 to A3 is selected. Further, instead of the sound source selection switch 15, an operation signal indicating which of the noise sources A1 to A3 is selected according to an operation input to the computer may be input to the sound source selection unit 142. Good. Further, any one of the noise sources A1 to A3 is automatically selected from among the noise sources A1 to A3, such as a noise source having the highest noise wave intensity and other noise sources excluding the lowest noise source. One or a plurality of noise sources may be automatically selected.
Specifically, the sound source selection unit 142 may select one or more of the residual signals E (z) from the noise sources A1 to A3 separated by the sound source separation unit 141 by the sound source selection switch 15. Only the residual signal E (z) corresponding to noise sources other than the noise source is individually input to the ANC processing unit 143.

The ANC processing unit 143 performs ANC only on noise from other noise sources except for one or a plurality of noise sources selected by the sound source selection switch 15 among the noise sources A1 to A3, The sound deadening for reducing only the noise is generated and output from the noise control speakers 13a to 13c. Specifically, the ANC processing unit 143 includes the noise sources A1 to A3 based on the noise wave detected by the noise detection microphones 11a to 11c and the residual wave separated by the sound source separation unit 141. In order to cancel noise waves from other noise sources excluding one or a plurality of noise sources selected by the sound source selection unit 142, a sound wave to be output from the noise control speakers 13a to 13c is generated. For example, when ANC is performed on noise from the noise source A1, the reference signal X (z) detected by the noise detection microphone 11a and the noise source A1 separated by the sound source separation unit 141 are used. Based on the corresponding residual signal E (z), a silencer to be output from the noise control speaker 13a is generated.
Thereby, in the active noise control device Y1, a noise reduction effect can be obtained only for noise from a noise source not selected by the sound source selection switch 15.

In the active noise control device Y1 configured as described above, an ANC process is executed in accordance with, for example, the flowchart shown in FIG.
First, in step T11, the sound source separation unit 141 of the DSP 14 is based on the residual signals E (z) detected by the residual detection microphones 12a to 12c, and the residuals corresponding to the noise sources A1 to A3. The difference signal E (z) is separated and extracted and input to the sound source selection unit 142.
Next, in step T12, the sound source selection unit 142 of the DSP 14 determines whether or not a noise source to be excluded from ANC targets is selected by the sound source selection switch 15. Here, if it is determined that the noise source to be excluded from the ANC target is selected (Yes side of T12), the process proceeds to step T13, but the noise source to be excluded from the ANC target is not selected. If judged (No side of T12), the process proceeds to step T14.
In step T13, the sound source selection unit 142 of the DSP 14 corresponds to the noise source selected by the sound source selection switch 15 among the residual signals corresponding to the noise sources A1 to A3 input from the sound source separation unit 141. Only the residual signals of other noise sources excluding the residual signal are output to the ANC processing unit 143.

Thereafter, in step T14, the ANC processing unit 143 performs ANC on the noise from the noise sources A1 to A3.
At this time, when one or a plurality of noise sources are selected by the sound source selection unit 142, the ANC processing unit 143 only detects noise from other noise sources other than the selected noise source. ANC processing is executed.
On the other hand, when the noise source is not selected by the sound source selection unit 142, the ANC processing unit 143 executes ANC processing for all the noises from the noise sources A1 to A3.
Thus, in the active noise control device Y1, each of the reference signals X (z) detected by the noise detection microphones 11a to 11c and each of the noise sources A1 to A3 after being separated by the sound source separation unit 141 are obtained. Since the ANC processing by the ANC processing unit 143 is performed based on each of the residual waves E (z) corresponding to the noise source A1, the noise source A1 is different from the conventional configuration for reducing noise in a specific frequency band. Even if the noise wave from each of A3 to A3 includes the same frequency band, only the noise wave from any preselected noise source among the noise waves generated from each of the noise sources A1 to A3 The noise reduction effect by ANC can be obtained for noise waves from other noise sources.

Hereinafter, an example of the ANC process of the simultaneous equation method executed by the ANC processing unit 143 will be described with reference to FIGS.
Here, for convenience of explanation, an example in which the noise sources A2 and A3 are selected by the sound source selection switch 15 and excluded from the target of the ANC process, and only the noise source 1 is the target of the ANC process. Will be described. That is, the ANC processing unit 143 executes ANC for attenuating only noise from the noise source A1. If a plurality of noise sources are not selected, the same ANC is executed for each noise source.
In addition, the ANC processing in the ANC processing unit 143 can be realized by calculating in either the time domain or the frequency domain as in the sound source separation unit 141. However, in the active noise control device Y1, since the sound source separation unit 141 performs sound source separation processing in the frequency domain by the FDICA method as described above, the ANC processing unit 143 also performs ANC processing in the frequency domain as described later. It is desirable to perform. Accordingly, the ANC processing unit 143 can handle the acoustic signal output from the sound source separation unit 141 in the frequency domain.

As shown in FIG. 3, the ANC processing unit 143 is provided with a noise control filter 104, an auxiliary filter 105, and a feedback control filter 106 corresponding to the noise source A1. Here, P (z), C (z), H (z), S (z), B (z), and B ′ (z) shown in FIG. 3 are respectively a primary system, a secondary system, and a noise control filter. 104, transfer functions of the auxiliary filter 105, the feedback system, and the feedback control filter 106 are shown. Further, N (z), X (z), E (z), and D (z) shown in FIG. 3 are respectively the primary noise generated in the noise source A1, the input signal of the noise control filter 104, and the noise source A1. A corresponding residual signal, an error signal between the residual signal E (z) and the output signal of the auxiliary filter 105 is shown.
The ANC processing unit 143 updates the filter coefficient of the transfer characteristic S (z) of the auxiliary filter 105 so that the error signal D (z) is minimized (0) in the ANC based on the simultaneous equation method. The overall system characteristics from the input of the noise control filter 104 to the residual detection microphones 12a to 12c are identified.

ここで，Ｃ'（ｚ）は，下記の式（２）で表される。

従って，二次系の伝達関数Ｃ'（ｚ）はΔＢ（ｚ）によって変化する。ここに，Ｂ（ｚ）は帰還系であって，騒音制御スピーカ１３ａからの制御音が騒音検出マイクロフォン１１ａに入る系である。そして，その帰還系Ｂ（ｚ）の相殺誤差ΔＢ（ｚ）は，下記の式（３）で表される。

なお，帰還制御フィルタ１０６のＢ'（ｚ）は，ハウリングが発生することを防止するものである。ここでは，前記帰還制御フィルタ１０６のＢ'（ｚ）が適切に設定されており，ハウリングは発生していないことを前提とする。
前記式（１）から一次系Ｐ（ｚ）を相殺し，残差検出マイクロフォン１２ａからの残差信号Ｅ（ｚ）をゼロ，即ち補助フィルタ１０５の伝達関数Ｓ（ｚ）＝０となるときの騒音制御フィルタ１０４の伝達関数Ｈｏｐｔ（ｚ）は，下記の式（４）で表される。

但し，このとき一次系の伝達特性Ｐ（ｚ）及び二次系の伝達特性Ｃ'（ｚ）は未知であるため，これでは騒音制御フィルタ１０４の伝達関数Ｈｏｐｔ（ｚ）を求めることはできない。 Here, FIG. 4 shows the structure of the total system identified by the auxiliary filter 105. In contrast to this structure, the transfer function S (z) of the auxiliary filter 105 is expressed by the following equation (1) if the error after identification is ignored.

Here, C ′ (z) is represented by the following formula (2).

Accordingly, the transfer function C ′ (z) of the secondary system varies with ΔB (z). Here, B (z) is a feedback system in which the control sound from the noise control speaker 13a enters the noise detection microphone 11a. The cancellation error ΔB (z) of the feedback system B (z) is expressed by the following equation (3).

Note that B ′ (z) of the feedback control filter 106 prevents howling from occurring. Here, it is assumed that B ′ (z) of the feedback control filter 106 is appropriately set and no howling has occurred.
When the primary system P (z) is canceled from the above equation (1), the residual signal E (z) from the residual detection microphone 12a is zero, that is, the transfer function S (z) = 0 of the auxiliary filter 105 The transfer function Hopt (z) of the noise control filter 104 is expressed by the following equation (4).

However, since the transfer characteristic P (z) of the primary system and the transfer characteristic C ′ (z) of the secondary system are unknown at this time, the transfer function Hopt (z) of the noise control filter 104 cannot be obtained.

この連立方程式はＨ１（ｚ）≠Ｈ２（ｚ）であるため解くことが可能であり，その解は下記の式（７），（８）で表される。

そして，前記式（７），（８）を前記式（４）へ代入すると，騒音制御フィルタ１０４の最適フィルタ係数の伝達特性Ｈｏｐｔ（ｚ）は，下記の式（９）で与えられる。

これにより，入力された騒音波を伝達特性Ｈｏｐｔ（ｚ）の騒音制御フィルタ１０４に通すことで同振幅・逆位相の消音波を生成し，該消音波を騒音制御スピーカ１３ａから出力すれば，残差検出マイクロフォン１２ａ〜１２ｃが設けられた騒音制御対象地点Ｂ１ではその騒音がキャンセルされる。 Therefore, in the simultaneous equation method, the transfer characteristic P (z) of the primary system and the transfer characteristic C ′ (z) of the secondary system are calculated by setting two different filter coefficients for the noise control filter 104. For example, filter coefficients H 1 and H 2 (where H 1 ≠ H 2) are sequentially given as transfer functions H 1 (z) and H 2 (z) of the noise control filter 104, and the overall system transfer function S 1 (z), If S2 (z) is identified, the following two relational expressions (5) and (6) are obtained.

This simultaneous equation can be solved because H1 (z) ≠ H2 (z), and the solution is expressed by the following equations (7) and (8).

Then, when the expressions (7) and (8) are substituted into the expression (4), the transfer characteristic Hopt (z) of the optimum filter coefficient of the noise control filter 104 is given by the following expression (9).

As a result, when the input noise wave is passed through the noise control filter 104 having the transfer characteristic Hopt (z), the sound wave having the same amplitude and the opposite phase is generated and output from the noise control speaker 13a. The noise is canceled at the noise control target point B1 where the difference detection microphones 12a to 12c are provided.

ここで，騒音制御フィルタ１０４の最適フィルタ係数の伝達特性Ｈｏｐｔ（ω）の係数は，先に設定した２通りの騒音制御フィルタ１０４のフィルタ係数Ｈ１，Ｈ２とそれに伴う２通りの補助フィルタ１０５のフィルタ係数Ｓ１，Ｓ２との離散フーリエ変換を前記式（１０）に代入した後に，それを逆フーリエ変換することで得ることができる。
実際にフーリエ変換は離散的にまた高速演算可能な高速フーリエ変換（ＦＦＴ）で実行するため，角周波数ωから周波数ビン番号を示す要素番号ｋに置き換えると，騒音制御フィルタ１０４の周波数特性Ｈｏｐｔ（ｋ）は，下記式（１１）で表される。なお，周波数ビン各々は所定幅ごとの周波数帯域に対応する。

そして，全ての周波数ビン番号ｋについて周波数特性を求めた後，逆フーリエ変換で騒音制御フィルタ１０４の最適フィルタ係数Ｈｏｐｔを算出する。 At this time, the conversion from the transfer function Hopt (z) of the noise control filter 104 to the optimum filter coefficient Hopt may be performed using an inverse Fourier transform. Specifically, when the above equation (9) that gives the transfer function Hopt (z) of the noise control filter 104 is replaced with the frequency domain and the angular frequency is ω, the frequency characteristic Hopt (ω) of the noise control filter 104 is expressed by the following equation: It is represented by (10).

Here, the coefficient of the transfer characteristic Hopt (ω) of the optimum filter coefficient of the noise control filter 104 is the filter coefficients H1 and H2 of the two noise control filters 104 set in advance and the filters of the two auxiliary filters 105 associated therewith. After substituting the discrete Fourier transform with the coefficients S1 and S2 into the equation (10), it can be obtained by inverse Fourier transform.
Actually, the Fourier transform is executed by a fast Fourier transform (FFT) that can be performed discretely and at high speed. Therefore, when the angular frequency ω is replaced with the element number k indicating the frequency bin number, the frequency characteristic Hopt (k ) Is represented by the following formula (11). Each frequency bin corresponds to a frequency band for each predetermined width.

Then, after obtaining the frequency characteristics for all the frequency bin numbers k, the optimum filter coefficient Hopt of the noise control filter 104 is calculated by inverse Fourier transform.

他方，ブロック実行型の周波数領域適応アルゴリズムで総合系を推定する場合には，各周波数ビンにおいて下記の式（１３）の更新式（周波数領域ＬＭＳ:Least Mean Square）で求めることができる。

ここに，ｊは更新回数を示す整数，μはステップサイズ，ｉはＦＦＴの区間数をカウントする値，Ｉは１ブロックに含まれるＦＦＴ区間数である。Ｉを１とすると１回のＦＦＴ区間ごとに総合系の周波数特性Ｓ（ｋ）を更新することとなるが，通常Ｉは２以上とする。その理由は，分母がゼロもしくはゼロに近い値となることを避けることにより，分数部分が発散する確率を下げるためである。 By the way, when the transfer characteristic H (z) of the noise control filter 104 is set, the frequency characteristic S (k) of the auxiliary filter 105 is estimated by a method using a cross spectrum method or a block execution type frequency domain adaptive algorithm. Etc.
For example, when the cross spectrum method is used, the output spectrum E (k) of the residual detection microphone 12a and the output spectrum X (k) of the noise detection microphone 11a can be obtained according to the following equation (12). Yes ( ^* represents a complex conjugate).

On the other hand, when the total system is estimated by the block execution type frequency domain adaptive algorithm, it can be obtained by the update formula (frequency domain LMS: Least Mean Square) of the following formula (13) in each frequency bin.

Here, j is an integer indicating the number of updates, μ is the step size, i is a value for counting the number of FFT sections, and I is the number of FFT sections included in one block. If I is 1, the frequency characteristic S (k) of the total system is updated every FFT interval, but I is usually 2 or more. The reason is to reduce the probability that the fractional part diverges by avoiding the denominator to be zero or close to zero.

FIG. 5 is a flowchart showing an example of the procedure for updating the optimum filter coefficient by the simultaneous equation method. According to the procedure shown in FIG. 5, the optimum filter coefficient Hopt of the noise control filter 104 is sequentially updated from the time when the system power is turned on, so that it is possible to cope with fluctuations in the transmission system. Note that the frequency characteristics H1 (k), H2 (k), S1 (k), S2 () are the frequency characteristics of the following filter coefficients H1, H2, total system characteristics S1 (z), S2 (z), and optimum filter coefficient Hopt. k) and Hopt (k) are each stored in a predetermined memory in the DSP 14.
First, the filter coefficient H1 is set in the noise control filter 104 (step T101). Further, the frequency coefficient H1 (k) of the noise control filter 104 is calculated by performing discrete Fourier transform on the filter coefficient H1 of the noise control filter 104 at this time (step T102). Then, the frequency characteristic S1 (k) of the overall system characteristic is identified according to the equation (13) (step T103).
Next, H2 different from the filter coefficient H1 is set in the noise control filter 104 (step T104). Further, the frequency coefficient H2 (k) of the noise control filter 104 is calculated by performing discrete Fourier transform on the filter coefficient H2 of the noise control filter 104 at this time (step T105). Then, the frequency characteristic S2 (k) of the total system is identified according to the equation (13) (step T106).
Thereafter, the frequency characteristic Hopt (k) of the optimum filter coefficient Hopt is calculated according to the equation (11) (step T107). Subsequently, the frequency characteristic Hopt (k) is subjected to inverse discrete Fourier transform to calculate the optimum filter coefficient Hopt (step T108). Then, the optimum filter coefficient Hopt is set in the noise control filter 104 (step T109).
At this time, the frequency characteristics S1 (k) and H1 (k) stored in the memory are overwritten and saved on the frequency characteristics S2 (k) and H2 (k) (step T110). Thereafter, the frequency characteristic H2 (k) stored in the memory is overwritten and saved on the frequency characteristic Hopt (k) (step T111).
Thereafter, the processes of steps T106 to T111 are repeatedly executed, whereby the frequency characteristic S2 (k) of the total system is identified by the frequency characteristic H2 (k), and the latest frequency characteristics S1 (k), H1 (k ), S2 (k), H2 (k), the optimum filter coefficient Hopt is updated. As a result, when the generated noise is uniform white noise over the entire frequency band or stationary sound with little frequency change, it is possible to reduce the noise in response to fluctuations in the noise transmission system. It becomes.

Next, a schematic configuration of the active noise control device Y2 according to the embodiment of the present invention will be described with reference to FIGS. Here, only differences from the active noise control device Y1 described in the above embodiment will be described.
In the active noise control device Y1 according to the embodiment, the reference signal X (z) indicating the noise wave from the noise source A1 (A2, A3) is detected by the one noise detection microphone 11a (11b, 11c). However, at this time, there is a possibility that other noises than the noise sources A1 to A3 are mixed in each of the reference signals X (z). In this case, normal ANC cannot be performed on the noise waves from the noise sources A1 to A3, and the noise reduction effect obtained by the ANC becomes low.

Therefore, as shown in FIG. 6, the active noise control device Y2 according to the present embodiment replaces the noise detection microphones 11a to 11c with three microphones (separation) for detecting noise from the noise sources A1 to A3. Separation noise microphones 21a to 21c having an example of noise detection means are provided.
Further, in the active noise control device Y2, the DSP 14 separates and detects only the noise from the noise sources A1 to A3 from the noise detected by the three microphones for each of the separation noise microphones 21a to 21c. Noise separation units 144a to 144c are provided. Each of the noise separation units 144a to 144c individually inputs noise signals corresponding to the detected noise sources A1 to A3 to the ANC processing unit 143. Here, the configuration of each combination of the separation noise microphones 21a to 21c and the noise separation units 144a to 144c is an example of the noise detection means.
Here, it is conceivable that the noise separation units 144a to 144c execute sound source separation processing by the blind sound source separation method based on the independent component analysis method, similarly to the sound source separation unit 141. As described above, in the BSS method sound source separation processing based on the ICA method, even when a plurality of noise waves contain the same frequency band, the noise waves can be separated. In addition, the structure which provides separately the electric circuit (ASIC), DSP, etc. which have the function similar to the said noise separation part 144a-144c separately from the said DSP14 may be sufficient.

In the active noise control apparatus Y2, as shown in FIG. 7, the noise separation units 144a to 144c of the DSP 14 separate and extract noise from the noise sources A1 to A3, and the separated and extracted reference signal X Noise separation processing (step T131) for inputting (z) to the ANC processing unit 143 is executed. At this time, since the noise source selected by the sound source selection unit 142 is excluded from the target of the ANC processing, it is not necessary to execute the noise separation processing in the noise separation units 144a to 144c. Therefore, in the step T131, the noise separation units 144a to 144c execute the noise separation process only when the corresponding noise source is not selected in accordance with the control command from the sound source selection unit 142. Conceivable.
According to the active noise control device Y2 configured as described above, the reference signal X (z) of the noise sources A1 to A3 input from the noise separation units 144a to 144c to the ANC processing unit 143 is converted into the noise. Since other noise mixed with the noises of the sources A1 to A3 is highly accurate, the ANC processing unit 143 can perform ANC processing based on each of the reference signals X (z) with high accuracy. .

11a-11c: Noise detection microphones 21a-21c: Separation noise detection microphones 12a-12c: Residual detection microphones 13a-12c: Noise control speaker 14: DSP
141: Sound source separation unit 142: Sound source selection unit 143: ANC processing units 144a to 144c: Noise separation unit 15: Sound source selection switches A1 to A3: Noise source B1: Noise control target points Y1, Y2: Active noise control device

Claims (4)

A plurality of noise detection means for individually detecting noise waves generated from a plurality of noise sources;
A silencer output means provided individually for each of the noise sources and outputting a silencer for reducing noise waves of the noise sources;
The same number of residual detection means as the noise sources for detecting residual waves at a noise control target point where the noise waves are to be reduced;
Residual wave separation means for separating the residual wave corresponding to each of the noise sources based on the residual wave detected by each of the residual detection means;
An exclusion selection means for selecting one or a plurality of noise sources to be excluded from noise wave reduction targets by the silencer output means among the plurality of noise sources;
One or a plurality of noise sources selected by the exclusion selection unit among the plurality of noise sources based on the noise wave detected by each of the noise detection units and the residual wave separated by the residual wave separation unit A silencer generating means for generating a silencer to be output from each of the silencer output means in order to cancel noise waves from other noise sources other than
An active noise control device comprising:   Each of the noise detection means includes a plurality of separated noise detection means for detecting a noise wave from the corresponding noise source, and based on the noise wave detected by each of the separated noise detection units, The active noise control device according to claim 1, wherein the noise wave is separated and detected.   The said exclusion selection means selects the 1 or several noise source excluded from the reduction target of the noise wave by the said muffling output means according to the operation input with respect to a predetermined operation input means. The active noise control device according to any one of the above.   The active noise control device according to claim 1, wherein the sound source separation unit and / or the noise separation unit performs sound source separation processing by a blind sound source separation method based on an independent component analysis method.

Images