JPH0769707B2 - Noise control device - Google Patents

Description

TECHNICAL FIELD The present invention relates to an active noise control device that suppresses indoor noise by radiating a noise control signal synthesized from noise received by an observation microphone from a speaker. It is a thing.

"Prior Art" FIG. 5 is a block diagram of a conventional noise control device. In the figure, 1 is a noise source, 2 is an observation microphone that receives a noise signal emitted by the noise source 1, 3 ₁ and 3 ₂ are filters that generate a noise control signal using the noise signal received by the observation microphone 2, 4 ₁ and 4 ₂ are speakers that radiate the noise control signals generated by the filters 3 ₁ and 3 ₂ into the room, and 5 is an error represented by a difference signal between the noise signal emitted from the noise source 1 and the noise control signal. It is a noise control point for observing signals.

When the noise signal generated by the noise source 1 at time j in FIG. 5 is x _j , the error signal y _j observed at the noise control point 5 is expressed as follows.

y _j = {g _n −g _m・ (g _s1 h ₁ ＋ g _s2 h ₂ )} ・ x _j …… where g _m : Room transfer function between noise source 1 and observation microphone 2 g _s1 , g _s2 : Indoor transfer function between speakers 4 ₁ and 4 ₂ and noise control point 5 h ₁ and h ₂ : Transfer function of filter 3 ₁ and 3 ₂ g _n : Indoor transfer between noise source 1 and noise control point 5 To control (zero) the error signal y _j from the function expression,
Filters 3 ₁ and 3 ₂ having transfer functions h ₁ and h ₂ satisfying the following expression
You can see that

g _m · (g _s1 h ₁ + g _s2 h ₂ ) ＝ g _n …… By transforming the formula, the following ′ formula is obtained.

g _s1 h ₁ ＋ g _s2 h ₂ ＝ g _n · 1 / g _m …… ′ Here, the transfer functions g _s1 and g _s2 have no common zero.
If the speakers 4 ₁ and 4 ₂ are properly arranged, the left side of the formula g _s1 h ₁ + g
The existence of the filters 3 ₁ , 3 ₂ that synthesize _s2 h ₂ into an arbitrary transfer function is guaranteed by the multi-input multi-output inverse filtering theorem MINT [1].

However, when the observation microphone 2 is not placed near the noise source 1, the room transfer function g _m between the observation microphone 2 and the noise source 1 is a non-minimum phase function, so g _n · 1 / g _m
Does not exist as stable. That is, when the observation microphone 2 cannot be installed in the vicinity of the noise source 1 in the conventional method and the indoor reflected sound is added to the received noise signal, the indoor noise observed at the noise control point 5 should be controlled. You cannot do it.

An object of the present invention is to provide an active noise control device that exhibits good noise suppression performance even when an observation microphone cannot be installed near a noise source and an indoor reflected sound is added to a received noise signal. It is to be.

"Means for Solving the Problem" The present invention uses a plurality of noise control units each consisting of a noise observation microphone, a filter, and a speaker, so that at any noise control point less than the noise control units, an arbitrary sound pressure Is realized. The present invention will be described below with reference to FIG. 1 in the case where the number of noise control points is one. In the figure, the same parts as those in FIG. 5 are designated by the same reference numerals and the description thereof is omitted. Also in the figure
6 ₁ and 6 ₂ represent the noise control unit of the present invention, and 7 ₁ , 7 ₂ , 8 ₁ ,
Reference numerals 8 ₂ and 9 ₁ and 9 ₂ represent a noise observation microphone, a filter, and a speaker, which constitute the noise control unit 6 ₁ and 6 ₂ , respectively.

The noise control system of FIG. 1 can be represented by a block diagram in FIG. In Fig. 2, g _m1 and g _m2 are the indoor transfer functions from the observed noise source 1 to the microphones 7 ₁ and 7 ₂ , and g _s1 and g _s2 are the indoor transfer functions from the speakers 9 ₁ and 9 ₂ to the noise control point 5. Function, h ₁ ,
h ₂ represents a transfer function of the filters 8 ₁ and 8 ₂ , and g _n represents an indoor transfer function between the noise source 1 and the noise control point 5. Further, when replacing the room transfer function g _s1, g _s2 and transmission filters _81, 82 function h _1, h _2, FIG. 2 can be modified as Figure 3. Third
From the figure, the error signal y observed at the noise control point 5 at time j
_j is y _j = {g _n − (g _m1 g _s1 h ₁ + g _m2 g _s2 h ₂ )} · x _j, where x _j is a noise signal generated by the noise source at time j. Therefore, in order to control (set to zero) the error signal y _j , the filter 8 ₁ having the transfer functions h ₁ and h ₂ satisfying the following expression
8 ₂ should be realized.

g _m1 g _s1 h ₁ ＋ g _m2 g _s2 h ₂ ＝ g _n …… Then, the following equation is obtained.

g ₁ h ₁ + g ₂ h ₂ = g _n ...... 'Therefore, if the arrangement of observation microphones 7 ₁ and 7 ₂ and speakers 9 ₁ and 9 ₂ is determined so that g ₁ and g ₂ do not have a common zero point, many The existence of the filters 8 ₁ and 8 ₂ that make the error signal y _j zero is guaranteed by the input multi-output inverse filtering theorem MINT [1], and their coefficients are expressed as follows.

z ^-1 : delay operator (unit delay) m: order of transfer function g _n , g ₁ , g ₂ i: order of transfer function h ₁ , h ₂ T: transposed matrix However, observation microphones 7 ₁ , 7 ₂ and noise It may happen that the room transfer functions g _m1 and g _m2 to and from the source 1 become difficult to measure, and the filters 8 ₁ and 8 ₂ cannot be defined using the transfer functions clearly as in the equation. In consideration of such a case, the noise control device of the present invention obtains the filters 8 ₁ and 8 ₂ by using the following method.

First, assume the following conditions.

I Noise signal emitted from noise source 1 is steady noise II Transfer function from loudspeakers 9 ₁ and 9 ₂ to noise control point 5
g _s1 and g _s2 are known Here, I can be considered that many noises such as air-conditioning noise are stationary, and II radiates M-sequence signals from speakers 9 ₁ and 9 ₂ and should be measured in advance. , Is a realistic condition. Further, no special constraint condition is given to the transfer functions g _m1 and g _m2 from the observation microphones 7 ₁ and 7 ₂ to the noise source 1.

Now, in FIG. 3, input signals x _1j , x of h ₁ , h ₂ at time j
_2j is obtained by the convolution operation of the known transfer functions g _s1 and g _s2 and the noise signals u ₁ and u ₂ received by the observation microphones 7 ₁ and 7 ₂ to which the indoor reflected sound is added. Then, at time j, noise control point 5
The error signal y _j observed at is given by the following equation.

y _j = r _j −H ^T X _j …… Accordingly, the filter coefficient [H ₁ ^T H ₂ ^T ] ^T that minimizes the expected value E {y _j ² } of the error power is obtained as follows.

[H ₁ ^T H ₂ ^T ] ^T = E { X _j X _j ^T } ⁻¹ · E { X _j r _j } ... The specific calculation method of the filter coefficient [H ₁ ^T H ₂ ^T ] ^T is as follows. In addition to the method of directly solving the expression from X _j , r _j observed for a sufficient time, the filter coefficient [H ₁ j ^T H ₂ j ^T ] ^T at time j is calculated using the iterative calculation based on the successive approximation algorithm. There is also a method that can be converged to. As the successive approximation algorithm, for example, the LMS algorithm [2] H _{j + 1} = H _j + α X _j y _j. alpha: high computational convergence factor is increased, fast recursive least squares algorithm convergent [3] _{H j + 1 = H j +} μψ j -1 X j y j ...... b ψ j + 1 = αψ j + X j X ^T ...... c however μ, α: Coefficients can be used.

Next, make sure that the solution [H ₁ ^T H ₂ ^T ] ^T given by the equation agrees with the solution of the equation. When the equation is modified by using the fact that X _j = [G ₁ G ₂ ] ^T X _j ..., The following equation is obtained.

[H ₁ ^T H ₂ ^T ] ^T = E {[G ₁ G ₂ ] ^T X _j X _j ^T [G ₁ G ₂ ]} ^-1 · E {[G ₁ G ₂ ] ^T X _j X _j ^T R} ...... Further, if the above formula is modified in consideration of the assumption I, [H ₁ ^T H ₂ ^T ] ^T = [G ₁ G ₂ ] ^-1 R ...... ', which agrees with the solution represented by the formula. Therefore, the filters 8 ₁ , 8 ₂ that make the error signal y _j observed at the noise control point 5 zero
Was confirmed to have the filter coefficient shown by the formula.

That is, according to the noise control device of the present invention, even when the observation microphones 7 ₁ and 7 ₂ cannot be installed in the vicinity of the noise source, and therefore the indoor reflected sound is added to the received noise signal, the noise control point 5 It is possible to control the indoor noise observed at.

In the above, the case of controlling the noise observed at one point in the room has been described as an example, but even in the case of noise control targeting a plurality of points in the room, the “prior art” and the “means for solving the problem” It can be explained exactly as described in ".

"Embodiment" An embodiment of the present invention will be described with reference to FIG. After the noise signal emitted from a noise source 1, the room reflected sound are added in the figure, the noise control unit 6 ₁ ~6 _{k (k} =
2, 3 ...) Observation microphones 7 _{1 to} 7 _k and noise control points less than the number of noise units 5 _{1 to} 5 _L installed at noise control points microphones 100 _{1 to} 100 _L To be done. The observation microphones 7 _{1 to} 7 _k convert the received signals into digital signals u _{1 to} u _k and output the respective signals to the filters 8 ₁ to 8 _k and the convolution calculator 101. In the filters 8 ₁ to 8 _k , the noise control signals s _{1 to} s _k are combined by the convolution operation of u _{1 to} u _k and the filter coefficients H _{1 to} H _k . Noise control signal s ₁ ~s _k synthesized in the filter 8 ₁ to 8 _k is the speaker 9 ₁ to 9 _k
After being input to the analog signal and converted into an analog signal, it is radiated into the room from these speakers and becomes an error signal represented by the difference signal with the noise signal radiated from the noise source 1. The noise control point microphones 100 _{1 to} 100 _L Observed in. Convolution calculator
The transfer function g _VW from the speaker 9 ₁ to 9 _k measured in advance in 101 to the noise control point microphone _{100 1 ~100 L (v = 1,2} ......
k; w = 1,2 ... L) and observation microphone 7 _{1 at} time j
And an output signal u ₁ ~u _k of _{~7 k, x VW = g VW} · u V ...... where u _V: output signal of the observation microphone constituting the first v-th noise control unit is synthesized. The output signal XX _j of the convolution calculator 101 and the error signal Y _j observed and digitized by the noise control point microphones 100 _{1 to} 100 _L are input to the filter coefficient calculator 102. In the filter coefficient calculator 102, the following algorithm [4] obtained by expanding the successive approximation algorithm a described in “Means for solving the problem” for multipoint control [4] H _{j + 1} = H _j + α XX _j Y _j , H _j : Coefficients of filters 8 ₁ to 8 _k at time j i: order of the filter 8 ₁ ~8 _k y ₁ (j) ... y _L (j): Error signal observed at time j α: Convergence coefficient is executed, coefficients of filters 8 ₁ to 8 _k for time j + 1 are determined, and then coefficient setting input of each filter Send the coefficient to. The coefficients of the filters 8 ₁ to 8 _k are set to the noise control point microphones 100 _{1 to} 100 _L by repeating the above process.
It converges to a coefficient that makes the expected value of the error signal power observed at 0 be zero.

That is, if the noise control device of the present invention is used, a microphone for observing a noise signal cannot be installed in the vicinity of the noise source, and the indoor noise observed at the noise control point is controlled even when the indoor reflected sound is added to the received signal. It becomes possible.

It should be noted that the present embodiment does not incorporate means for removing acoustic coupling (howling) that occurs when the noise control signals radiated from the speakers 9 _{1 to} 9 _{k are} received by the observation microphones 7 _{1 to} 7 _k . However, considering that the room transfer function between the speakers 9 _{1 to} 9 _k and the observation microphones 7 _{1 to} 7 _k can be measured, it is clear that the above acoustic coupling can be eliminated by using the acoustic echo canceller. Let's do it.

[Advantages of the Invention] As described above, according to the noise control device of the present invention, not only when an observation microphone that receives a noise signal can be installed near the noise source, but also when an observation microphone that receives a noise signal is detected near the noise source. Even if the microphone cannot be installed and the indoor reflected sound is added to the received signal, the indoor noise observed at the noise control point can be controlled.

"Documents" [1] Miyoshi, Kanada: Inverse filter problem of sound field, Proc. Of the 2nd Digital Signal Processing Symposium B-1-1 (19
87) [2] Haykin, Translated by Takebe: Introduction to Adaptive Filters, Contemporary Engineering, p.108 (1987) [3] Haykin, Translated by Takebe: Introduction to Adaptive Filters, Contemporary Engineering, p.140 (1987) [4] ] SJ Elliott et al .: A Multiple Error LMS Algo
rithm and Its Application to the Active Control of
Sound and Vibration, IEEE trans.ASSP, vol.35, pp1423
-1434 (1987)

[Brief description of drawings]

FIG. 1 is a block diagram of a noise control device of the present invention used for noise control at one point in a room, and FIG.
FIG. 3 is a block diagram of the noise control device of the present invention, FIG. 3 is a block diagram obtained by exchanging transfer functions g _s1 , g _s2 and h ₁ , h ₂ in FIG. _2, and FIG. FIG. 5 is a diagram showing an embodiment of a noise control device of the present invention used when performing noise control in a room, and FIG. 5 is a configuration diagram of a conventional noise control device used when performing noise control at one point in a room. . 1: Noise source, 2: Observation microphone used in conventional noise control device, 3 ₁ , 3 ₂ : Filter used in conventional noise control device, 4 ₁ , 4 ₂ : In conventional noise control device Speakers used, 5, 5 _{1 to} 5 _k : noise control points provided in the room, 6 _{1 to} 6 _k : noise control unit of the present invention, 7 _{1 to} 7 _k : observation microphones constituting the noise control unit, 8 ₁ to 8 _k : Filter configuring noise control unit, 9 _{1 to} 9 _k : Speaker configuring noise control unit, 100 _{1 to} 100 _L : Noise control point microphone installed at noise control point, 101: Convolution calculator , 102: filter coefficient calculator, g _n : transfer function from noise source 1 to noise control point 5, g _m1 , g _m2 : indoor transfer function from noise source 1 to observation microphones 7 ₁ , 7 ₂ , g _s1 , g _s2 : Speaker 9
Room transfer function from ₁ , 9 ₂ to noise control point 5, h ₁ , h ₂ : transfer function of filter 8 ₁ , 8 ₂ , N: noise source, u ₁ ~ u _k : observation microphone 7 ₁ ~ 7 _k Sound signal received and digitized, s ₁ ~
s _k : noise control signal synthesized by the filters 8 ₁ to 8 _k , y _j : time j at the noise control point when noise control is performed at one point in the room
Error signal observed at x _j : noise signal emitted from noise source 1 at time j, r _j : noise signal emitted from noise source 1 to noise control point 5 at time j, x _1j , x _2j : Output signals u ₁ and u ₂ of the observation microphone and transfer functions g ₁ and g _{2 at} time j, respectively.
, Y _{1 to} y _L : noise control point microphones 100 _{1 to} 100 _{L at} time j in the embodiment of the present invention.
XX _j : output signal of convolutional computing unit 101, H _{j + 1} : coefficients of filters 8 ₁ to 8 _{k with} respect to time j + 1 (output signal of filter coefficient computing unit 102).

Claims (1)

[Claims] 1. An observation microphone for observing a noise signal generated by a noise source, a filter for synthesizing a noise control signal for suppressing the indoor noise by using the noise signal received by the observation microphone, and the noise signal A plurality of noise control units each consisting of a speaker radiating into the room, noise control point microphones provided at each of the noise control points less than the number of these noise units, output signals of the noise observation microphones, and Convolution calculation means for performing convolution calculation of the transfer function from each speaker to each noise control point, the calculation result of the convolution calculation means, and the output of each noise control point microphone are input, and these noise control point microphones are input. The coefficient of each filter is calculated by the successive approximation algorithm that minimizes (zeroizes) the sum of squares of the output of Noise control apparatus characterized by comprising a filter coefficient calculating means for setting the filter.