JPH07107590A - Howling canceller - Google Patents

Abstract

PURPOSE:To eliminate noise generated from the mechanical noise generating source and to maintain the noise elimination function even when the acoustic transmission function in the noise eliminated space is fluctuated. CONSTITUTION:The white noise from the white noise generation device 207 is outputted from a speaker 205 to the inside of a duct 221 before the noise is generated from a noise generating source 200. The sound is captured by a microphone 222 and an signal (e) is given to an adaptive filter 213. The adaptive filter 213 estimates the acoustic transmission function Gao between the speaker 205 and the microphone 222 based on the white noise signal and the acoustic signal (e). The estimated tap coefficient is set to a characteristic compensation filter 202 and the output of the white noise is stopped. The noise from the noise generating source 200 is outputted inside the duct 221 instead. With the use of the adaptive filter 203, the acoustic transmission function between the speaker 201 and the microphone 222 is estimated.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a howling canceller which makes it difficult to generate howling in an acoustic space.

[0002]

2. Description of the Related Art In recent years, research and development have been conducted to realize various acoustic control and acoustic processing by digital signal processing.

For example, a technique of digital signal processing for preventing howling is disclosed in, for example, the following document.

Literature: IEICE Technical Report 199
2 years, EA92-85, "Application of active noise control to howling prevention system".

Such a technique can be applied to reduce noise in an automatic or an airplane, and is required as a technique necessary for an acoustic system.

[0006] Here, with respect to the system described in the above document, a conventional technique will be schematically described with reference to FIG. 2.

In FIG. 2, the howling prevention system mainly includes a noise generation source 1, a noise generation speaker 2, a characteristic compensation filter 3, an adaptive control unit 4, a fixed filter 5, and noise elimination. It is composed of a speaker 6, a microphone 7, and a duct 8.

In the system of FIG. 2, the noise n output from the noise generation source 1 is output from the speaker 2 and also becomes a pseudo noise signal n ^* through the fixed filter 5 from the noise erasing speaker 6. Is output. The noise n output from the speaker 2 is
When propagating through the duct 8, the noise N shown in FIG. 2 is generated by the action of the acoustic transfer function Gso, and reaches the microphone 7.

On the other hand, the pseudo noise signal n ^* output from the noise canceling speaker 6 of FIG. 2 becomes noise N ^{* due} to the action of the acoustic transfer function Gao between the noise canceling speaker 6 and the microphone 7. , Reaches the microphone 7. At this time, if the pseudo noise N ^* and -N have a substantially equal relationship, a noise canceling effect occurs in the space of the duct 8 and the noise N can be reduced to almost zero.

(Pseudo Noise N ^* Generation Method):
A method of generating the pseudo noise N ^* will be described. First, the speaker 2 is turned off so that the signal n from the noise source 1 is input to the fixed filter 5 and the signal is output only from the noise elimination speaker 6. In this case, the fixed filter 5 uses the digital FI.
As is well known, the R filter is mainly composed of a combination of a tap coefficient register and a delay register, and realizes a filter function by a convolution operation of both. The tap coefficient of the fixed filter 5 is set by the adaptive control unit 4, but the initial value is set to 0 for all taps.

As a method of generating the tap coefficient in the adaptive control unit 4 of FIG. 2, the noise canceling residual signal E supplied from the microphone 7 and the noise generation source 1 are supplied via the characteristic compensation filter 3. If the method uses the noise signal n, for example, learning identification method, LMS method, R
Any known algorithm such as the LS method may be adopted.

Then, the adaptive control unit 4 of FIG. 2 controls the tap coefficient of the fixed filter 5 and, when the adaptive control algorithm converges after a certain time T0 (for example, 1 sec), copies the tap coefficient to the characteristic compensation filter 3. To do. The characteristic compensation filter 3 is composed of an FIR filter, and the estimated acoustic characteristic G ^* ao obtained by estimating the acoustic characteristic Gao is set by the copy processing of the adaptive control unit 4.
As a result, the characteristic compensation filter 3 thereafter performs filtering processing on the noise signal n with the estimated acoustic characteristic G ^* ao.

In this case, the tap coefficient of the characteristic compensation filter 3 is not updated after the initial copy processing is performed. Further, the application control unit 4 completes copying the tap coefficient to the characteristic compensation filter 3, and then the fixed filter 5
The tap coefficient of is returned to the initial value of 0.

After the coefficient of the characteristic compensation filter 3 is fixed as described above, the adaptive control unit 4 and the fixed filter 5 are connected.
The adaptive filter is constructed by and the pseudo noise n ^* is generated.

(Operation after the lapse of time T0): Next, the operation after the lapse of time T0 will be described. First, the noise n output from the speaker 2 is filtered by the acoustic transfer function Gso in the duct 8 and becomes the noise N and reaches the microphone 7. In this case, the signal N is expressed by the following equation (1).

N = n · Gso (1) On the other hand, the noise n output from the noise source 1 in FIG.
Is an estimated acoustic transfer function G by the characteristic compensation filter 3.
^* Filtered by ao, it is input to the adaptive controller 4 becomes n · ^G * ao.

Then, the adaptive control unit 4 newly adds the fixed filter 5 based on the signal n · G ^* ao supplied from the characteristic compensation filter 3 and the noise canceling residual signal E supplied from the microphone 7. Find the tap coefficient of.

This tap coefficient is set at the initial operation (time T0
It may be obtained using the same algorithm as used in the previous). However, the tap coefficient obtained in this case is the acoustic transfer function-the transfer function obtained by estimating Gso-G ^*.
It corresponds to so.

The tap coefficient generated by the adaptive control unit 4 of FIG. 2 is used as the filter coefficient of the fixed filter 5, whereby the pseudo noise n ^* output from the fixed filter 5 is given by the following (2). It is represented by a formula.

N ^* = − n · G ^* so (2) Formula Then, the pseudo noise n ^* represented by the above formula (2) is output from the noise elimination speaker 6, and the noise N output from the speaker 2 is output. (= N · Gso) is added in the duct 8.

That is, the following operation (3) is performed in the duct 8.

N + n ^* = n · Gso−n · G ^* so Equation (3) Then, the value of this Equation (3) becomes almost 0, and the duct 8
It is configured to eliminate noise within.

[0023]

However, it is considered that the above-mentioned conventional technique has the following problems.

That is, (1) when the noise source is not an electrical one but mechanical vibration noise of the motor or wind noise generated by the rotation of the fan, a signal is output from the noise canceling speaker 6. However, since the sound of other mechanical system is mixed, the acoustic transfer function (Gao) between the noise elimination speaker 6 and the microphone 7 cannot be estimated. Furthermore, in such a situation, since the tap coefficient of the characteristic compensation filter 3 cannot be set in the initial operation, it is impossible to remove noise in the first place.

(2) Acoustic transfer function between the noise eliminating speaker 6 and the microphone 7 (Gao in FIG. 2)
Is estimated at the initial operation and fixed after that, when the acoustic transfer function in the duct changes (in case of temperature characteristic change, shape change, foreign matter intrusion, aging speaker or microphone, etc.) However, there is a problem in that the noise erasing characteristic is significantly deteriorated and howling occurs.

The present invention has been made in view of the above problems, and an object of the present invention is to remove noise (noise) even when there is a non-electrical mechanical noise source. (EN) It is possible to provide a howling canceller which is capable of preventing the howling cancellation action from deteriorating even when the acoustic transfer characteristics of the acoustic space to be howling canceled fluctuate.

[0027]

Therefore, the first invention is
In order to achieve the above object, a sound generating unit that outputs a sound to the howling cancellation target sound space, a sound capturing unit that captures the sound propagating in the howling cancellation target space, and to perform howling cancellation of the sound Acoustic output means for howling cancellation, which outputs acoustics for howling cancellation, and a first acoustic transfer characteristic between the acoustic output means for howling cancellation and the acoustic capturing means in the acoustic space of the howling cancellation target, or the acoustic generation means and the above. The howling canceller including the howling canceling acoustic signal generating means for generating the howling canceling acoustic signal from the second acoustic transfer characteristic with the acoustic capturing means is realized by the following characteristic configuration.

That is, white noise generating means for generating white noise, and control means for outputting the white noise from the acoustic output means for howling cancellation to the acoustic space for howling cancellation when the acoustic generating means is not operating. , The first acoustic transfer characteristic is estimated from the acoustic signal obtained by the acoustic capturing means by the output of the white noise to the acoustic space for howling cancellation, and the estimation result is given to the acoustic signal generating means for howling cancellation. And a first acoustic transfer characteristic estimating means.

Further, it is characterized in that the howling canceling acoustic signal generating means is caused to set the estimation result of the first acoustic transfer characteristic estimating means.

The acoustic space subject to howling cancellation is, for example, a duct, a room, an automobile, a wider acoustic space, or the like.

The above-mentioned sound generating means is, for example,
It means a speaker device or the like.

Further, the above-mentioned sound capturing means is, for example,
It means a sound collecting device such as a microphone device.

Furthermore, the acoustic output means for howling cancellation described above means, for example, a speaker device.

In a second aspect of the invention, in the configuration of the first aspect of the invention, the acoustic signal is generated after the howling cancellation acoustic signal generating means is set with the estimation result of the first acoustic transfer characteristic estimating means. A second means for operating the generating means, operating the howling canceling acoustic signal generating means and the howling canceling acoustic output means, and estimating the second acoustic transfer characteristic from the acoustic signal obtained by the acoustic capturing means. An acoustic transfer characteristic estimating means is provided.

Further, in a third aspect of the present invention, in the configuration of the above-mentioned second aspect, during the operation of the sound generating means, the first aspect
The change in the first acoustic transfer characteristic is detected from the estimation result of the acoustic transfer characteristic estimating unit and the acoustic signal of the sound generating unit, and is set in the howling canceling acoustic signal generating unit according to the change. It is characterized by further comprising first acoustic transfer characteristic changing means for changing the above-mentioned first acoustic transfer characteristic.

Furthermore, in a fourth aspect of the invention, in the configuration of the above-described second aspect or third aspect, the estimation result of the second acoustic transfer characteristic estimating means during the operation of the acoustic generating means, A change in the second acoustic transfer characteristic is detected from the acoustic signal of the acoustic capturing means, and the second acoustic transfer characteristic set in the acoustic signal generating means for howling cancellation is changed according to the change. The second acoustic transfer characteristic changing means is provided.

A fifth aspect of the invention is to compare the change of the first acoustic transfer characteristic with the change of the second acoustic transfer characteristic in the structure of the fourth aspect of the invention, and the result of the comparison is compared. According to the third aspect of the present invention, there is provided a comparison control unit that preferentially changes any acoustic transfer function set in the howling cancellation acoustic signal generation unit.

[0038]

According to the configuration of the first invention, the first acoustic transfer characteristic (for example, acoustic transfer function) between the acoustic output means for howling cancellation and the acoustic capturing means is accurately measured by using white noise. Can be estimated.

The white noise generating means can be realized by various means, for example, it can be realized by an analog circuit configuration (thermal noise generating circuit or the like), or a pseudo random code generating circuit of a digital circuit. It can also be achieved with.

According to the configuration of the second invention, in addition to the configuration of the first invention, the operation of the sound generating means is performed after the first acoustic transfer characteristic is set in the howling cancellation acoustic signal generating means. By operating and estimating the second acoustic transfer characteristic, it is possible to accurately estimate the second acoustic transfer characteristic.

Therefore, according to the first and second inventions described above, it is possible to generate the howling canceling sound with higher accuracy and prevent howling from occurring than in the conventional case.

Further, according to the structure of the third invention, in addition to the structure of the second invention, even if the first acoustic transfer characteristic of the acoustic space to be howling canceled is changed, this change is detected and this change is detected. Since the first acoustic transfer characteristic of the howling cancellation acoustic signal generation means is changed in accordance with
Howling canceling action does not easily deteriorate.

Furthermore, according to the structure of the fourth invention, in addition to the structure of the second invention or the third invention, even if the second acoustic transfer characteristic of the acoustic space to be howling canceled changes, this change also occurs. Is detected and the second acoustic transfer characteristic of the acoustic signal generation means for howling cancellation is changed according to this change, the howling cancellation effect is unlikely to deteriorate.

According to the structure of the fifth invention, in addition to the structure of the fourth invention, when a change occurs in the first acoustic transfer characteristic and the second acoustic transfer characteristic, which change occurs. By changing the acoustic transfer characteristic of any one of the howling cancellation acoustic signal generation means with priority depending on whether it is large or small, it is possible to quickly generate the howling cancellation acoustic signal with respect to the characteristic change.

[0045]

DESCRIPTION OF THE PREFERRED EMBODIMENTS A preferred embodiment of the howling canceller of the present invention will be described with reference to the drawings.

Therefore, in this embodiment, a white signal generator is provided on the noise canceling speaker side. Further, an adaptive filter for estimating the acoustic transfer function between the noise canceling speaker and the microphone in the initial state is provided. Furthermore, an adaptive filter is provided that follows a time-varying acoustic transfer function between the noise canceling speaker and the microphone. Furthermore, it comprises an adaptive filter which follows a time-varying acoustic transfer function between the noise source and the microphone. Further, a comparator for comparing the output power of the adaptive filter for time-variable tracking with the output power of the filter for initial state estimation is provided.

The above-described components will be described in detail with reference to the drawings.

FIG. 1 is a functional block diagram of the howling canceller of this embodiment. In FIG. 1, 201
Is a noise generation unit that outputs noise based on the signal generated by the noise generation source 200, and is configured by a speaker, a mechanical vibration source, or the like. And 221 is a duct for transmitting sound. And 222 is a microphone.

Reference numerals 203, 209 and 213 are adaptive filters. And 205 is a speaker for noise elimination. 207 is a white noise generator. And 208, 206, 212, 214, 219, 22
0 is a switch. And 215, 216, 218
Is an adder. 217 is a comparator.

In FIG. 1, Gao is an acoustic transfer function between the noise canceling speaker 205 and the microphone 222. Gso is an acoustic transfer function between the noise generator 201 and the microphone 222.

Further, each switch 206, 208, 21
2, 214 are control signals C1 to C from the control circuit 300.
It is controlled to be turned on or off by 4.

Next, the operation of the howling canceller will be described in detail with reference to the functional block diagram of FIG. 1 and the operation timing chart of FIG.

[Estimation of Acoustic Transfer Function Gao between Noise Eliminating Speaker 205 and Microphone 222]: In FIG. 1, in the initial state, the noise generation source 200 is stopped and the adaptive filter 213 is converged. In this case, the applied filter 213 controls the noise canceling speaker 20.
Acoustic transfer function Gao between the microphone 5 and the microphone 222
Is estimated, and this is obtained as an estimated acoustic transfer function G ^* ao. The convergence operation of the applied filter 213 will be described below.

First, the switches 206 and 214 are turned on, and the switches 212, 208, 220 and 219 are turned off (S1 in FIG. 3). In this state, the white noise signal n ^* 0 output from the white noise generator 207 is output.
Is input to the adaptive filter 213 via the switch 206 and is output from the noise elimination speaker 205. The noise signal n ^* 0 output from the noise elimination speaker 205 is input to the microphone 222 after being filtered by the acoustic transfer function Gao. The output signal e of the microphone 222 is added to the adder 21.
5 is input as an addition signal.

The adder 215 includes an adaptive filter 21.
The output signal n ^* 1 of 3 is input as the subtraction signal,
As a result, the output signal e1 of the adder 215 becomes (e−n
^* 1) Then, the output signal e1 of the adder 215
Is supplied to the adaptive filter 213 via the switch 214 and used for updating the tap coefficient.

In this case, the tap coefficient may be updated by an algorithm that uses the signal e1 which is the residual signal and the input signal of the adaptive filter 213 and is an algorithm that minimizes the residual signal. Learning identification method, LMS
(Least Mean Square) method, RLS
An algorithm such as the (Recursive Least Square) method may be used.

Then, after the convergence time T1 (FIG. 3) of the adaptive filter 213 in FIG.
14 is turned off (S2 in FIG. 3), the switch 212,
220 is turned on (S3). Then, the inverse function of the acoustic transfer function G ^* ao estimated by the adaptive filter 213 (1
/ G ^* ao) is copied to the coefficient register (not shown) of the characteristic compensation filter 202.

"Noise removing operation": After the convergence time T1 (FIG. 3) of the adaptive filter 213 elapses,
The noise source 200 is driven, which causes the noise n
Is generated from the noise generator 201. Noise generator 2
The noise n output from 01 is added to the acoustic transfer function Gso and becomes n · Gso, which is input to the microphone 222.

On the other hand, the noise n output from the noise generation source 200 is input to the characteristic compensation filter 202. In this case, the noise n is not limited to the electric signal. Therefore, when the noise n cannot be directly obtained as the electric signal, a sound collecting microphone (not shown) is installed in the vicinity of the noise generation source 200. May be obtained.

Next, since the characteristic compensation filter 202 has the characteristic of 1 / G ^* ao as described above, its output becomes n / G ^* ao, and the adaptive filter 20
Input to 3. Then, the adaptive filter 203, but generates a pseudo-noise ^{n *,} pseudo-noise ^{n *} is an input signal n ^{/ G} * ao, signal e inputted from the switch 220
4 (however, the signal e4 at this time is the signal e supplied via the switch 220). The method of generating the pseudo noise n ^* is based on the adaptive filter 213.
It is similar to the case of.

When the adaptive algorithm of the adaptive filter 203 converges, its transfer function becomes a characteristic G ^* so obtained by estimating and approximating the acoustic transfer function Gso. Therefore, the pseudo noise n ^* output by the adaptive filter 203 is expressed by the following equation (4).

N ^* = n · G ^* so / G ^* ao (4) The pseudo noise signal n ^* is supplied to the noise canceling speaker 205 after being phase-inverted by the phase inverter 204, and the duct 221. Is output inside. Duct 221
The pseudo noise n ^* output within is the acoustic transfer function Gao
Is input to the microphone 222. Therefore, the output signal e of the microphone 222 is expressed by the following equation (5).

E = n · Gso−n · G ^* so · Gao / G ^* ao = n · Gso−n · G ^* so (5) Equation (5) Then, the value of this e is almost zero. As is clear from the equation (5), it is understood that noise is removed by the convergence of the adaptive filter 203.

"Adaptive Operation for Variation of Acoustic Transfer Function": Next, the noise removing operation when the acoustic transfer function in the duct 221 varies will be described. Here, as an example, the noise elimination speaker 20 is provided between the times t2 and t3 (FIG. 3) when the adaptive filter 203 converges.
Acoustic transfer function Gao between the microphone 5 and the microphone 222
Is changed. The amount of change at this time is ΔGao, and the period from time t2 to t3 (FIG. 3) is Δt.

Then, in this embodiment, the following operation is performed at every constant period T (preferably T> T1).

First, the switches 212 and 208 are turned on (S4 in FIG. 3) and the switches 206 and 214 are turned off (S5 in FIG. 3). Next, the adaptive filter 203
Calculates the inverse characteristic of the tap coefficient and sets it as the coefficient of the fixed filter 211. Thus, the acoustic transfer function 1 / G is stored as the tap coefficient of the fixed filter 211.
It corresponds to the characteristics of so.

Further, the adaptive filter 213 determines that the time t1
By (FIG. 3), the estimation of the acoustic transfer function Gao between the noise removal speaker 205 and the microphone 222 has almost converged. Next, the noise n generated from the noise generation source 200 passes through the characteristic compensation filter 202, the adaptive filter 203, and the phase inverter 204 as described above, and the signal n
It becomes ^* and is output from the noise elimination speaker 205.

The signal n ^* output from the noise elimination speaker 205 is added to the microphone 222 after the acoustic transfer function (Gao + ΔGao at this point) between the noise elimination speaker 205 and the microphone 222 is added. Is entered. Therefore, the microphone 2
The output signal e of 22 is expressed by the following equation (6).

E = Gso · n− (G ^* so / G ^* ao) · (Gao + ΔGao) · n = Gso · n− (G ^* so · Gao · n / G ^* ao) · n−G ^* so · ΔG ao · n / G ^* ao = −G ^* so · ΔGao · n / G ^* ao (6) The output signal e of the microphone 222 is supplied to the switch 212, the adder 216, the comparator 217 and the switch 220. Is entered. Here, since the switch 214 is in the off state, the tap coefficient of the adaptive filter 213 is not updated. In this case, since the acoustic transfer function G ^* ao is already set in the adaptive filter 213 by the above-described operation, the output n ^* 1 of the adaptive filter 213 is expressed by the following equation (7).

N ^* 1 = −G ^* so · ΔGao · n · G ^* ao / G ^* ao (7) Equation (7) Then, the value of n ^* 1 is almost −G ^* so · ΔGao.
・ It becomes n. Then, the signal n ^* 1 is sent to the fixed filter 21.
Input to 1. Since the fixed filter 211 has the characteristic of 1 / Gso as described above, the output signal n ^* 2 of the fixed filter 211 is expressed by the following equation (8).

N ^* 2 = −G ^* so · ΔGao · n / G ^* so = −ΔGao · n (8) Expression (8) This signal n ^* 2 is input to the adder 218. on the other hand,
The output of the noise source 200 is input to the adaptive filter 209 via the switch 208. Adaptive filter 209
, Like the deviation at which the signal e3 and the output signal n ^{* 2} of its own output signal n ^{* 3} and the fixed filter 211 is minimized, update control the tap coefficients.

The control method of the tap coefficient in this case is the same as that of each of the adaptive filters described above. Therefore, the adaptive filter 209 estimates the change amount (change amount) ΔGao (change amount in the period Δt) of the acoustic transfer function Gao between the noise elimination speaker 205 and the microphone 222.

As a result, the tap coefficient ΔG ^* of the adaptive filter 209 becomes substantially equal to ΔGao.

Then, the adaptive filter 209 copies the tap coefficient estimated as described above to the fixed filter 210. On the other hand, the output signal n ^* of the phase inverter 204 is input to the fixed filter 210, and as a result, the output signal n ^* 4 of the fixed filter 210 is expressed by the following equation (9).

N ^* 4 = n ^* · ΔGao = −G ^* so · ΔGao / G ^* ao (9) The output signal n ^* 4 of the fixed filter 210 is input to the adder 216. On the other hand, since the output signal e of the microphone 222 is input to the adder 216, the adder 216 adds the signal n ^* 4 and the signal e.

Since the signal e and the signal n ^* 4 are expressed by the above equations (6) and (9), respectively, the output signal e2 of the adder 216 is obtained by the following equation (10). expressed.

E2 = e−n ^* 4 = −G ^* so · ΔGao · n / G ^* ao + G ^* so · ΔGao / G ^* ao (10) The value of e2 is almost zero. The output signal e2 of the adder 216 is input to the comparator 217. The comparator 217 compares the magnitudes of the signals e2 and e. In this case, the magnitude comparison may be performed by any method as long as the magnitude of the signal amplitude is reflected, such as the power or absolute value of the signal or the average of the amplitudes.

Further, the comparator 217 outputs the smaller signal to the adaptive filter 203 as the signal e4 as a result of comparing the signal magnitudes. That is, the comparator 217 determines that e> e
When it is 2, it is determined that the change in the acoustic transfer function Gso (hereinafter, referred to as ΔGso) in the period T is larger, and the switch 220 is turned on and the switch 219 is turned off, e4 = e And

On the other hand, when e2> e, the comparator 217
It is determined that ΔGao is larger, so that the switch 219 is turned on and the switch 220 is turned off to set e4 = e2. Therefore, the residual input e4 to the adaptive filter 203 has the relationship of the following expression (11).

E4 = e (when e <e2), or e4
= E2 (when e> e2) Equation (11) Then, the adaptive filter 203 uses the signal e4 to update the tap coefficient. In this case, the updating method is the same as that of the adaptive filters 209 and 213.

According to the operation described above, the switch 21
When 9 is in the ON state, ΔGao, which is the change amount (change amount) of the acoustic transfer function Gao, is removed from the residual signal e4, so the adaptive filter 203 estimates the acoustic transfer function Gso more accurately. it can. Therefore, the noise erasing characteristic is improved, and howling can be caused.

Further, when the switch 219 is in the ON state, it is judged that the change of the acoustic transfer function Gao is large, and in this case, the adaptive filter 213 determines its tap coefficient and the tap coefficient of the adaptive filter 209. To generate an acoustic transfer function Gao _NEW . This acoustic transfer function Gao _NEW is expressed by the following equation (12).

Gao _NEW (K) = Gao + ΔGao = G ₂₁₃ (K) + G ₂₀₉ (K) (12) Expression G ₂₁₃ (K) is a tap coefficient of the adaptive filter 213. Further, G _{2 09} (K) is the adaptive filter 20.
9 tap coefficient. And K is a tap coefficient order, for example, K = 0, 1, 2, 3, ...

The tap coefficient Ga represented by the above equation (12)
o _NEW (K) is the new estimated acoustic transfer function G ^* ao
It is _NEW . The characteristic compensation filter 202 generates an inverse function 1 / G ^* ao _NEW of the acoustic transfer function G ^* ao _NEW , and thereafter uses this as a new acoustic transfer function G ^* ao.

On the other hand, when the switch 220 is in the on state, the calculation of the equation (12) is not performed and the coefficient register of the characteristic compensation filter 202 is held as it is.

When the noise source 200 is stopped, the operation from the initial state described above is repeated.

(Effect of One Embodiment): According to the howling canceller of the above one embodiment, the adaptive filter 213 for estimating the acoustic transfer function Gao between the noise canceling speaker 205 and the microphone 222 in the initial state is provided. Since it is provided, the acoustic transfer function Gao can be accurately estimated even if the noise source is a non-electrical noise source such as mechanical vibration.

Further, an adaptive filter 209 for estimating the optimum tap coefficient by following the change of the acoustic transfer function Gao between the noise canceling speaker 205 and the microphone 222 in the duct 221 which changes with time is provided. so,
It is possible to suppress a decrease in howling suppression effect.

Furthermore, the noise generator 20 in the duct 221
1 and microphone 222 acoustic transfer function Gso
By including the adaptive filter 203 that estimates the optimum tap coefficient in accordance with the temporal variation of the above, it is possible to prevent howling from occurring due to the temporal variation of the acoustic transfer function Gso.

Furthermore, the output of the adder 216 to which the fixed filter 210 in which the coefficient processed by the adaptive filter 209 that follows the fluctuation of the acoustic transfer function Gao between the noise canceling speaker 205 and the microphone 222 is set is input. Even by comparing the power e2 with the output power e of the microphone 222 and optimally updating the tap coefficient of the adaptive filter 203 based on the comparison result, it is possible to prevent the suppression effect of howling from being lowered. . That is, the comparison between the output powers e and e2 is G
The magnitude of the change between ao and Gso is monitored, and the tap coefficient of the adaptive filter 203 is updated according to the magnitude of this change.

(Other Embodiments) (1) Since various embodiments can be realized within the scope of the present invention, the present invention is not limited to the embodiment described above. For example, as the configuration of the howling canceller, the number of speakers and microphones is not limited to the configuration of FIG. Further, the operation is not limited to the operation timing shown in FIG.

(2) Further, in FIG. 1 described above, howling cancellation of noise from the noise generation source 200 is described, but not limited to howling cancellation for such noise, it is also applicable to howling cancellation for other sounds. can do. For example, it can be applied to various howling cancellations such as a vibration sound of a mechanical device, a human voice, a bark of an animal, and a sound of a musical instrument.

[0093]

As described above, the structure of the howling canceller of the first invention comprises the white noise generating means, the control means, and the first acoustic transfer characteristic estimating means, and generates howling canceling acoustic signals. By causing the means to set the estimation result of the first acoustic transfer characteristic estimating means, it is possible to accurately estimate the first acoustic transfer characteristic necessary for performing howling cancellation with respect to mechanical or physical noise. it can.

In addition, the configuration of the howling canceller of the second invention includes the second acoustic transfer characteristic estimating means in addition to the configuration of the first invention, so that the second acoustic transfer characteristic is more accurate than the conventional one. Can be accurately estimated.

Further, in the structure of the howling canceller of the third invention, in addition to the structure of the second invention, the first acoustic transfer characteristic changing means is provided, and the first acoustic transfer characteristic changes in the acoustic space to be howling canceled. However, since the first acoustic transfer characteristic of the howling canceling acoustic signal generating means is changed in response to this change, the howling canceling action is unlikely to deteriorate even with such a change.

Furthermore, the structure of the howling canceller of the fourth invention comprises, in addition to the structure of the second invention or the third invention, a second acoustic transfer characteristic changing means, and a second acoustic transfer target acoustic space is provided. Even if the acoustic transfer characteristic changes, the second acoustic transfer characteristic of the acoustic signal generating means for howling cancellation is changed in response to the change, and therefore the howling canceling action can be achieved even if such a change occurs. Hard to deteriorate.

In addition to the structure of the fourth invention, the structure of the howling canceller of the fifth invention is provided with comparison control means, and the first acoustic transfer characteristic and the second acoustic transfer characteristic are provided in the acoustic space of the howling cancellation target. If any change in sound transmission characteristics that has a dominant influence on howling cancellation occurs,
It is possible to quickly perform howling cancellation in response to changes in acoustic transfer characteristics.

Therefore, according to the howling cancellers of the above-mentioned first to fifth inventions, howling cancellation with respect to mechanical or physical noise or the like can be performed more accurately than before, and the howling cancellation target is Even if the acoustic transfer characteristics of the acoustic space fluctuate, the howling canceling action is not deteriorated.

[Brief description of drawings]

FIG. 1 is a functional block diagram of a howling canceller according to an embodiment of the present invention.

FIG. 2 is a functional block diagram of a conventional howling canceller.

FIG. 3 is an operation timing chart of one embodiment.

[Explanation of symbols]

Reference numeral 200 ... Noise generating source, 201 ... Noise generating unit, 202
... Characteristic compensation filter, 203, 209, 213 ... Adaptive filter, 205 ... Noise removal speaker, 206 ... Switch, 207 ... White noise generator, 210, 211 ... Fixed filter, 215, 216, 218 ... Adder, 217
... comparator.

Claims (5)

[Claims] 1. A sound generation means for outputting a sound to a howling cancellation target acoustic space, a sound capturing means for capturing the sound propagating in the howling cancellation target space, and a howling canceling for canceling the howling of the sound. Acoustic output means for howling cancellation for outputting a sound, first acoustic transfer characteristics between the acoustic output means for howling cancellation and the acoustic capturing means in the acoustic space of the howling cancellation target, or the acoustic generating means and the acoustic capturing means A howling canceling acoustic signal generating means for generating the howling canceling acoustic signal from the second acoustic transfer characteristic between the white noise generating means for generating white noise and the acoustic generating means When operating, the white noise above Control means for outputting from the acoustic output means for howling cancellation to the acoustic space of the howling cancellation target; and output of the white noise to the acoustic space of the howling cancellation target, from the acoustic signal obtained by the acoustic capturing means to the first acoustic transmission. A first acoustic transfer characteristic estimating means for estimating a characteristic and giving the estimation result to the howling cancel acoustic signal generating means, wherein the howling cancel acoustic signal generating means is provided for the first acoustic transfer characteristic estimating means. A howling canceller characterized in that the estimation result of is set. 2. The howling canceller according to claim 1, wherein the acoustic generation means is operated after setting the estimation result of the first acoustic transfer characteristic estimation means in the howling cancellation acoustic signal generation means. Second acoustic transfer characteristic estimating means for operating the howling cancel acoustic signal generating means and the howling cancel acoustic output means to estimate the second acoustic transfer characteristic from the acoustic signal obtained by the acoustic capturing means. A howling canceller comprising: 3. The howling canceller according to claim 2, wherein, during the operation of the sound generating means, the first sound transfer characteristic estimating means estimates a sound signal from the sound generating means, and A first acoustic transfer characteristic changing means for detecting a change in the acoustic transfer characteristic of No. 1 and changing the first acoustic transfer characteristic set in the acoustic signal generating means for howling cancellation according to the change. A howling canceller characterized by that. 4. The howling canceller according to claim 2 or 3, wherein during the operation of the sound generating means, the estimation result of the second acoustic transfer characteristic estimating means and the acoustic signal of the acoustic capturing means are used. Second acoustic transfer characteristic changing means for detecting a change in the second acoustic transfer characteristic and changing the second acoustic transfer characteristic set in the howling cancellation acoustic signal generating means in accordance with the change. A howling canceller characterized by having. 5. The howling canceller according to claim 4, wherein a change in the first acoustic transfer characteristic and a change in the second acoustic transfer characteristic are compared, and the comparison result indicates that
A howling canceller comprising comparison control means for preferentially changing any of the acoustic transfer characteristics set in the acoustic signal generation means for howling cancellation.