JPH10247087A - Three-dimensional space active noise/vibration control system and its prior learning method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce the number of error connection system filters and to shorted a time for prior learning in a three-dimensional active noise/vibration control system. SOLUTION: Noise/vibration is detected by a first sensor 2 to be inputted to a first filter 3 of active noise/vibration control system 1a-1d. Its output is imparted parallel to plural sound/vibration generation means 4, and the noise/ vibration in a three-dimensional closed space is canceled. Further, an additional value of the outputs of second sensors 5 and the output of a second filter 6 are imparted to a coefficient adjustment means 7, and the coefficient of the first filter 3 is updated. When the prior learning of the second filter 6 is performed, the output of a reference signal source 8 is imparted to respective sound/vibration generation means 4 and the filter 6, and the coefficient of the filter 6 is adjusted so that differences between the detection outputs of the second sensors 5 and the output of the filter 6 are reduced. Then, after one side filter 6 is learned, the learned coefficient is copied by other side filter 6.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a three-dimensional active noise / vibration control system for generating an output for canceling noise / vibration from a speaker / vibration generator or the like and reducing the noise / vibration, and pre-learning thereof. More particularly, the present invention relates to an active noise / vibration control system capable of performing simultaneous learning of an error coupling filter in an active noise / vibration control system, and a method for pre-learning the active noise / vibration control system. The system of the present invention can be applied to an active noise control system for canceling noise, a vibration suppression control system in the construction field, an echo canceller such as a telephone, and various measurement / control systems such as image processing as a ghost countermeasure. It is.

[0002]

2. Description of the Related Art The prior art will be described below using an active noise control system as an example. FIG. 8 is a diagram showing a general configuration of an active noise control system. In the figure, Ns is a primary noise source, and the noise generated by the primary noise source Ns propagates in the direction of the arrow in the figure. SP denotes a speaker, which generates an erasing sound from the speaker SP to cancel the noise. RM is a reference microphone, which detects noise generated by the primary noise source Ns. EM is an error microphone, and error microphone E
M detects noise (error) that could not be canceled by the erasing sound generated by the speaker SP. The sounds detected by the sensor microphone RM and the error microphone EM are sent to the active noise control device Cnt.

In the active noise control device Cnt, F1 is a noise control filter (NCF) which simulates a sound transmission system {H} until the noise detected by the reference microphone RM reaches the error microphone EM ( The characteristic of the noise control filter F1 is denoted by {h}). F2 is an error coupling filter (EPF: Erro
r Path Filter), which simulates an acoustic transmission system from the speaker SP to the error microphone EM (hereinafter referred to as an error coupling system and denoted as {Hpf}) (the characteristic of the error coupling system filter F2 is denoted by {hpf}). write). AL is an adaptive algorithm that adjusts the coefficient of the noise control filter F1 based on the error detected by the error microphone EM and the output of the error coupling filter F2.
1 is controlled so that the transfer characteristic {hf} approaches the actual transfer characteristic {H}.

In FIG. 8, active noise control is performed as follows. The noise Xj generated by the primary noise source Ns is detected by the reference microphone RM. The noise Xj detected by the reference microphone RM is the active noise control device Cnt.
Of the noise control filter F1
Numeral 1 generates an erasure signal having substantially the same magnitude as the noise of the primary noise source Ns and having the opposite phase. This erase signal is output from the speaker SP
To generate an erasing sound from the speaker SP to cancel the noise. On the other hand, an error Ej is detected by the error microphone EM and sent to the adaptive algorithm AL, and the signal Xj is sent to the error coupling filter F2 (here, the transmission characteristic of the error coupling filter F2 is Δh
pf} coincides with the transfer characteristic {Hpf} of the error coupling system). The adaptive algorithm AL uses the normalized least squares method (NLMS) or the least squares method (LMS) based on the output of the error coupling filter F2 and the error Ej detected by the error microphone EM to reduce the noise Ej. The coefficient of the control filter F1 is calculated, and control is performed such that the transfer characteristic of the noise control filter F1 matches the actual transfer characteristic {H}.

In the above noise control system, since the coefficient of the noise control filter F1 is constantly adjusted during the operation of the system, the noise control filter F1 should have a transmission characteristic close to the acoustic transmission system {H} of the air duct D. Can be
The coefficients of the error coupling filter F2 are not adjusted during the operation of the system. For this reason, prior learning is performed on the error coupling system filter F2 and the filter F
2 with the transfer characteristic {H
pf}.

FIG. 9A shows the error coupling filter F2.
FIG. 6 is a diagram showing a configuration at the time of learning. Error coupling system filter F
2, a calibration generator CG that generates an M-sequence signal or a white noise signal is provided, and the M-sequence signal or the white noise signal generated by the calibration generator CG is provided as shown in FIG. This is given to the speaker SP and the error coupling system filter F2.
The speaker S by the M-sequence signal or the white noise signal
The sound generated from P is detected by the error microphone EM.
On the other hand, the M-sequence signal or the white noise signal appears at the output of the error coupling filter F2, is compared with the sound detected by the error microphone EM, and the difference is provided to the coefficient updating circuit R1. The coefficient updating circuit R1 adjusts the coefficient of the error coupling filter F2 so that the difference becomes smaller. By performing the pre-learning as described above, the transfer characteristic {Hpf} of the error coupled system and the transfer characteristic {h
pf} can be substantially matched.

[0007]

To apply the above active noise control system to a three-dimensional closed space such as a room,
A plurality of noise control devices shown in FIG. 8 are provided in accordance with the distance intervals of the erasing space, a plurality of speakers are installed on a wall or the like of the room, and the erasing sound is generated from each speaker. FIG. 10 is a diagram showing a state in which a plurality of speakers and an error microphone are installed on the wall surface as described above, and FIG.
The case where error microphones EM0 to EM3 are provided corresponding to SP3 is shown. Multiple speakers SP on each wall
When the error microphones EM0 to EM3 are provided, the erasing sounds output from the plurality of speakers SP0 to SP3 are input to the error microphones EM0 to EM3, as shown in FIG. That is, C00, C01,..., C10, C shown in FIG.
A plurality of paths are formed as shown in FIG.

Therefore, when a plurality of speakers and an error microphone are provided as described above, it is necessary to provide a number of error coupling filters corresponding to these paths. Requires N (= M ² ) error coupling filters. In addition, it is necessary to pre-learn these error coupling filters, but when a plurality of speakers and an error microphone are provided, the error microphone EM is connected to the error microphone EM as shown in FIG. Since radiation is input, it is necessary to install N pre-learning systems in parallel and learn the error coupling system filter including the influence of radiation from other speakers. To update the coefficient of the error coupling system filter, an evaluation function J is determined, and the evaluation function J
Is determined by finding the coefficients of the error coupling system filter that minimizes the number of speakers and error microphones. As the number increases, pre-learning becomes virtually impossible.

FIG. 11 is a diagram showing the time transition of the evaluation functions J1 to J4 when four speakers and four pre-learning systems are used. As shown in FIG.
Since the time when J4 becomes the minimum value differs, it is necessary to determine the coefficient so that the mean square error of the evaluation functions J1 to J4 becomes the evaluation function J and the evaluation function J becomes the minimum value in order to update the coefficient. Pre-learning takes a very long time. As described above, when the conventional active noise control device is applied to a three-dimensional confined space, and a plurality of speakers and error microphones are provided, the number of error-coupling filters increases correspondingly and the cost increases. Requires a lot of time. The present invention has been made in order to solve the above-mentioned problems of the prior art, and it is an object of the present invention to reduce the number of error coupling filters and significantly reduce the time for pre-learning. It is to provide a three-dimensional space active noise / vibration control system that can be shortened.

[0010]

As shown in FIG. 10, when a plurality of erasing speakers and an error microphone are arranged, an erasing sound enters the error microphone from a diagonal direction. A system filter is required, and much time is required for learning. Therefore, the radiated power from each erasure speaker is made sufficiently large,
The above-mentioned problem can be solved by making the incident signal from the erasing speaker disposed on the front of the error microphone sufficiently large and making the distance between each speaker and the error microphone equal.

That is, in FIG.
There are a plurality of paths such as 0, C01, ..., C10, C11, ..., but it is sufficient to make the power radiated by each erasing speaker large enough so that oblique paths can be ignored. Here, the correlation coefficient R required for learning by the error coupling system filter can be expressed by R = Gj / ΣGjx. In this equation,
Gj is the output of the convolution of the above filter, and ΣGjx is the signal incident obliquely. If the signal obliquely incident is large, the denominator is large and R is small. That is, if the output of the erasing speaker is increased to increase the self-interaction function, the influence on other systems is also increased. If it is necessary to reduce this effect, a hood may be provided at the edge of the erasing speaker to reduce a signal incident on the error microphone from an oblique direction.

With the above-described configuration, most of the signals incident on each error microphone become signals radiated from the erasing speaker disposed on the front surface thereof, and a plurality of error coupling filters are provided as described above. There is no need to do so, and the learning can be performed in a short time.
Furthermore, to eliminate noise / vibration in a three-dimensional enclosed space,
As described above, it is necessary to arrange a plurality of speakers and an error microphone on each side of the three-dimensional closed space and provide an active noise / vibration control system for each side. Equal distance, one active noise /
By learning only the error coupling filter of the vibration control system and copying the learning result to the error coupling filter of the active noise / vibration control system provided on the other side, learning can be performed in a shorter time. Becomes

The present invention has been made in view of the above points, and FIG. 1 shows a basic configuration diagram of the present invention. FIG.
In FIG. 1, noise / vibration generated by a noise / vibration source (not shown) is detected by the first sensor 2 and input to the active noise / vibration control systems 1a to 1d. In addition, a sound / vibration generating means 4 for generating an output for canceling the noise / vibration is provided around the three-dimensional enclosed space, and a second sensor 5 for detecting an error that cannot be canceled is provided on the front surface thereof. Is provided. The sound / vibration generating means 4
Is the wavelength of the upper limit frequency of noise / vibration to cancel.
A second sensor 5 for error detection is installed at an equal distance from the sound / vibration generating means on the front surface of each of the plurality of sound / vibration generating means 4. The output of each sound / vibration generating means is mainly input to a second sensor disposed on the front surface thereof.

Further, the active noise / vibration control system 1a-
1d is a first filter 3 for generating an output for canceling noise / vibration in a three-dimensional enclosed space based on an output of the first sensor 2, and a plurality of sound / vibration generating means to a second sensor. And a coefficient adjusting means 7 for adjusting the coefficient of the first filter. Then, the added value of the outputs of the second sensors is input to the coefficient adjusting means 7 of each active noise / vibration control system, and the output of the first filter 3 is output to the sound / vibration generating means 4 in parallel. .

FIG. 3B is a diagram showing a state of the second filter at the time of pre-learning. The output of the reference signal source 8 is given to each sound / vibration generating means 4 and the second filter 6 is given to the second filter 6. give. Then, the coefficient of the second filter 6 is adjusted so that the difference between the output of the second sensor 5 and the output of the second filter 6 becomes small. The learning as described above is performed on the active noise / vibration control system 1a provided on one side of the three-dimensional closed space, and the learning result (coefficient of the second filter) is applied to the other side as shown in FIG. By copying to the second filters of the provided active noise / vibration control systems 1b to 1d, the coefficients of the second filters of the active noise / vibration control systems 1b to 1d on each side can be set.

As described above, in the present invention, the above-mentioned problem is solved as follows. (1) First to detect noise generated by noise / vibration source
, A first filter to which an output of the first sensor is input and which generates an output for canceling noise / vibration in the three-dimensional enclosed space; and a canceler connected in parallel to the output of the first filter. A plurality of sound / vibration generating means arranged at an interval of a half of a wavelength of an upper limit frequency of noise / vibration, and a front surface of each of the plurality of sound / vibration generating means being equal to the sound / vibration generating means. A plurality of second sensors for error detection, which are installed at a distance, and configured so that the output of each sound / vibration generating means is mainly input to a second sensor disposed on the front surface thereof; An output of the first sensor is input, a second filter simulating transfer characteristics from the plurality of sound / vibration generating means to a second sensor, an output of the second filter, and a plurality of second filters. 2 sensors Based on the sum of outputs constituting the three-dimensional active noise / vibration control system and a coefficient adjusting means for adjusting the coefficients of the first filter.

(2) A plurality of active noise / vibration control systems are provided, and a plurality of sound / vibration generating means and a second sensor of each system are arranged around a three-dimensional enclosed space.
(3) Second of the active noise / vibration control system of the above (1)
In performing pre-learning of the filter, the reference signal is input to the second filter, and the reference signal is input to the plurality of sound / vibration generating means of the system in parallel, and the output of the second filter is output. By adjusting the coefficient of the second filter so that the difference between the sum of the output of the second sensor and the sum of the outputs of the second sensor is reduced, the second filter can be moved from the sound / vibration generating means to the second sensor. Learning to simulate transfer characteristics.

(4) A plurality of the active noise / vibration control systems of (1) are arranged around a three-dimensional closed space, and when the second filter is pre-learned, the reference signal is transmitted to the plurality of three-dimensional filters. Input to a second filter of one of the spatial active noise / vibration control systems;
The reference signal is input in parallel to the plurality of sound / vibration generating means of the system, and the second signal is output so as to reduce the difference between the output of the second filter and the added value of the output of the second sensor. By adjusting the coefficient of the filter, the second filter is learned so as to simulate the transfer characteristic from the sound / vibration generating means to the second sensor. Then, after learning the second filter, the learned filter coefficient is copied to the second filter of another system.

[0019]

Next, an embodiment of the present invention will be described. In the following embodiment, an active noise control system for canceling noise will be described. However, the present invention is applicable not only to the active noise control system but also to various measurement and control systems such as the above-described vibration suppression control system. It is possible to apply. FIG. 2 is a diagram showing a configuration of an active noise control system according to an embodiment of the present invention. In the figure, six speakers and an error microphone are arranged on each of four walls of a room, and active noise control is performed for each of the walls. The example which provided the apparatus is shown. In the figure, Cnta to Cntd are noise control devices provided corresponding to each wall surface, SP0 to SP5 are loudspeakers each generating a cancellation sound attached to each wall surface,
Error microphones EM0 to EM5 are installed on the front of each speaker.

The distance between the speakers SP0 to SP5 and the error microphones EM0 to EM5 is equidistant as shown in FIG. 3, and the interval between the error microphones EM0 to EM5 is set to a half wavelength λ / 2 of the upper limit frequency to be erased. Is set. The outputs of the error microphones EM0 to EM5 are added for each wall surface, and the outputs * 1 to * 4 are added to the noise control devices C
nta to Cntd. The outputs of the noise control devices Cnta to Cntd are output to the speakers SP0 to SP5 installed on the respective walls. In addition, a reference microphone RM for detecting noise generated by the noise source is provided, and outputs * A to * D of the reference microphone RM.
Is input to each of the noise control devices Cnta to Cntd.

FIG. 4 is a diagram showing one of the systems of the active noise control system shown in FIG. 2. As shown in FIG. 4, the output gi of the noise control filter F1 is controlled by the amplifiers AS0 to AS6.
And is input to the speakers SP0 to SP5. The outputs of the error microphones EM0 to EM5 are
Active noise control device C
nt adaptive algorithm AL. Here, as described above, there are a plurality of paths between each of the speakers SP0 to SP5 and the error microphones EM0 to EM5, but in the present embodiment, among these paths, diagonal paths can be ignored and mainly System microphone sound is error microphone EM0
To EM5, the gains of the amplifiers AS0 to AS6 that amplify the input signal to the speaker are set sufficiently large. In addition, the error microphone EM
As described above, a hood may be provided for each of the speakers SP0 to SP5 in order to further reduce the influence of the signals incident on the speakers 0 to EM5.

In FIGS. 2 and 4, the active noise control is performed as follows. The reference microphone RM detects noise generated by a noise source (not shown). The noise detected by the reference microphone RM is the active noise control device Cnt.
The noise control filter F1 is sent to the noise control filters F1 to Cntd, and the noise control filter F1 generates an erasure signal having substantially the same magnitude and opposite phase as the noise of the primary noise source Ns. This erase signal is sent to speakers SP0 to SP5, and speakers SP0 to SP5
An erasing sound is generated from P5 to cancel the noise.

On the other hand, an error is detected by the error microphones EM0 to EM5, and the sum thereof is calculated by the adaptive algorithm AL.
And the output of the reference microphone RM is sent to the error coupling system filter F2 (here, it is assumed that the error coupling system filter F2 has been learned). The adaptive algorithm AL is based on the output of the error coupling filter F2 and the errors detected by the error microphones EM0 to EM5, and is based on the normalized least squares method (NLMS) or the least squares method (LMS).
And a noise control filter F for reducing the error.
A coefficient of 1 is calculated, and control is performed so that the transfer characteristic of the noise control filter F1 matches the actual transfer characteristic {H}.

FIG. 5 is a diagram showing the basic structure of a digital filter used as the noise control filter F1 and the error coupling filter F2. The digital filter used in this embodiment is as shown in FIG. A non-recursive digital filter (FIR) composed of multipliers Mx0 to Mxm of filter coefficients a0 to aM and delay means TD0 to TDm of time T can be used. The output yk of the non-recursive digital filter is generally the non-recursive M filter shown in FIG.
Described by the following (the number of taps: equal to the number M + 1 of coefficients {am}) constant coefficient linear difference equation, and when a unit impulse is given, an impulse response having a peak value corresponding to the coefficient value is shown.

FIG. 6 is a diagram showing the configuration of the error-coupling filter F2 in this embodiment at the time of pre-learning. FIG. 6 shows one of the active noise control systems Cnta to Cntd shown in FIG. The case of learning is shown. In this embodiment, in order to perform the pre-learning of the error coupling system filter F2, as described above, the calibration generator CG that generates the M-sequence signal or the white noise signal is used.
To provide the M-sequence signal or the white noise signal Xj generated by the calibration generator CG to the speakers SP0 to SP5 and the error coupling filter F2 provided on one side of the three-dimensional closed space. Sounds generated from the speakers SP0 to SP5 due to the M-sequence signal or the white noise signal are detected by the error microphones EM0 to EM5. On the other hand, the output Gj of the error coupling system filter F2 is compared with the sum of the outputs of the error microphones EM0 to EM5, and the difference Ej is given to the coefficient updating circuit R1. The coefficient updating circuit R1 adjusts the coefficient of the error coupling filter F2 so that the difference Ej becomes small.

The following equations (1), (2), and (3) are the equation for updating the coefficient of the error coupling system filter.
(M) and Hw + 1 (m) are the coefficients of the m and m + 1 taps of the error coupling filter, K is the step gain, and Gj, Ej, and Xj are the error coupling filters shown in FIG. The output of F2, the difference between Xj and Gj, and the input of error microphones EM0 to EM5. Note that Pw (m) is w
It represents the power at the m-th tap at the time.

[0027]

(Equation 1)

In the present embodiment, at the time of pre-learning, the step gain K in the above-mentioned coefficient updating equation (1) is changed according to the correlation coefficient between the speaker and the error microphone. That is, when the correlation coefficient increases, the step gain K decreases, and when the correlation coefficient decreases, the step gain increases. This makes it possible to perform pre-learning of the filter without being significantly affected by the obliquely incident signal. As described above, the active noise control system Cnta
After learning the error coupling system filter F2, the coefficients of the learned error coupling system filter F2 are replaced with the error coupling system filters Fnt of the active noise control systems Cntb to Cntd on the other sides.
Copy to 2. That is, since the distance between the speaker and the error microphone on each side is equal as described above, the coefficient of the error coupling filter F2 provided on each side is equal. Therefore, one active noise control system Cnta
Can be copied to the active noise control systems Cntb to Cntd on the other sides.

FIG. 7 is a diagram showing a sequence at the time of starting up the system of this embodiment. In the present embodiment, as shown in the figure, after the system power is turned on, a coefficient is determined by performing pre-learning of the error coupling filter F2 on one side, and the coefficient is provided on the other 2 to 4 sides. Copy to error coupling filter F2 of active noise control system. Then, the erase operation of each side is started.

[0030]

As described above, in the present invention,
The following effects can be obtained. (1) In a three-dimensional active noise / vibration control system, the number of error coupling filters can be significantly reduced. Therefore, the configuration of the system can be simplified. (2) Since the number of error-coupling filters can be reduced, the time required for pre-learning can be greatly reduced, and the start-up of the system can be speeded up. Further, after learning the error coupling system filter of the active noise / vibration control system provided on one side of the three-dimensional closed space, the coefficient is applied to the error coupling system filter of the active noise / vibration control system provided on the other side. The system can be started up in a shorter period of time by copying to the system.

[Brief description of the drawings]

FIG. 1 is a basic configuration diagram of the present invention.

FIG. 2 is a diagram illustrating a configuration of a three-dimensional spatial active noise control system according to an embodiment of the present invention.

FIG. 3 is a diagram showing an arrangement of each speaker and an error microphone in the embodiment.

FIG. 4 is a diagram showing a configuration of one of the active noise control systems of FIG. 2;

FIG. 5 is a diagram showing a basic structure of a digital filter used in the present embodiment.

FIG. 6 is a diagram illustrating a configuration at the time of pre-learning of the error coupling system filter in the present embodiment.

FIG. 7 is a diagram showing a sequence when starting up the system of the present embodiment.

FIG. 8 is a diagram showing a general configuration of an active noise control system.

FIG. 9 is a diagram showing a configuration of the error coupling system filter shown in FIG. 8 at the time of learning.

FIG. 10 is a diagram showing a state where a plurality of speakers and an error microphone are installed on a wall surface.

FIG. 11 is a diagram showing a time transition of evaluation functions J1 to J4.

[Explanation of symbols]

1a to 1d Active noise / vibration control system 2 First sensor 3 First filter 4 Sound / vibration generating means 5 Second sensor 6 Error coupling system filter 7 Coefficient adjusting means 8 Reference signal source SP0 to SP5 Speaker RM Reference microphone EM0 to EM5 Error microphone Cta to Cntd Active noise control system F1 Noise control filter F2 Error coupling system filter AS0 to AS6 Amplifier AE0 to AE5 Amplifier AL Adaptive algorithm CG Calibration generator

Claims (4)

[Claims] A first sensor for detecting noise generated by a noise / vibration source; an output of the first sensor being input to generate an output for canceling noise / vibration in a three-dimensional enclosed space. A first filter; a plurality of sound / vibration generating means connected in parallel to an output of the first filter and arranged at a half interval of a wavelength of an upper limit frequency of noise / vibration to be canceled; The sound / vibration generating means is installed at the same distance from the sound / vibration generating means, and the output of each sound / vibration generating means is mainly input to the second sensor disposed on the front of the sound / vibration generating means. A plurality of second sensors for error detection, configured to receive the output of the first sensor,
A second filter simulating a transfer characteristic from the vibration generating means to the second sensor; an output of the second filter; and an output of the first filter based on a sum of outputs of the plurality of second sensors. An active noise / vibration control system for a three-dimensional space, comprising: a coefficient adjusting means for adjusting a coefficient. 2. The active noise / vibration control system according to claim 1, wherein a plurality of sound / vibration generating means and a second sensor of each system are arranged around a three-dimensional enclosed space. Active noise / vibration control system in three-dimensional space. 3. A first sensor for detecting noise / vibration generated by a noise / vibration source, and an output of the first sensor is input to generate an output for canceling noise in a three-dimensional enclosed space. A first filter; a plurality of sound / vibration generating means connected in parallel to an output of the first filter and arranged at a half interval of a wavelength of an upper limit frequency of noise / vibration to be canceled; The sound / vibration generating means is installed at the same distance from the sound / vibration generating means, and the output of each sound / vibration generating means is mainly input to the second sensor disposed on the front of the sound / vibration generating means. A plurality of second sensors configured to detect an error, an output of a second filter to which an output of the first sensor is input, and a sum of outputs of the plurality of second sensors. Based on the first A pre-learning method for an active noise / vibration control system in a three-dimensional space, comprising: a coefficient adjusting means for adjusting a coefficient of a filter; And the reference signal is input in parallel to the plurality of sound / vibration generating means of the system, so that the difference between the output of the second filter and the sum of the outputs of the second sensor is reduced. Adjusting the coefficient of the second filter to make the second filter learn so as to simulate the transfer characteristic from the sound / vibration generating means to the second sensor. Prior learning method of active noise / vibration control system. 4. A first sensor for detecting noise / vibration generated by a noise / vibration source, and an output for receiving an output of the first sensor and canceling noise / vibration in a three-dimensional enclosed space. And a plurality of sound / vibration generating means connected in parallel to the output of the first filter and arranged at a half interval of the wavelength of the upper limit frequency of the noise / vibration to be canceled. A second sensor disposed on the front surface of each of the plurality of sound / vibration generating units at an equal distance from the sound / vibration generating unit, and the output of each sound / vibration generating unit is disposed mainly on the front surface thereof A plurality of second sensors for error detection configured to be input to the second filter, an added value of the outputs of the plurality of second sensors, and a second filter to which the output of the first sensor is input Based on the output of A plurality of active noise / vibration control systems each including a coefficient adjusting means for adjusting a coefficient of the first filter are arranged around a three-dimensional enclosed space. Inputting the reference signal to the second filter of one of the active noise / vibration control systems in a plurality of three-dimensional spaces, and inputting the reference signal to the plurality of sound / vibration generating means of the system in parallel; By adjusting the coefficient of the second filter so that the difference between the sum of the output of the second filter and the output of the second sensor is reduced, the second filter is controlled by the sound / vibration generating means. Learning to simulate transfer characteristics to a second sensor, learning the second filter, and then copying the learned filter coefficient to a second filter of another system. Learning method for active noise / vibration control system in three-dimensional space.