JP3439245B2 - Noise cancellation system - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [0001] BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a noise canceling system.
Especially at observation points originating from multiple noise sources.
Noise cancellation system to cancel noise
I do. [0002] 2. Description of the Related Art As a countermeasure against noise, a sound absorbing material is conventionally used.
(Passive control) is known. But sucking
In the method using sound material, a quiet area with low noise is formed.
Is difficult to perform and bass cannot be effectively eliminated.
There is a title. In particular, to prevent noise in the cabin of a car
Increases the weight of the car and effectively reduces noise.
There is a problem that can not be done. As a result, the noise capa
Method to reduce noise by emitting cancel sound from speaker
(Active control) is in the limelight, such as factories and offices
Is being put to practical use in some indoor spaces. In addition,
Active control reduces noise even in the passenger compartment
A scheme has been proposed. FIG. 4 shows a conventional system for canceling engine sound.
It is a lineblock diagram of a noise canceling device. 11 is a noise source
Engine, 12 is the engine speed for detecting the engine speed R
The sensor 13 has a frequency corresponding to the engine speed R.
A sinusoidal signal having a constant amplitude as a reference signal xn
It is a reference signal generator. If the noise source is an engine,
The noise generated by gin rotation has periodicity,
The wave number depends on the engine speed. For example, 4-cylinder engine
In the case of gin, the periodic noise generated in the cabin
The second harmonic of the rotation speed is dominant and the rotation speed is 600r
pm (10 rpm), noise around the vehicle
The wave number is 20 Hz, and the rotation speed is 6000 rpm (100 rpm).
In the case of s), the frequency of the noise generated in the vehicle interior is 200H
z. The reference signal generator 13 is a sine wave of the second harmonic
Store the data in ROM and save the data as needed.
Generate reference signal xn by reading and outputting
I do. The read / output timing of this data is
It is controlled according to the engine speed R,
Frequency of periodic noise generated according to engine speed R
A reference signal having a number is output. [0004] 14 is a noise canceling controller.
The reference signal xn generated from the reference signal generator 13 is input.
The noise canceling position (observation point)
And the noise Sn near the driver's ear, for example).
A synthesized sound signal of the cancel sound Sc is input as an error signal en.
Adaptive signal processing so that the error signal is minimized.
And outputs a noise cancellation signal yn. Noise can
The cell controller 14 includes an adaptive signal processing unit 14a,
An adaptive filter 14b having a digital filter configuration and a reference signal
Cancellation from speaker to noise cancellation point at xn
Filter by convolving the propagation characteristics (transfer function) of the sound propagation system
To generate a code X signal (reference signal for signal processing) rn.
It has a filter 14c for creating a Lutard X signal. 1
5 is an adaptive filter output (noise cancel signal yn).
DA converter to convert to analog noise cancellation signal
And 16 are power amplifiers for amplifying the noise canceling signal.
And 17 are cancellations that emit a noise cancellation sound Sc.
The speaker 18 is located at the noise canceling point, and the noise S
n and the cancel sound Sc are detected, and the synthesized sound signal is
An error microphone which outputs the error signal en
-Amplifier to amplify signal en, 20 removes alias
The low-pass filter 21 outputs the low-pass filter output.
This is an AD converter that converts digital data. [0005] The adaptive signal processing section 14a has a noise canceling point.
Input through the filter 14c and the error signal en at
The signal processing reference signal (filtered X signal) rn
Input and use these signals at the noise cancellation point.
Adaptive signal processing to cancel noise
The coefficient of the adaptive filter 14b is determined. For example, adaptive signal processing
The processing unit 14a is a well-known filtered X LMS (Least Mean
 Square) according to the adaptive algorithm.
Adapted so that the error signal en input from
The coefficients of the filter 14b are determined. Adaptive filter 14b
According to the coefficient determined by the adaptive signal processing unit 14a
Digital filter processing is applied to the reference signal xn to generate noise
The cancel signal yn is output to cancel the noise. still,
The reference signal xn is a signal having a high correlation with the noise Sn to be eliminated.
Sound that has no correlation with the reference signal must be eliminated.
No. [0006] As shown in FIG.
It is composed of an FIR type digital filter.
Delay elements DL and D that sequentially delay signals by one sampling time
L and a coefficient w for each delay element output₁(n), w_Two(n), w
_Three(n) ・ ・ ・ w_Nmultiplying units ML, ML,...
, And adders AD, AD,.
・ It is realized by That is, the reference at the current time n · Ts
The illumination signal is xn, and the coefficient of each multiplier at that time is w₁(n), w
_Two(n), w_Three(n) ・ ・ ・ w_N(n), output (noise cancellation signal)
Is yn, the adaptive filter 14b becomes   yn = Σwi (n) · x (n−i + 1) (i = 1 to N) ·· (1) And outputs a noise cancellation signal yn. [0007] Filter 14c for creating a filtered X signal
Consists of FIR digital filters as shown in Fig.6
For example, the input signal is sequentially delayed by one sampling time.
And the delay elements DL, DL.
Number c₁, C_Two, C_Three... c_MMultiplying units ML, ML,.
.. and adders AD, AD for sequentially adding the outputs of the respective multipliers
... Coefficient c₁, C_Two, C_Three... c_MIs the secondary sound
The propagation characteristics of the propagation system (system from the speaker to the observation point) are simulated.
It has been decided to imitate. Participation at time n · Ts
Let xn be the illumination signal and r (n) be the output (filtered X signal).
If the filter 14c is   r (n) = ΣciΣx (n-i + 1) (i = 1 to M) And outputs a filtered X signal r (n). The adaptive signal processing unit 14a performs one sampling.
Adaptive filter at next time (n + 1) · Ts after time Ts
14b coefficient w₁(n + 1), w_Two(n + 1), w_Three(n + 1) ・ ・ ・ w
_N(n + 1) is calculated by using the coefficient and the error signal en at the current time n · T.
And the filtered X signal rn, the following equation (coefficient update equation) w_j(n + 1) = w_j(n) + μr (n-j + 1) en (3) (Where j = 1, 2,... N). (3)
In the equation, (n) is the value of the current sampling time, and (n + 1) is 1
Value after sampling time, (n-1) is one sampling time
The previous value, (n-2) means the value two sampling times earlier, ...
are doing. Μ is a step for updating the coefficients of the adaptive filter.
Constants (step size parameters)
And appropriate value according to the noise cancellation system
Is set to The above is the harmonic components of the engine speed, for example
In the case where the second harmonic component is canceled,
In the vehicle, road noise is generated in addition to the engine noise.
Road noise is caused by tire → suspension due to uneven road surface
→ The entire vehicle body shakes via the path of the suspension
It is generated by moving and propagating into the passenger compartment. Such b
To cancel the input noise by adaptive signal processing
Is a coherent reference to the road noise in the cabin.
Light signals need to be obtained. By the way, road noise is
It does not have periodicity like harmonic noise of engine speed,
Moreover, it is difficult to uniquely identify the noise source.
For this reason, a plurality of acceleration sensors
And load the output of each acceleration sensor
This is a reference signal for cancel. And for each reference signal
Noise canceling controller in each noise canceling
The controller uses a predetermined reference signal and synthesized sound signal
Performs adaptive signal processing and outputs a noise cancellation signal.
Combine the noise cancellation signal and input it to the speaker for noise
A cancel sound is generated. FIG. 7 shows a conventional road noise noise canceller.
FIG. 1 is a configuration diagram of a system. In the figure, 31a, 31b, 3
1c is a signal x corresponding to road noise detected by detecting vehicle vibration.
₁n, x_Twon, x_ThreeThis is an acceleration sensor that generates n. Road
The noise is caused by the tire → suspension →
The entire body vibrates through the path of the spence support,
It is generated by propagating into the passenger compartment. Such Rhodeno
The noise source in Izu cannot be uniquely identified. this
Therefore, the plurality of acceleration sensors 31a to 31c are
Attached to each part of the car support section to detect vehicle vibration as a whole
Like that. 32a, 32b, 32c are acceleration sensors
The output signals of₁n, x_Twon, x_Threeas n
The noise is canceled according to the LMS adaptive algorithm
Adaptive signal processor (noise processing) that performs cell processing (adaptive signal processing)
Cancel controller). Each adaptive signal processing unit 3
2a to 32c have the same configuration, and have a conventional noise control.
The adaptive signal processing is performed similarly to the cancel controller (see FIG. 4).
Control unit LMS, adaptive filter ADF, filtered X signal
It has a filter FXF for creation. Each adaptive signal processing
The unit LMS outputs an error signal (noise at the noise canceling point).
Composite sound signal of cancellation sound and noise) en and filter FX
Reference signal for signal processing output from F (filtered X
Signal) r₁n, r_Twon, r_Threen, and use these signals
To cancel the noise at the noise cancellation point
Adaptive signal processing to determine the coefficients of the adaptive filter ADF
Set. Each adaptive filter ADF has an adaptive signal processing unit LMS
Signal x according to the coefficient determined by₁n, x_Twon, x
_Threenoise canceling signal by applying digital filter processing to n
y₁n, y_Twon, y_ThreeOutputs n. Reference numeral 33 denotes each of the noise canceling processing sections 32a to 32a.
Noise cancel signal y output from 2c₁n, y_Twon, y
_ThreeNoise to combine n and cancel road noise
The adder outputs a cancel signal yn. 34 is noisy
Sound cancellation signal yn is converted to analog noise cancellation signal
DA converter (DAC) for converting to noise
A power amplifier for amplifying the cancel signal;
Cancel speaker that emits cell sound Sc, speaker
Cancel sound output from the secondary sound propagation system (cancel
A noise canceling point is reached via a sound propagation system 37. 3
8 is located at the noise cancellation point (observation point), and the noise Sn
And a cancel sound Sc, and a synthesized sound signal is detected.
An error microphone which outputs the error signal en
-A microphone amplifier that amplifies the signal en, 39 is a microphone amplifier
Converter (ADC) for converting loop output to digital
It is. [0013] SUMMARY OF THE INVENTION Reduction in adaptive signal processing
A signal that correlates with the target noise (for example, road noise)
If the coherent reference signal xn is not generated, the noise
Sound cannot be canceled effectively. In other words, load
Noise suppression performance is determined by the output of each acceleration sensor (reference signal)
Of multiple coherence with noise (road noise)
Depends on size. Therefore, conventionally, multiple coherent
Increase the number of acceleration sensors to secure
are doing. Note that multiple coherence refers to each sensor.
Calculated from the coherence between power and noise
It is a multiple association function. However, as the number of sensors increases
A noise canceling processing unit is required accordingly, and
Problems that cause large size and cost increase in conventional systems
was there. From the above, the object of the present invention is to reduce the number of sensors
Noise reduction processing unit effectively reduces road noise
By providing a noise cancellation system that can mute
is there. Another object of the present invention is to reduce the number of noise canceling processing units.
System is smaller than the number of sensors,
Noise cancellation system that can prevent mold and cost increase
System. [0014] FIG. 1 illustrates the principle of the present invention.
FIG. 101a to 101c are n (n> m) noises
The detection sensors 102a to 102b are n sensor detections.
Output signals and create m uncorrelated signals
Conversion filters, 103a to 103b are from conversion filters
Adaptive signals are used for each of the output m signals as reference signals.
M adaptive signal processing units for performing signal processing
The output signal of the signal processing unit is combined to generate a noise cancellation signal.
Adder 107 for inputting the noise cancel signal
Sound generator that outputs noise canceling sound
Peaker) and 109 are noise Sn and noise capacities at the observation point.
Detects the synthesized sound with the cancel sound Sc and sends it to the adaptive signal processing unit.
This is a noise detection unit that performs feedback. [0015] The function has n measured inputs, but has m measured inputs.
Systems that have only transparent noise sources
More m mutually incoherent virtual inputs
It can be a gathering of power. So, m independent
N (n> m) noise detection to detect the noise of the noise source
Output sensors 101a to 101c, and the n sensors are provided.
Input the output signal to the conversion filters 102a to 102b
And generating m uncorrelated signals from the conversion filter,
Each of the m signals output from the conversion filter is referenced.
M adaptive signal processing units 103a to 103b as illumination signals
Performs adaptive signal processing on the output signals of the m adaptive signal processing units.
Signal into the speaker 107
Force to cancel the noise at the observation point.
Outputs cancel sound. In this way, the number of sensors
A small number of noise keys, in other words equal to the number of independent noise sources
Noise generated from the noise source by the cancel processing unit (for example,
Road noise) can be effectively silenced,
And cost increase can be prevented. [0016] 【Example】 (a) Principle of the present invention First, the power spectrum of the signal will be briefly described.
You. xi (t) is a time-series input signal (the number of inputs is n), hi (t)
Is a time series weighting function, x (f) = [X₁(t), x_Two(t), ..., x_n(t)] (4) h (t) = [h₁(t), h_Two(t), ..., h_n(t)] (5) Can be expressed as Fourier transform of equations (4) and (5) gives X (f) = [X₁(f), X_Two(f), ..., X_n(f)] (4) ′ H (f) = [H₁(f), H_Two(f), ..., H_n(f)] (5) ′ Given as Where the mutual spectrum between the input signals
The density function Sxx (f) is expressed by the following equation. [0017] (Equation 1) Where S₁₁, S_{twenty two}, ... S_nnIs the power spectrum
And the others represent the mutual spectrum. Between each input signal
If there is no relevance, mutual
The rams are all zero, and Sxx (f) is a diagonal matrix. As a real system, when driving a car
The road noise generated in the cabin as an example
Looking at the road noise, depending on the condition of the road surface,
Vibrations are generated, and the vibrations become noise sources
Is propagated to Here, adaptive algorithms such as LMS
To realize a noise cancellation system using a car
Need to obtain coherent reference signals with noise in the room
There is. The noise sources of road noise are suspension and body
The noise is transmitted to the cabin through the
It is difficult to specify
Paste to each part of the pension, its output and vehicle interior noise
Noise cancellation by evaluating multiple coherence
An optimal reference signal for the stem will be obtained. At this time,
The cross spectrum between the sensor outputs is the n × n row of equation (6).
Can be represented by columns. Here, n is the number of sensors. Now, assuming that the number of independent noise sources is m,
If the number of sensors is at least m, all noise sources can be detected.
It is possible to install sensors from noise sources in reality.
Has a predetermined transmission characteristic between
The number of sensors is n, because there are zero points in the characteristics
(N> m). Sensor with such a configuration
The cross spectrum of the signal obtained from the
In this way, it is represented by an n × n matrix.
You. S.M.Price and R.J.Bernhard stated, "n measured inputs.
Powerful but with only m incoherent noise sources
The system without m is incoherent by the linear transformation.
Can be expressed by a collection of virtual inputs. "
(S.M.Price and R.J.Bernhard. "Virtual
Coherence: A Digital Signal ProcessingTechnique fo
r Incoherent Source Identification; Proceedings of4
th Inter-national Modal Analysis Conference, page
1256-1262, 1986). Therefore, the above system (m
Independent noise source, a system with n noise detection sensors
To the series of processes described in this document.
Therefore, m virtual coherences (incohere
Virtual input of the component). In practice, n
Virtual coherences are calculated, but mn
Has almost zero virtual coherence, and the output
It has been confirmed that it has no effect. Further, m virtual inputs which affect the output terminal
Sii 'is Sxx '= P^HSxxP (7) Required by (Equation 2)Is described in the above document. Equation (7)
Where P is a unitary matrix with the eigenvectors of Sxx as columns
And the virtual input Sii 'is the diagonal element in equation (8).
Given. From the above, virtual coherence using Sxx '
N that affect the output end by finding the source function
Can be specified. Soshi
Of (n−m) virtual inputs out of n virtual inputs Sii ′
Since the force is almost zero, m noise sources and n sensors
In the case of one system, it affects the output end (observation point in the cabin)
Can be set to m. m
Once the virtual inputs are obtained, m reference signals
Adaptive signal processing is performed by the adaptive signal processing unit.
Synthesizes the output signals of the sections to create a noise cancellation signal,
Input the noise cancellation signal to the speaker and send it to the observation point.
Noise cancel sound to cancel the noise
Power. (B) Outline of the present invention FIG. 1 is a schematic explanatory diagram of the present invention. First, shown in equation (4)
Input x (t) = [x₁(t), x_Two(t), ... x_n(t)]
D conversion and X (f) = [X₁(f), X_Two(f), ... X_n(f)]
(Step 1). Next, the mutual spectrum between the input signals
The vector density function Sxx (f) is obtained by equation (6) (step
2). If the cross-spectral density function Sxx (f) is found, (7)
The diagonalized cross spectrum Svii is obtained from the equation (step
3). From this diagonalized cross spectrum Svii
Calculate the virtual coherence function and calculate the number of independent virtual inputs m
(Step 4). Note that the number of independent virtual inputs is zero.
This is the number of virtual inputs Svii. Thereafter, the actual n inputs (sensor outputs)
Transform filter h that produces m virtual inputs from₁,
h_Two, ... h_mIs determined (step 5). For example, x
₁(t), x_Two(t), x_ThreeVirtual coherence for 3 inputs of (t)
As a result, it is found that there are two independent virtual inputs.
Suppose you have been rejected. At this time, x₁(t), x_Two(t), x_Three(t) 3
One output from input y₁The conversion filter for obtaining (t) is h₁
(The conversion filter can be either FIR type or IIR type.
Although good, the IIR type is excellent for simplifying the structure).
Conversion filter h₁Output y₁Fourier transform (t)
-Spectrum Y₁Find (f) and S₁And This S₁Is virtual
Input Sv₁₁H so that₁Is determined, x
₁(t), x_Two(t), x_ThreeConvert the three inputs of (t) to the first virtual input
First conversion filter h for₁Is determined.
Similarly, S_TwoIs the virtual input Sv_{twenty two}H so that_TwoTo
If you decide, x₁(t), x_Two(t), x_Three(t) is input to the second temporary
Second conversion filter h for converting to a virtual input_TwoIs found
It will be. This conversion filter h₁, H_TwoIn real time
There is no need to find each time, and a fixed filter determined in advance
It can be. As described above, if the conversion filter is obtained,
The noise cancellation system is configured as shown in
M adaptations with each of the permutation filter outputs as reference signals
Adaptive signal processing is performed by the signal processing unit, and the
The output signals are combined to create a noise cancellation signal, and the noise
Input the sound cancellation signal to the speaker and
Output cancel sound to cancel noise
(Step 6). Note that a method for calculating the coefficient of the conversion filter h
There are various methods. The first approach is to use h as an adaptive filter
And make adaptive operation so that Si-Svii → 0.
This is a method for obtaining coefficients. The second approach uses adaptive action
Instead, it uses the well-known linear random search method.
You. This is done by randomly changing the coefficients of the transform filter
Find Si-Svii sequentially, and look for coefficients that make Si-Svii → 0.
It is used to identify the system using a computer.
It is a technique that has been used. (C) Noise cancellation system of the present invention ·Constitution FIG. 2 is a configuration diagram of the noise cancellation system of the present invention.
You. In the figure, 101a, 101b and 101c are from the noise source.
Signal x corresponding to output noise₁n, x_Twon, x_Threegenerates n
Sensors (for example, detecting vehicle body vibration to reduce road noise)
(An acceleration sensor that generates a corresponding signal). 102
a and 102b are sensor detection signals x₁n, x_Twon, x_ThreeEnter n
Input, first and second incoherent virtual inputs
Force x₁n ', x _Twoa conversion filter that outputs n ', 103a,
103b Output signal x of conversion filters 102a and 102b
₁n ', x_Twon ′ are input as reference signals, respectively, and LM
Noise cancellation processing (adaptive
(Signal processing). Each adaptive signal processing
The processing units 103a to 103b have the same configuration.
Adaptive signal processing similar to conventional noise cancellation controllers
Section LMS, adaptive filter ADF, filtered X signal
It is provided with a forming filter FXF. Each adaptive signal processing unit
The LMS and the error signal en at the noise cancellation point
Signal processing reference signal (filter signal) output from the filter FXF.
Lutard X signal) r₁n, r_Twon and input these signals
To cancel the noise at the noise cancellation point
The adaptive filter ADF performs the adaptive signal processing as
To determine. Each adaptive filter ADF has an adaptive signal processing unit L
The reference signal x according to the coefficient determined by the MS₁n ', x
_Twon ′ is subjected to digital filter processing to
No. y₁n, y_TwoOutputs n. Reference numeral 104 denotes each noise canceling processing unit 103a
Noise cancel signal y output from to 103b₁n, y
_TwoCombine n to cancel noise (eg road noise)
Adder that outputs the noise cancellation signal yn
is there. Reference numeral 105 denotes a noise canceling signal yn for analog noise.
DA converter (DA
C) and 106 are powers for amplifying the noise canceling signal.
The pump 107 emits a noise canceling sound Sc.
Cancel sound output from the speaker at the cell speaker
Is via a secondary sound propagation system (cancellation sound propagation system) 108
The noise cancellation point is reached. 109 is the noise cancellation point
(Observation point), the sum of noise Sn and cancel sound Sc
Detects synthesized sound and outputs synthesized sound signal as error signal en
Error microphone 110 amplifies the error signal en
Microphone amplifier 111 digitalizes microphone amplifier output
An AD converter (ADC) for conversion. Operation The noise of each noise source is detected by the sensors 101a to 101c.
Output and detection signal x₁n, x_Twon, x_ThreeConvert each to n
Input to the data 102a and 102b. Conversion filter 102
a and 102b are the sensor detection signals x₁n, x_Twon, x
_Threen is subjected to a predetermined filter processing to generate a virtual input x₁n ', x
_TwoCreate n 'and output. Adaptive signal processing section 103a, 1
03b is the output signal of the conversion filters 102a and 102b
(Virtual input) x₁n ', x_Twon ′ as reference signals
Noise cancellation processing according to LMS adaptive algorithm
(Adaptive signal processing) and the noise cancel signal y₁n, y
_TwoOutputs n. The adder 104 is a noise canceling processor
Noise cancellation signals output from 103a and 103b
y₁n, y_Twonoise (for example, road noise)
Outputs noise cancel signal yn for canceling
You. The DA converter 105 outputs the noise cancel signal yn
Convert to analog noise cancellation signal, power amplifier
106 amplifies the noise canceling signal and a speaker 107
And the noise canceling sound Sc is radiated from the speaker.
I do. The cancellation sound output from the speaker is 2
Noise through the secondary sound propagation system (cancellation sound propagation system) 108
At the cancellation point (observation point), the noise at the observation point
Cancel. The error microphone 109 controls the noise Sn and the noise.
The synthesized sound of the cancel sound Sc is detected, and the synthesized sound
No. en via a microphone amplifier and AD converter
Feedback is provided to each adaptive signal processing unit. After that,
The operation is repeated and finally the noise at the observation point is considerable
Canceled. The present invention has been described with reference to the embodiments.
However, the present invention follows the gist of the present invention described in the claims.
Various modifications are possible, and the present invention excludes these.
Not something. [0029] According to the present invention, m independent noises are provided.
N (n> m) noise detection cells to detect the source noise
A sensor for converting the n sensor output signals.
M signals that are uncorrelated from the conversion filter
And m signals output from the conversion filter are
Each of the m adaptive signal processing sections serves as an adaptive signal as a reference signal.
Signal processing and synthesizes the output signals of m adaptive signal processing units
Input the noise cancellation signal to the speaker
Noise canceling sound to cancel noise
Output so that the number of sensors
Noise generated by the noise source by the noise canceling processing unit
(For example, road noise)
Can be prevented from increasing in size and cost. Also, in the present invention
According to this, since there is no correlation between reference signals,
The convergence speed in response signal processing has been improved,
Can be planned.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a diagram illustrating the principle of the present invention. FIG. 2 is a schematic explanatory diagram of the present invention. FIG. 3 is a configuration diagram of a noise cancellation system of the present invention. FIG. 4 is a configuration diagram of a conventional noise canceling device (in the case of canceling periodic noise). FIG. 5 is a configuration diagram of an adaptive filter. FIG. 6 is a configuration diagram of a filter for creating a filtered signal. FIG. 7 is a configuration diagram of a conventional noise canceling system for canceling road noise. [Description of Signs] 101a to 101c Sensors 102a and 102b Conversion filters 103a and 103b Adaptive signal processing unit 104 Adder 107 Speaker 109 Error microphone

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁷ , DB name) G10K 11/178 F01N 1/00 H03H 17/00 601 H03H 21/00 H04R 3/00 320

Claims (1)

(57) [Claim 1] A noise canceling system for canceling noise at an observation point generated from m noise sources, comprising: n (n> m) noise detection sensors; A conversion filter that receives the sensor detection signal and creates m uncorrelated signals; m adaptive signal processing units that perform adaptive signal processing using the m signals output from the conversion filter as reference signals; A canceling sound generating unit that receives a noise canceling signal obtained by synthesizing an output signal of the adaptive signal processing unit and outputs a noise canceling sound; Noise canceling system having a noise detection unit that feeds back to the system.