JPH11259077A - Noise reduction device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a noise reduction device capable of obtaining in a short time the optimum value of transfer characteristics used for muffling noises. SOLUTION: When determining the optimum values of transfer characteristics Gs, Ge by an operation part 10, random noises are discharged from a secondary sound source 16, and the transfer characteristics Gs, Ge are updated by each of the convergence coefficients μs, μe so that a difference between an output signal of a sensor microphone 14 and a random noise which has passed a filter 17, and a difference between an output of an error microphones 15 and a random noise which has passed filter 18 are minimized respectively, based on the random noises which have passed each of the filters 17, 18 having each output signal and transfer characteristics Gs, Ge of the microphones 14, 15, and the convergence coefficients μs, μe are altered based on changes in the respective transfer characteristics Gs, Ge.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a noise reduction device for reducing noise of an air conditioning duct or the like.

[0002]

2. Description of the Related Art Conventionally, as a method of silencing noise propagating in an air-conditioning duct or the like, a passive silencing method such as a method of silencing by a sound absorbing material lined in a duct has been generally used. There are problems such as pressure loss and the size when silencing low frequencies, and active silencing methods that muffle sound by emitting noise in opposite phase to sound waves have been studied. However, many problems remain.

FIG. 3 is a block diagram showing a configuration of a conventional noise reduction device. In the figure, noise transmitted through a duct 63 is detected as an analog electric signal by a microphone 64 (hereinafter referred to as a sensor microphone), and the signal is amplified by a microphone amplifier 51. The frequency cutoff filter 52 is applied. Thereafter, an analog-digital converter (hereinafter, A / D)
The signal is converted into a digital signal through a converter 53. The digital signal thus obtained is input to the arithmetic unit 60.

[0004] Further, the system is placed in a place where noise is to be reduced due to the operation of the system (in the case of the duct shown in FIG. 3, it is located downstream of a noise reduction speaker 66 (hereinafter referred to as a secondary sound source) as viewed from a noise source. Similarly to the sensor microphone 64, the analog signal of the microphone 65 (hereinafter, referred to as an error microphone) for detecting the mute state is also input to the microphone amplifier 5
After amplification at 7 and passing through a frequency cutoff filter 58, A
The signal is converted into a digital signal through a / D converter 59.

[0005] The converted digital signal e (n) representing the muffling state is input to the calculation unit, and the characteristic G between the secondary sound source 66 and the sensor microphone 64 simulating the transfer characteristic of the sound in the space is obtained.
s and a signal obtained by filtering using the characteristic Ge between the secondary sound source 66 and the error microphone 65 to extract only the sound from the upstream of the noise, and an optimal filter coefficient such that the signal e (n) always approaches zero. W is calculated by the adaptive algorithm 62, and a convolution operation is performed on the input signal from the sensor microphone 64 as a filter coefficient W of the muffling signal generation filter 61 to create a muffling digital filter.

Here, the secondary sound source 66 and the sensor microphone 64
The characteristic Gs between the secondary sound source 66 and the characteristic Ge between the secondary sound source 66 and the error microphone 65 are obtained before the sound is muted.
4 is configured as shown in FIG. 4 and the M-sequence noise output unit 6
9, the M-sequence random noise is output from the secondary sound source 66, and the adaptive algorithm is applied to the signals obtained from the sensor microphone 64 and the error microphone 65 through the respective electric transfer characteristics and the M-sequence random noise. It is required by making it correspond.

Here, the Filtered-X LMS algorithm well known as the steepest descent method will be described as the adaptive algorithm 62. First, a signal e representing the mute state
(n) is a signal d transmitted from the noise source through the duct 63.
(n) and the signal Y (n) output from the secondary sound source 66 pass through the duct 63 and enter the error sensor 65.
The transfer function from the secondary sound source 66 to the error sensor 65 is G
If e is given, it can be expressed as the following equation (1).

E (n) = d (n) + Ge · Y (n) (1) where Y (n) = [Y (n), Y (n−1), Y (n−2), Y (n-3
),..., Y (nj)] ^T Ge = [Ge0, Ge1, Ge2, Ge
3, Λ, Gej] j: the number of elements preset as elements of the transfer function Ge

The mute signal generation filter 61 uses a renewable filter coefficient W to output a signal X (n) from the sensor microphone 64.
Is convolved to generate the signal Y (n) output from the secondary sound source 66, so that it can be expressed as the following equation (2).

Y (n) = W.X (n) (2) where X (n) = [X (n), X (n-1), X (n-2), X (n-
3),..., X (ni)] ^T W = [W0, W1, W2, W3,
Λ, Wi] i: number of elements preset as elements of filter coefficient W

That is, the signal e (n) indicating the mute state can be described as the following equation (3) from the above equations (1) and (2). e (n) = d (n) + Ge ｛{W ・ X (n)｝ = d (n) + W ｛{Ge ・ X (n)｝ = d (n) + WR ・ (n) (3) , R (n) are represented by the following reference signals. R (n) = Ge.X (n)

Next, if the evaluation function J = [e (n)] ² is set in order to minimize the signal e (n) representing the mute state, the following equation (4) can be obtained. J = [e (n)] ² = [d (n) + W · R (n)] ² (4) Since the evaluation function J is a quadratic function with respect to the filter coefficient W, when the minimum is obtained, This is when the derivative shown in equation (5) becomes zero.

[0013]

(Equation 1)

Therefore, since the control object changes with time, the filter coefficient W is sequentially updated using the following equation (6), which is a steepest descent method, so that the evaluation function J is minimized for that state. do it.

[0015]

(Equation 2)

The signal Y (n) output from the secondary sound source 66 is generated from the equation (2) using the filter coefficient W obtained here. This signal Y (n) is converted into an analog signal by a digital-analog signal converter 56 (hereinafter, referred to as a D / A converter), and a noise cutoff analog signal is generated through a frequency cutoff filter 55 by analog signal processing. This analog signal is amplified by the speaker amplifier 54 to drive the speaker of the secondary sound source 66,
By radiating the noise to be silenced, the noise and the sound wave for silencing collide, and the noise is canceled out by causing interference to cancel each other.

In the example shown in FIG. 3, the above Filtered-X LMS
Using the spatial characteristic Gs between the secondary sound source 66 and the sensor microphone 64 and the signal Y (n) output from the secondary sound source 66 in the algorithm, the sound output from the secondary sound source 66 that has entered the sensor microphone 64 Is to be removed.

As described above, according to the active noise type noise reduction apparatus using the conventional adaptive filter, the filter coefficient of the adaptive filter is sequentially updated so that the noise that cannot be completely eliminated in the duct to be controlled is minimized. While silencing. The frequency cut-off filter used above is used to remove signals such as noise and to process only signals in a frequency region to be silenced.
Generally, as shown in FIG. 5, a frequency cutoff filter having a frequency characteristic that passes only a signal having a frequency lower than a certain frequency fc with respect to an input signal is used. .

[0019]

However, in the conventional noise reduction apparatus as described above, when performing the process of identifying the space between the secondary sound source 66 and the sensor microphone 64 or the space between the secondary sound source 66 and the error microphone 65, By using an adaptive algorithm to calculate the noise reduction signal generation filter coefficient such that the error signal always approaches zero by changing the convergence coefficient μ,
In order to converge quickly, the convergence coefficient μ may be increased. However, if the convergence coefficient μ is increased, no matter how many operations are repeated, the optimum value cannot be reached,
If the convergence coefficient μ is reduced, there is a problem that it takes a long time to reach the optimum value.

[0020]

According to the present invention, there is provided a noise reduction device comprising:
Sensor, a secondary sound source and a second sensor, a first transfer characteristic between the secondary sound source and the first sensor, a second transfer characteristic between the secondary sound source and the second sensor, A calculating unit that radiates a sound wave having the same sound pressure and opposite phase to be silenced with respect to the noise sound wave from the secondary sound source based on the output signal and the output signal of the second sensor, thereby muting the noise sound wave by interference of both sound waves The arithmetic unit causes the random noise to be emitted from the secondary sound source when determining the optimum values of the first and second transfer characteristics, and the output signals of the first and second sensors, and the first and second A difference between the output signal of the first sensor and the random noise passed through the filter having the first transfer characteristic, based on the random noise passed through each filter having the second transfer characteristic, and the output signal of the second sensor And a filter having a second transfer characteristic. The first and second transfer characteristics are respectively updated by the first and second convergence coefficients so that the difference from the random noise obtained is minimized, and the first and second convergence coefficients are respectively updated by the first and second convergence coefficients. First, it is changed based on a change in the second transfer characteristic.

[0021]

FIG. 1 is a block diagram showing a configuration at the time of space identification processing in a noise reduction device according to an embodiment of the present invention, wherein a noise is transmitted through a duct 13 from left to right. One-dimensional duct sensor microphone, error microphone,
The positional relationship between the secondary sound sources and the calculation unit are shown.

In the figure, an arithmetic unit 10 includes an M-sequence noise output unit 19 for generating random noise, adaptive control algorithms 11 and 12 for performing an LMS algorithm, and 2
A transfer characteristic Gs filter 17 between the secondary sound source 16 and the sensor microphone 14, a transfer characteristic Ge filter 18 between the secondary sound source 16 and the error microphone 15, subtraction units 22 and 23, and convergence coefficient change units 20 and 21 are included. Make up.

The M-sequence noise signal generated by the arithmetic unit 10 is converted into an analog signal by the D / A converter 6 and then passes through the frequency cutoff filter 5 and the speaker amplifier 4 to provide a speaker 16 as a secondary sound source. Output as sound. The sound detected by the sensor microphone 14 passes through the microphone amplifier 1 and the frequency cutoff filter 2, is converted into a digital signal by the A / D converter 3, and is input to the arithmetic unit 10. Similarly, the error microphone 15 also passes through the microphone amplifier 7 and the frequency cutoff filter 8, is converted into a digital signal by the A / D converter 9, and is then input to the arithmetic unit 10.

Next, the operation of this embodiment will be described. First, the M-sequence noise signal created by the arithmetic unit 10 is converted from a digital signal to an analog signal through the D / A converter 6, passes through the frequency cutoff filter 5 and the speaker amplifier 4, and outputs a speaker 16 as a secondary sound source. Output as sound. The sound output from the secondary sound source 16 passes through the space of the one-dimensional duct 13 and is detected by the sensor microphone 14 and the error microphone 15 which are detection microphones, and amplifies the signal through the microphone amplifiers 1 and 7.
After only the target signal is extracted by the frequency cutoff filters 2 and 8, the signal is converted into a digital signal by passing through the A / D converters 3 and 9 and input to the arithmetic unit 10.

The signals from the sensor microphone 14 and the error microphone 15 input to the arithmetic unit 10 are subtracted by filtering the signal of the M-sequence noise output unit 19 with the transfer characteristic filters 17 and 18, and the adaptive control algorithm 11 and 12 are input. Then, the adaptive control algorithms 11 and 12 sequentially update the transfer characteristic filters Gs and Ge by the convergence coefficients μs and μe by the LMS algorithm so that the subtracted value approaches zero as much as possible.

The values of the transfer characteristic filters Gs and Ge are sent to the convergence coefficient changing units 20 and 21, and the convergence coefficient μs determined by the convergence coefficient changing units 20 and 21 will be described later.
, Μe are sent to adaptive control algorithms 11 and 12,
The values of the transfer characteristic filters Gs and Ge are updated with this value. This is repeated to converge the transfer characteristic filter G
s and Ge simulate the effects of transfer functions unique to each such as a time delay and a change in frequency characteristics when a sound wave propagates in the duct 13 and a signal passes through the circuit.

Here, the operation of the convergence coefficient changing units 20 and 21 in the identification processing will be described. FIG. 2 is a flowchart showing the operation of the convergence coefficient changing unit. Since the identification process is performed in the same manner on the sensor microphone 14 side and the error microphone 15 side, the operation will be described as the operation on the sensor microphone 14 side.

First, the initial value of the convergence coefficient μs and the final convergence coefficient μend which is the minimum value of the convergence coefficient μs are preset, and the initial value of the convergence coefficient μs is μ
It is set to a value that is larger than end and that the convergence of the transfer characteristic filter is fast and the accuracy is considered to be good to some extent.

As the operation of changing the convergence coefficient in the convergence coefficient changing unit, first, the individual coefficients of the new transfer characteristic filter Gs and the previously stored transfer characteristic filter Gs' are subtracted (P (n) = Gs ( n) -Gs' (n)) (S100), S10
0, the maximum value B is obtained from the subtracted value P (n) (S
101). Then, the maximum value B obtained in S101 is compared with the current convergence coefficient μs (S102). If “maximum value B <current convergence coefficient” in S102, the maximum value B is set to a preset final convergence coefficient. Is compared with the coefficient μend (S10
3) In the case of “maximum value B <final convergence coefficient μend” in S103, the identification process ends (S104).

If “maximum value B ≧ current convergence coefficient” in S102, the new transfer characteristic filter is
This is stored in s' (n) (S105), and this process ends.
If “maximum value B ≧ final convergence coefficient μend” in S103, the value of the convergence coefficient μs is reduced (S106),
This new transfer characteristic filter is stored in Gs' (n) (S105), and this process ends.

Then, using the convergence coefficient μs obtained by the operation of the convergence coefficient changing unit 20 as shown in FIG. 2, the adaptive control algorithm 11 sequentially updates the transfer characteristic filter Gs to obtain the optimum transfer characteristic. The value of the filter Gs is obtained, and the optimum value of the transfer characteristic filter Ge is obtained by the same processing. Then, by using the obtained values of the transfer characteristic filter Gs and the transfer characteristic filter Ge, the operation of the calculation unit as shown in FIG. 3 performs actual silencing processing in the duct.

Here, the flowchart shown in FIG. 2 is set to be performed every time the transfer characteristic changes (every sampling). However, the present invention is not limited to this. Can be given the same meaning as the above flow by comparing the maximum change rate of Gs (n) after a certain time has elapsed with the convergence coefficient μs.

In this embodiment, in the identification processing,
Compared to the case where the convergence coefficient of the adaptive algorithm is fixed, the time required for the identification process can be significantly reduced, and a highly accurate transfer characteristic filter can be obtained. Is set, the accuracy of the transfer characteristic filter is determined, so that a stable identification filter can be obtained even in various environments.

In this embodiment, an example of a noise reduction system for noise in a duct has been described. However, even in a noise reduction device of another form and shape, and in another adaptive algorithm, a filter coefficient is included in the adaptive algorithm. The same effect can be obtained for an object having the identification processing.

[0035]

As described above, according to the present invention, when the calculation unit determines the optimum values of the first and second transfer characteristics,
Random noise is emitted from the secondary sound source, and the output of the first sensor is determined based on the output signals of the first and second sensors and the random noise that has passed through the filters having the first and second transfer characteristics. The difference between the signal and the random noise passed through the filter having the first transfer characteristic and the difference between the output signal of the second sensor and the random noise passed through the filter having the second transfer characteristic are minimized. Then, the first and second transfer characteristics are updated by the first and second convergence coefficients, respectively, and the first and second convergence coefficients are updated based on the changes in the first and second transfer characteristics, respectively. , The convergence coefficient is kept constant, and the time required to determine the transfer characteristics can be significantly reduced as compared with the case where the optimal transfer characteristics are determined. An effect that can be obtained sex.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration at the time of space identification processing in a noise reduction device according to an embodiment of the present invention.

FIG. 2 is a flowchart illustrating an operation of a convergence coefficient changing unit.

FIG. 3 is a block diagram illustrating a configuration of a conventional noise reduction device.

FIG. 4 is a block diagram showing a configuration at the time of space identification processing in a conventional noise reduction device.

FIG. 5 is an explanatory diagram for explaining characteristics of a frequency cutoff filter.

[Explanation of symbols]

Reference Signs List 10 arithmetic unit 11, 12 adaptive control algorithm 13 duct 14 sensor microphone (first sensor) 15 error microphone (second sensor) 16 secondary sound source 17 transfer characteristic Gs filter (filter having first transfer characteristic) 18 transmission Characteristic Ge filter (filter having second transfer characteristic) 19 M-sequence noise output unit 20 Convergence coefficient μs changing unit 21 Convergence coefficient μe changing unit 22, 23 Subtraction unit

Claims (1)

[Claims] A first sensor, a second sound source, and a second sensor arranged in order with respect to a transmission direction of the noise sound wave; a first transfer characteristic between the second sound source and the first sensor;
A second transfer characteristic between the secondary sound source and the second sensor;
On the basis of the output signal of the first sensor and the output signal of the second sensor, by radiating from the secondary sound source a sound wave of the same sound pressure opposite phase to be silenced with respect to the noise sound wave, An arithmetic unit that silences the noise sound wave by interference, wherein the arithmetic unit emits random noise from the secondary sound source when determining the optimal values of the first and second transfer characteristics, and Based on each output signal of the second sensor and the random noise that has passed through each of the filters having the first and second transfer characteristics, the output signal of the first sensor and the filter having the first transfer characteristic The difference between the passed random noise and the output signal of the second sensor and the second
The first and the second so that the difference from the random noise passed through the filter having the transfer characteristic of
Are updated by the first and second convergence coefficients, respectively, and the first and second convergence coefficients are changed based on the changes of the first and second transfer characteristics, respectively. A noise reduction device characterized by the above-mentioned.