JP3084883B2 - Noise reduction device - 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 reduction device, and more particularly to a noise reduction device for reducing a noise component of a microphone output.

[0002]

2. Description of the Related Art Many microphones have a structure in which a change in sound pressure of a sound wave is converted into mechanical vibration of a diaphragm and an electroacoustic conversion system is operated based on the vibration. Therefore, when sound is picked up by the microphone, noise is generated if the diaphragm is affected by some factor.

[0003] If the above-mentioned factor is wind, noise due to the wind (hereinafter referred to as wind noise) is generated. Conventional techniques for reducing the above-mentioned wind noise include, for example, the following. (1) Use of windscreen (windshield) (2) Use of electric / acoustic high-pass filter (3) Adoption of configuration showing omnidirectionality in low frequency range

[0004]

The prior arts for reducing the above-mentioned wind noise have the following problems, respectively. If (1) is provided in the device, it is against the downsizing of the device. In general, the wind noise becomes smaller as the outer dimension of the windshield is larger and as the distance between the microphone and the inner wall of the windshield is larger.

[0005] With respect to (2): the sound pickup quality is reduced. Since wind noise is mainly composed of low frequency components, cutting the low frequency is effective for wind noise. However, in this case, not only the wind noise but also the low frequency components of the voice are cut off in the same manner.

[0006] For (3): The level at which the wind noise is reduced is insufficient. Although the level of wind noise is lower in omnidirectional microphones than in directional microphones, the `` configuration showing omnidirectionality in the low frequency range '' is actually due to the effects of the housing around the microphone. Recruitment alone is not at a sufficiently low level.

[0007] Therefore, in the current situation where the equipment provided with the microphone is further downsized and higher sound pickup quality is desired, the wind noise can be further reduced only by the above-mentioned prior art. Is becoming more difficult.

SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to provide a noise reduction apparatus which can be reduced in size and can reliably remove noise and external influence on voice input.

[0009]

SUMMARY OF THE INVENTION This invention comprises a first and a second microphone provided near the first and
And a second adaptive noise canceler, the first micro
Output of the first and second adaptive noise cancellers.
Each of which is supplied as a primary input and
The output is provided as a reference input for a first adaptive noise canceller.
And the residual output of the first adaptive noise canceller is
Supply as reference input of adaptive noise canceller
A noise reduction device characterized by being configured.

[0010]

The output of each of a pair of microphones provided close to each other includes an audio signal component and a noise component.
Since the pair of microphones are provided close to each other, the correlation between the audio signal components is high. Further, the correlation of the noise component is low, and is substantially uncorrelated.

In the preceding adaptive processing means, if the correlation of the audio signal components is high in principle, the difference in microphone sensitivity, the phase difference (time difference) between the input audio signals, and the like are absorbed. Therefore, the audio signal component having a high correlation is removed by the preceding adaptive processing means, and only the noise component is output as the residual component.

The main input of the subsequent adaptive processing means is the output of one microphone, and the reference input is the residual component of the preceding adaptive processing means.

In the subsequent adaptive processing means, the reference input is adaptively processed so as to be equal to the noise component of the main input, and the adaptively processed reference input is subtracted from the main input. Accordingly, only the noise component of the main input is minimized, that is, canceled, and the audio signal to noise ratio is maximized.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below with reference to FIGS. In the pair of microphones 1 and 2 arranged close to each other, surrounding sounds are collected together with wind noise, converted into electric signals, and output. Since the microphones 1 and 2 are arranged close to each other, substantially the same voice and wind noise are collected, converted into electric signals, and output. FIG. 3 shows an example of the frequency spectrum of the wind noise components included in the outputs of the microphones 1 and 2.
As is clear from FIG. 3, it can be understood that the wind noise is mainly composed of low frequency components.

The microphones 1 and 2 are not only arranged in the same direction, but are opposite to each other if, for example, the distance between the microphones 1 and 2 is within a wavelength range defined by the frequency of a desired signal. It may be. Microphone 1
Is supplied to the A / D conversion circuit 3, and the electric signal output from the microphone 2 is supplied to the A / D conversion circuit 4.

In the A / D conversion circuits 3 and 4, electric signals supplied from the microphones 1 and 2 are converted into digital signals. The digital signal converted by the A / D conversion circuit 3 is represented by (S + n), and the signal (S + n) is used as a main input to the adaptive noise canceller 6. A / D
The digital signal converted by the conversion circuit 4 is [(S *)
+ (N *)], and the signal [(S *) + (n
*)] Is a reference input to the adaptive noise canceller 6.

In the above signals, S and (S *) represent voice signal components, and n and (n *) represent wind noise components. Also, the noise components n and (n *) have an additive property,
The noise component (n *) and n have a correlation.

The main input (S + n) described above is delayed for a predetermined time by a delay circuit 7 provided in the adaptive noise canceller 6, and then supplied to an adder 8 of the adaptive noise canceller 6 and a delay circuit 22 of the adaptive noise canceller 21. You. The delay amount corresponds to a time delay required for the calculation for the adaptive processing, a time delay in the adaptive filter 9, and the like, and can be appropriately set according to the system configuration. Then, the output of the A / D conversion circuit 4 is supplied to an adaptive filter 9 provided in the adaptive noise canceller 6.

In the adaptive filter 9, the filter characteristics are successively self-adjusted, and based on the audio signal component (S *) of the reference input [(S *) + (n *)], the audio signal component of the main input (S + n) A signal Y0 is formed as a component similar to S. The configuration of the adaptive filter 9 and the algorithm for adaptively operating the adaptive filter 9 are the same as those of the adaptive filter 23 provided in the adaptive noise canceller 21.
And the duplicate description will be omitted.

The signal Y0 output from the adaptive filter 9 and given a negative sign is supplied to the adder 8. Also,
The main input (S + n) is supplied from the delay circuit 7 to the adder 8 and the delay circuit 22 of the adaptive noise canceller 21.

In the adder 8, the above-mentioned main input (S + n)
And the signal Y0 is subtracted. In the adder 8, the audio signal component S is removed by the signal Y0, and a noise component,
That is, only (n- (n *)) remains.

By the way, as described above, since the microphones 1 and 2 are arranged close to each other, the audio signal components S and (S *) show a high correlation especially in a low frequency band.
On the other hand, the noise components n and (n *) show low correlation.

In the adaptive processing of the adaptive noise canceller 6, if the correlation between the audio signal components S and (S *) is high in principle, the difference in sensitivity between the microphones 1 and 2 and the phase difference between the input audio signals [ Time difference] can be absorbed. Accordingly, the audio signal components S and (S *) are removed. As a result, only the noise component is used as the residual component ε0.

FIG. 4 shows an example of the coherence of the wind noise components generated by the pair of microphones 1 and 2. As shown in FIG. 4, it is generally known that wind noise components generated at two acoustic terminals have low correlation. Therefore, the difference between the outputs of the microphones 1 and 2 does not become zero even when the difference is obtained, and the above-mentioned residual component ε0 [= (n− (n
*))] Can be formed. This residual component ε0 [=
(N- (n *))] is shown in FIG.

The residual component ε0 [= (n− (n *))] is supplied to the adaptive filter 23 of the adaptive noise canceller 21 as a reference input, and is fed back to the adaptive filter 9.

The delay circuit 22 of the adaptive noise canceller 21
In this example, the main input (S + n) is output after being delayed for a predetermined time. The delay amount corresponds to a time delay required for the calculation for the adaptive processing, a time delay in the adaptive filter 23, and the like, and can be appropriately set according to the system configuration. The main input (S + n) that has passed through the delay circuit 22 is supplied to the adder 24.

In the adaptive filter 23, the filter characteristics are successively self-adjusted to form a signal Y1 as a component similar to the noise component n of the main input (S + n). That is, the reference input to the adaptive noise canceller 21 (n- (n *))
Is the main input (S +
The filter characteristics are self-adjusted sequentially so as to resemble the audio signal component S of n). The signal Y1 is supplied to the adder 24.

The configuration of the adaptive filter 23 will be described. As the adaptive filter 23, an FIR filter type adaptive linear combiner shown in FIG. 2 is used. In the configuration of FIG. 2, DL1 to DLL represent delay circuits,
P 0 to MPL represent coefficient multipliers. Reference numeral 16 denotes an adder, and reference numerals 15 and 17 denote terminals.

[Z ^-1 ] in the delay circuits DL1 to DLL represents a delay of a unit sampling time, and W _nk supplied to the coefficient multipliers MP 0 to MPL represents a weight coefficient. If the weighting coefficient W _nk is fixed, it is a normal FIR digital filter.

In the adder 24, the output from the delay circuit 22 and the signal Y1 which is given a negative sign and output from the adaptive filter 23 described later are added. This signal Y1 is
As will be described later, it is a component similar to the noise component n in the main input (S + n). Therefore, in the adder 24,
The noise component n is subtracted from the main input (S + n), and the audio signal component S is maximized. In other words, the main input (S +
The noise component n of n) is minimized and substantially canceled.

The audio signal component S is fed back to the adaptive filter 23 and supplied to the D / A conversion circuit 10. In the D / A conversion circuit 10, the audio signal component S is converted into an analog signal, and the analog signal is taken out from the terminal 11.

FIG. 6 shows the effect of noise reduction according to this embodiment. FIG. 6 shows a main input (S + n) indicated by a solid line, that is, an output of the microphone 1, and a system output indicated by a broken line, that is, an output of the adaptive noise canceller 21 extracted from the terminal 11. . The audio signal component S is represented as 500 in a pseudo manner.
Hz sine wave is added.

As is apparent from FIG. 6, the level of the noise component n at the output of the microphone 1 [solid line in FIG. 6] corresponds to the output of the adaptive noise canceller 21 [FIG.
(Broken line in the middle) is remarkable. Also, 500
It can be seen that the Hz sine wave maintains its level regardless of the presence or absence of the adaptive noise canceller 21.

Here, an algorithm for adaptively operating the adaptive filter 23 will be described. Various algorithms can be used for the operation in the adaptive filter 23. However, the LMS (least mean square) algorithm, which has a relatively small amount of calculation, is practical, and is frequently used, will be described below. This LMS (Least Mean Square) algorithm is also used in the adaptive filter 9 provided in the adaptive noise canceller 6, although not particularly specified in the following description.

If the input vector Xk is expressed as X _k = [X _k X _k-1 X _k-2 ... X _kL ], the output Y _k of the adaptive filter 23 becomes Given by

Assuming that the output of the delay circuit 22 is d _k , the difference output (residual output) is ε _k = d _k −X _k ^T W _k . In the LMS (Least Mean Square) method, the updating of the weight vector is performed according to the following equation. W _{k + 1} = W _k +2 με _k X _{k In the} above equation, μ is a gain factor that determines the stability of the speed of adaptation, that is, a so-called step gain.

By updating the weight vector as described above, an operation is performed to minimize the output power of the system. Hereinafter, this operation will be formalized and described. If the delay circuit 22 is ignored for simplicity, the difference output ε from the adder 24 is ε = S + n−Y.

The expected value of the square of ε is represented by the following equation. E [ε ² ] = E [S ² ] + E [(n−Y) ² ] + 2E
[S (n−Y)] Here, since S is uncorrelated with n and Y, E [S (n−Y)] = 0 in the above equation. Therefore, the expected value E of the squares of epsilon [epsilon ^2] is represented by the following equation. E [ε ² ] = E [S ² ] + E [(n−Y) ² ]

The adaptive filter 23 is adjusted so that E [ε ² ] is minimized, but since E [S ² ] is not affected, the following equation is obtained. Emin [ε ² ] = E [S ² ] + Emin [(n−Y) ² ]

Since E [S ² ] is not affected,
Minimizing [ε ² ] means that E [(n−Y) ² ] is minimized. Therefore, the output Y of the adaptive filter 23 is the best least squares estimated value of [n].

When E [(n−Y) ² ] is minimized,
Since [ε−S = n−Y], E [(ε−
S) ² ] is also minimized. Therefore, adjusting the adaptive filter 23 to minimize the total output power is equivalent to the difference output ε
Becomes the best least squares estimate of the audio signal component S.

The difference output ε is generally a sound signal component S with some noise components added. Since the noise component to be output is given by (n−Y), E [(n−Y )
² ] is equivalent to maximizing the signal-to-noise ratio of the output.

According to this embodiment, the main input (S + n) is formed based on the output of the microphone 1 among the outputs of the pair of microphones 1 and 2 arranged close to each other.
Reference input [(S *) + based on the output of microphone 2
(N *)] is formed.

The adaptive filter 9 of the adaptive noise canceller 6
Then, the self-adjustment is sequentially performed so that the audio signal component (S *) becomes the audio signal component S, and the signal Y0 formed based on the audio signal component (S *) is input to the main input (S +
n). As a result, the audio signal components S, (S
*) Are removed, and the adder 8 outputs a residual component ε0 [= (n− (n *))] including only a noise component. That is,
From the adaptive noise canceller 6, the main input (S + n) and the residual component ε0 [= noise component (n− (n *)) are supplied to the adaptive noise canceller 21.

That is, the main input (S + n) of the adaptive noise canceller 6 becomes the main input (S + n) of the adaptive noise canceller 21, and the residual component ε 0 of the adaptive noise canceller 6 becomes the reference input of the adaptive noise canceller 21.

In the adaptive filter 23 of the adaptive noise canceller 21, based on the above-mentioned reference input (n- (n *)),
Signal Y1 similar to noise component n in primary input (S + n)
Is formed. The signal Y1 is supplied to the main input (S
+ N), the noise component n is canceled and the audio signal component S is output.

Therefore, a pair of ordinary microphones 1, 2
Is used, it is possible to reliably cancel the wind noise component. In order to remove this noise component, a windshield is not used, and the microphones 1 and 2 are arranged close to each other, which can contribute to downsizing of the device. Also, it is necessary to use an electric / acoustic high-pass filter or the like. Since there is no sound quality, it is possible to prevent a decrease in sound pickup quality.

Further, since the adaptive noise cancellers 6 and 21 are used, even if the characteristics of the wind noise (for example, level or spectrum distribution) change, the adaptive filters 9 and 23 are changed.
Is automatically updated, and the wind noise component can be reduced stably.

Further, the difference in sensitivity between the microphones 1 and 2,
The phase difference [time difference] between the input audio signals can be absorbed, and the influence on the audio signal component S due to these external factors can be minimized.

The noise reduction device shown in this embodiment is
It can be applied to various sound collection systems. For example, the present invention can be applied to a small portable video camera device or a single microphone. The pair of microphones 1 and 2 shown in this embodiment can be used regardless of directivity.

In this embodiment, an example is shown in which adaptive noise cancellers 6 and 21 having almost the same configuration are used in cascade connection in two stages. However, the present invention is not limited to this. For example, the adaptive noise cancellers 6 and 21 may be realized by software and a DSP and used in a time-division manner.

[0052]

According to the noise reduction apparatus of the present invention, there is an effect that the noise component can be surely canceled. In addition, regarding the removal of noise components, a pair of microphones are disposed close to each other without using a windshield, thereby contributing to downsizing of the device. Further, there is no need to use an electric / acoustic high-pass filter or the like. There is an effect that a reduction in sound pickup quality can be prevented. Furthermore,
Even if a processing system is not prepared for each type of noise, there is an effect that good sound collection quality can be realized with a single processing system.

Further, according to the embodiment, since the adaptive noise canceller is used, even if the characteristics of the wind noise (for example, the level or the spectrum distribution) change, the characteristics of the adaptive filter are automatically updated and the wind noise is reduced. There is an effect that the noise component can be stably reduced.

The adaptive processing means in the preceding stage has an effect that the influence on the audio signal component due to external factors can be absorbed and the influence on the audio signal component can be minimized.

[Brief description of the drawings]

FIG. 1 is a block diagram showing one embodiment of the present invention.

FIG. 2 is a block diagram illustrating a configuration of an adaptive filter.

FIG. 3 is a diagram illustrating a frequency spectrum of a wind noise component.

FIG. 4 is a diagram showing the degree of correlation between wind noise components collected by a pair of microphones.

FIG. 5 is a diagram illustrating an example of a difference output of a wind noise component collected by a pair of microphones.

FIG. 6 is a waveform chart showing a noise reduction effect.

[Explanation of symbols]

 1,2 Microphone element 6,21 Adaptive noise canceller

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int. Cl. ⁷ , DB name) H04R 3/00 320 G10K 11/178 H03H 21/00 H04R 1/40 320

Claims (1)

(57) [Claims] And 1. A first and second microphone provided near, and first and second adaptive noise canceler, the outputs of the first and second first microphone
Supplied as main input of adaptive noise canceller
Then, the output of the second microphone is changed to the first adaptive noise.
Is supplied as a reference input of the
The residual output of the adaptive noise canceller to the second adaptive noise
Noise reduction apparatus according to claim Rukoto configured to supply a canceller of the reference input.