close

Вход

Забыли?

вход по аккаунту

?

JP2008283385

код для вставкиСкачать
Patent Translate
Powered by EPO and Google
Notice
This translation is machine-generated. It cannot be guaranteed that it is intelligible, accurate,
complete, reliable or fit for specific purposes. Critical decisions, such as commercially relevant or
financial decisions, should not be based on machine-translation output.
DESCRIPTION JP2008283385
The present invention provides a noise suppression device that suppresses noise components
included in a stereo signal while maintaining a natural sound image. A common component
extraction unit (110) extracts a component Mi (n) commonly included in a plurality of input
channels. The non-common component extraction unit 210 extracts a component Si (n) not
common to the input channel. The common component noise suppression unit 310 performs
noise suppression processing on the component Mi (n) to obtain a signal Mo (n). The noncommon component processing unit 410 attenuates or enhances the component Si (n) to obtain
a signal So (n). The multiple channel generation unit 510 removes noise of a common component
from the signal In (n, k) of each input channel using the signal Mo (n) and the signal So (n), and
attenuates or enhances non-common components. It is something like that. [Selected figure]
Figure 1
Noise suppressor
[0001]
BACKGROUND OF THE INVENTION Field of the Invention The present invention relates to a
noise suppression device that suppresses noise of a voice signal having a plurality of channels.
[0002]
As well known, as a conventional noise suppression method for stereo signals, for example, there
is a method of suppressing noise by superposing noise of opposite phase on each of L channel
and R channel (for example, see Patent Document 1).
08-05-2019
1
Although this method can suppress the noise components (ea, eb) commonly contained in both L
and R channels, it is difficult to suppress the noise of the stereo component. Moreover, since the
sound image after suppression is not considered, there is a problem that it becomes an unnatural
sound image. Unexamined-Japanese-Patent No. 2-205120 gazette.
[0003]
The conventional noise suppression device has a problem that the noise component included in
the stereo signal can not be suppressed while maintaining a natural sound image. The present
invention has been made to solve the above-mentioned problems, and it is an object of the
present invention to provide a noise suppression device that suppresses noise components
included in a stereo signal while maintaining a natural sound image.
[0004]
In order to achieve the above object, the present invention provides a common component for
extracting a component common to a first channel signal and a second channel signal in a noise
suppression device for suppressing noise components included in input signals of a plurality of
channels. Extracting means; common component noise suppressing means for suppressing noise
with respect to common components extracted by the common component extracting means;
uncommon component extracting means for extracting components not common to the first
channel signal and the second channel signal; The noise component is suppressed from the
uncommon component processing means for performing level adjustment, the output of the
common component noise suppression means, and the output of the uncommon component
processing means for the uncommon components extracted by the uncommon component
extraction means A plurality of channel generation means for respectively generating the first
channel signal and the second channel signal are provided.
[0005]
As described above, in the present invention, noise suppression is performed by obtaining a
component common to the first channel signal and the second channel signal, and level
adjustment is performed to obtain a component not common to the first channel signal and the
second channel signal. And generate the first channel signal and the second channel signal in
which the noise component is suppressed based on them.
08-05-2019
2
Therefore, according to the present invention, it is possible to provide a noise suppression device
capable of suppressing noise components included in a stereo signal while maintaining a natural
sound image.
[0006]
Hereinafter, embodiments of the present invention will be described with reference to the
drawings. FIG. 1 shows a configuration of a noise suppression apparatus according to a first
embodiment of the present invention.
[0007]
The noise suppression apparatus includes a common component extraction unit 110, a noncommon component extraction unit 210, a common component noise suppression unit 310, a
non-common component processing unit 410, and a multiple channel generation unit 510.
[0008]
The common component extraction unit 110 receives a plurality of input channel signals In (n, k)
(k = 1,..., Kin: Kin is the total number of input channels and n is time), and is common to a
plurality of input channels. Extract the component Mi (n) contained.
The non-common component extraction unit 210 receives signals In (n, k) of a plurality of input
channels and extracts components Si (n) not common to these input channels.
[0009]
The common component noise suppression unit 310 performs noise suppression processing on
the component Mi (n) extracted by the common component extraction unit 110, obtains a signal
Mo (n), and outputs this. The non-common component processing unit 410 performs
predetermined level adjustment processing (for example, attenuation or enhancement) on the
component Si (n) extracted by the non-common component extraction unit 210 to obtain a signal
08-05-2019
3
So (n). Output this.
[0010]
The multiple channel generation unit 510 uses the signal Mo (n) obtained by the common
component noise suppression unit 310 and the signal So (n) obtained by the non-common
component processing unit 410 to generate a plurality of input channels. Output signals Out (n,
k) (k = 1,..., Kout: Kout: total number of output channels, n) where noises of common components
are removed from the signal In (n, k) and non-common components are attenuated or
emphasized Outputs the time).
[0011]
Here, the values of the input channel numbers Kin and Kout may be two or more values, but to
simplify the following description, Kin = Kout = 2, that is, stereo input and stereo output as an
example. I will list and explain.
In this case, for example, the configuration of the noise suppression apparatus shown in FIG. 1
may be configured as shown in FIG.
[0012]
The common component extraction unit 111 receives signals In (n, 1) and In (n, 2) of two input
channels, and extracts a component Mi (n) commonly included in the two input channels. For
example, extraction is performed by the calculation as shown in the following equation (1). That
is, by adding the signals In (n, 1) and In (n, 2) of the two input channels, a signal in which the
components having common phases become more prominent is obtained. If necessary,
processing may be added to shift the time axis of the signals In (n, 1) and In (n, 2) to synchronize
the phases.
[0013]
The non-common component extraction unit 211 receives signals In (n, 1) and In (n, 2) of two
input channels, and extracts a component Si (n) that is not common to the two input channels.
08-05-2019
4
For example, it extracts by the operation as shown to the following Formula (2). That is, by
subtracting the signals In (n, 1) and In (n, 2) of the two input channels, a signal in which
components whose phases are not in common become more prominent is obtained. If necessary,
processing may be added to shift the time axis of the signals In (n, 1) and In (n, 2) to synchronize
the phases.
[0014]
The monaural component noise suppression unit 311 performs noise suppression processing on
the component Mi (n) extracted by the common component extraction unit 111, obtains a signal
Mo (n), and outputs this. Here, as a method of the noise suppression processing, a document (S.
F. Boll, "Suppression of acoustic noise in speech using spectral subtraction," IEEE Trans. There is
a so-called Spectral Subtraction (SS) method disclosed in ASSP, vol. 27, no. 2, pp. 113-120, Feb.
1979 (p. 114, Chapter 2 C)). Alternatively, Y. Ephraim et a1., “Speech enhancement using a
minimum mean-square error short-time spectral amplitude estimator,” “ASSP, vol. 32, no. 6,
pp. 1109-1121, 1984 (p. 53)), or the so-called MMSE-STSA method, JSLim and AVOppenheim,
"Enhancement and Bandwidth Compression of Noisy Speech," Proc. The Wiener filter method
disclosed in Of the IEEE, vol. 67, pp. 1586-1604, Dec. 1979 is representative.
[0015]
In these methods, as the signal-to-noise ratio SNR (k) of each frequency band is larger, the gain of
band k is closer to 1 (that is, the amount of suppression = O dB), and as SNR (k) is smaller, the
band k It is a method of suppressing the noise component in the input signal by bringing the gain
closer to 0 or the positive lower limit value.
[0016]
The stereo component processing unit 411 performs predetermined level adjustment processing
(for example, attenuation or enhancement) on the component Si (n) extracted by the noncommon component extraction unit 211 to obtain a signal So (n). Output this.
Here, in the case of attenuation, an operation according to the following expression (3) is
performed using the gain G.
08-05-2019
5
[0017]
However, 0 <G <1, for example, G = 0.5 is used. In the case of emphasis, Equation (3) is
calculated, where 1 <G, for example, G_low = 2. Also, in order to suppress distortion, G = 1, that
is, the stereo component processing unit 411 may output without performing any processing.
[0018]
The multiple channel generation unit 511 includes a channel generation unit 511a and a channel
generation unit 511b. The channel generation unit 511 a and the channel generation unit 511 b
use the signal Mo (n) obtained by the monaural component noise suppression unit 311 and the
signal So (n) obtained by the stereo component processing unit 411, respectively. Common signal
noise is removed from the signals In (n, 1) and In (n, 2) of one input channel, and output signals
Out (n, 1) and Out (n, 2) with attenuated or emphasized non-common components. Output). For
example, it calculates | requires and outputs by calculation as shown to following Formula (4),
(5).
[0019]
In order to match the amplitude between the signals In (n, 1) and In (n, 2) of the two input
channels and the respective output signals Out (n, 1) and Out (n, 2) , Instead of outputting Out (n,
1) and Out (n, 2) by adding a configuration for performing the operation as shown in the
following equations (6) and (7), Out_new (n, 1), Out_new (n, 2) may be output. Here, K is a
coefficient for matching the amplitudes of the input signal and the output signal, and may be, for
example, K = (1 + G).
[0020]
As described above, in the noise suppression device of the above configuration, noise of a sound
whose sound image is in the center direction among the signals of two input channels, that is,
stereo signals, is suppressed. An effect that makes the announcer's voice easy to hear is obtained,
and the stereo component can be attenuated or emphasized separately from the monaural
component. For this reason, if a stereo component is subjected to, for example, gain adjustment
to the extent of adjusting the volume, it is possible to hear a natural sound image.
08-05-2019
6
[0021]
Next, a noise suppression apparatus according to a second embodiment of the present invention
will be described. FIG. 3 shows the configuration. The noise suppression apparatus shown in this
figure includes a common component extraction unit 120, a non-common component extraction
unit 220, a common component noise suppression unit 320, a non-common component
processing unit 420, and a multiple channel generation unit 520. .
[0022]
The common component extraction unit 120 receives a plurality of input channel signals In (n, k)
(k = 1, ..., Kin: Kin is the total number of input channels and n is time), and is common to a
plurality of input channels. Extract the component Mi (n) contained. The non-common
component extraction unit 220 receives signals In (n, k) of a plurality of input channels and
extracts components Si (n) not common to these input channels.
[0023]
The common component noise suppression unit 320 performs noise suppression processing on
the component Mi (n) extracted by the common component extraction unit 120, obtains a signal
Mo (n), and outputs this. Further, the common component noise suppression unit 320 detects the
level of the noise signal included in the component Mi (n), that is, the noise signal level of the
common component, and notifies the non-common component processing unit 420 of this.
[0024]
The non-common component processing unit 420 performs predetermined level adjustment
processing on the component Si (n) extracted by the non-common component extraction unit
220 with an intensity corresponding to the level notified from the common component noise
suppression unit 320. (Eg, attenuation or enhancement) to obtain signal So (n) and output it.
[0025]
08-05-2019
7
The noise signal included in the common component among the plurality of input channels and
the noise signal included in the non-common component are considered to be correlated to some
extent.
The noise suppression apparatus according to the second embodiment pays attention to this
point, and for example, when noise of a component common to a plurality of input channels is
large, it is considered that noise of a component not common is also large. The suppression in the
common component processing unit 420 is made stronger. Or, conversely, when the noise level
of the component not common is detected in the non-common component processing unit 420
and the noise of the common component is large when the noise of the non common component
is large, the gain in the common component noise suppression unit 320 is considered. G may be
reduced.
[0026]
The multiple channel generation unit 520 uses the signal Mo (n) obtained by the common
component noise suppression unit 320 and the signal So (n) obtained by the non-common
component processing unit 420 to generate a plurality of input channels. Output signals Out (n,
k) (k = 1,..., Kout: Kout: total number of output channels, n) where noises of common components
are removed from the signal In (n, k) and non-common components are attenuated or
emphasized Outputs the time).
[0027]
Here, the values of the input channel numbers Kin and Kout may be two or more values, but to
simplify the following description, Kin = Kout = 2, that is, stereo input and stereo output as an
example. I will list and explain.
In this case, for example, the configuration of the noise suppression apparatus shown in FIG. 3
may be configured as shown in FIG.
[0028]
The common component extraction unit 121 receives the signals In (n, 1) and In (n, 2) of the two
08-05-2019
8
input channels as input, and the component Mi (n) commonly included in the two input channels
can Extract by the operation shown in 1). That is, by adding the signals In (n, 1) and In (n, 2) of
the two input channels, a signal in which the components having common phases become more
prominent is obtained.
[0029]
The non-common component extraction unit 221 receives signals In (n, 1) and In (n, 2) of two
input channels, and extracts a component Si (n) that is not common to the two input channels.
For example, it extracts by calculation as shown to above-mentioned Formula (2). That is, by
subtracting the signals In (n, 1) and In (n, 2) of the two input channels, a signal in which
components whose phases are not in common become more prominent is obtained.
[0030]
The monaural component noise suppression unit 321 performs noise suppression processing on
the component Mi (n) extracted by the common component extraction unit 121, obtains a signal
Mo (n), and outputs this. Here, a method similar to the monaural component noise suppression
unit 311 can be considered as a method of the noise suppression processing.
[0031]
The monaural component noise suppression unit 321 also detects the level of the noise signal
included in the component Mi (n), that is, the noise signal level of the monaural component, and
notifies the stereo component processing unit 421 of this.
[0032]
The stereo component processing unit 421 performs predetermined level adjustment processing
on the component Si (n) extracted by the non-common component extraction unit 221 with an
intensity corresponding to the level notified from the monaural component noise suppression
unit 321 ( For example, attenuation or enhancement is performed to obtain a signal So (n) and
output it.
08-05-2019
9
Here, in the case of attenuation, the gain G is used to perform the calculation according to the
above-mentioned equation (3). As a specific method of determining the gain G, for example, G =
α × Mi_SNR may be calculated using the SNR (signal to noise ratio) of the component Mi (n).
Here, α is a constant, and Mi_SNR is the SNR of the component Mi (n). Alternatively, from the
SNR of the component Mi (n) (or a priori SNR disclosed in the above-mentioned MMSE-STSA
method literature), the spectral subtraction method, the MMSE-STSA method, the Wiener filter
method disclosed in the above-mentioned document The gain for noise suppression may be
calculated using and this gain may be used as gain G.
[0033]
On the contrary, the noise level of the component which is not common is detected in the stereo
component processing unit 421, and when the noise is large, the noise G of the common
component noise suppression unit 320 is reduced considering that the noise of the common
component is also large. You may do so. For example, G = N_c × (N_si to the power of p) may be
set. Here, N_si is a noise level of the component Si (n), and p and N_c are constants. For example,
it may be p = −1, N_c = −16 [dB]. However, the value of G is limited to a value from 0 to 1.
[0034]
The multiple channel generation unit 521 includes a channel generation unit 521 a and a
channel generation unit 521 b. The channel generation unit 521 a and the channel generation
unit 521 b use the signal Mo (n) obtained by the monaural component noise suppression unit
321 and the signal So (n) obtained by the stereo component processing unit 421 respectively.
Common signal noise is removed from the signals In (n, 1) and In (n, 2) of one input channel, and
output signals Out (n, 1) and Out (n, 2) with attenuated or emphasized non-common components.
Find and output). For example, it calculates | requires and outputs by calculation as shown to
above-mentioned Formula (4), (5).
[0035]
In order to match the amplitude between the signals In (n, 1) and In (n, 2) of the two input
channels and the respective output signals Out (n, 1) and Out (n, 2) , Out_n (n, 1), instead of
outputting Out (n, 1), Out (n, 2) by adding a configuration for performing the operation as shown
in the above equations (6), (7) You may output Out_new (n, 2).
08-05-2019
10
[0036]
As described above, in the noise suppression device of the above configuration, noise of a sound
whose sound image is in the center direction among the signals of two input channels, that is,
stereo signals, is suppressed. An effect that makes the announcer's voice easy to hear is obtained,
and the stereo component can be attenuated or emphasized separately from the monaural
component.
For this reason, if a stereo component is subjected to, for example, gain adjustment to the extent
of adjusting the volume, it is possible to hear a natural sound image.
[0037]
Also, since noise suppression and gain adjustment of the common component and the noncommon component are performed based on the correlation between the common component
and the non-common component, noise contained in the common component is more naturally
suppressed. Or, the sound image of the non-common component can be kept more natural, and
the effect of being easy to hear is obtained. For example, in the case of 2-channel input, it is
possible to obtain an effect that it becomes easy to hear the announcer's voice described in a
sports broadcast program or the like.
[0038]
Next, a noise suppression apparatus according to a third embodiment of the present invention
will be described. FIG. 5 shows the configuration. The noise suppression apparatus shown in this
figure includes a common component extraction unit 130, a common component noise
suppression unit 330, a multiple channel generation unit 530, and gain adjustment units 630a
and 630b.
[0039]
The common component extraction unit 130 receives a plurality of input channel signals In (n, k)
(k = 1, ..., Kin: Kin is the total number of input channels and n is time), and is common to a
plurality of input channels. Extract the component Mi (n) contained.
08-05-2019
11
[0040]
The common component noise suppression unit 330 performs noise suppression processing on
the component Mi (n) extracted by the common component extraction unit 130, obtains a signal
Mo (n), and outputs this.
In addition, the common component noise suppression unit 330 detects the level of the noise
signal included in the component Mi (n), that is, the noise signal level of the common component,
and notifies the gain adjustment units 630a and 630b of this.
[0041]
The gain adjusting units 630a and 630b perform signal level suppression on the signal In (n, k)
of each input channel with a gain G according to the noise signal level notified from the common
component noise suppressing unit 330. As a specific calculation method of the gain G, for
example, the calculation method of the gain G described in the description of the stereo
component processing unit 421 in the second embodiment can be considered.
[0042]
A noise signal contained in a component common to a plurality of input channels and a noise
signal contained in the signal of each input channel are considered to be correlated to some
extent. The noise suppression apparatus according to the third embodiment focuses on this point,
and for example, when the noise of a component common to a plurality of input channels is
large, the noise included in the signal of each input channel is also large. In consideration of this,
the suppression in the gain adjustment units 630a and 630b is made stronger.
[0043]
Alternatively, as another configuration, the gain adjusters 630a and 630b detect the noise level
of each input channel, and when the noise level of the input channel is large, the common
component noise suppressor 330 considers that the noise of the common component is also
08-05-2019
12
large. The gain G in may be reduced. As a specific calculation method of the gain G, for example,
the calculation method of the gain G described in the description of the stereo component
processing unit 421 in the second embodiment can be considered.
[0044]
The multiple channel generation unit 530 uses the signal Mo (n) obtained by the common
component noise suppression unit 330 and the signals obtained by the gain adjustment units
630 a and 630 b to generate the signal In (n, n) of each input channel. k) remove the noise of the
common component from each of the output signals Out (n, k) (k = 1,..., Kout: Kout is the total
number of output channels and n is the time) Output each.
[0045]
Here, the values of the input channel numbers Kin and Kout may be two or more values, but to
simplify the following description, Kin = Kout = 2, that is, stereo input and stereo output as an
example. I will list and explain.
In this case, for example, the configuration of the noise suppression apparatus shown in FIG. 5
may be configured as shown in FIG.
[0046]
The common component extraction unit 131 receives the signals In (n, 1) and In (n, 2) of the two
input channels as input, and the component Mi (n) commonly included in the two input channels
can Extract by the operation shown in 1).
[0047]
The common component noise suppression unit 331 performs noise suppression processing on
the component Mi (n) extracted by the common component extraction unit 131, obtains a signal
Mo (n), and outputs this.
Here, a method similar to the monaural component noise suppression unit 311 can be
considered as a method of the noise suppression processing.
08-05-2019
13
[0048]
Further, the common component noise suppression unit 331 detects the level of the noise signal
included in the component Mi (n), that is, the noise signal level of the monaural component, and
notifies this to the gain adjustment units 631a and 631b.
[0049]
The gain adjusting units 631a and 631b perform signal level suppression on the signal In (n, k)
of each input channel with a gain G according to the noise signal level notified from the common
component noise suppressing unit 331.
[0050]
Further, as another configuration, the noise level of the signal of each input channel is detected
by the gain adjustment units 631a and 631b, and when the noise level is large, the common
component noise suppression unit 331 is considered to have a large noise of common
components. The gain G in may be reduced.
[0051]
The multiple channel generation unit 531 includes a channel generation unit 531a and a channel
generation unit 531b.
The channel generation unit 531a uses the signal Mo (n) obtained by the common component
noise suppression unit 331 and the signal obtained by the gain adjustment unit 631a to common
to the signal In (n, 1) of the input channel. The output signal Out (n, 1) from which the
component noise has been removed is obtained and output.
Similarly, the channel generation unit 531 b uses the signal Mo (n) obtained by the common
component noise suppression unit 331 and the signal obtained by the gain adjustment unit 631
b to generate the signal In (n, 2) of the input channel. The output signal Out (n, 2) from which the
noise of the non-common component has been removed is calculated and output.
[0052]
08-05-2019
14
In order to match the amplitude between the signals In (n, 1) and In (n, 2) of the two input
channels and the respective output signals Out (n, 1) and Out (n, 2) , Out_n (n, 1), instead of
outputting Out (n, 1), Out (n, 2) by adding a configuration for performing the operation as shown
in the above equations (6), (7) You may output Out_new (n, 2).
[0053]
As described above, in the noise suppression device of the above configuration, noise of a sound
whose sound image is in the center direction among the signals of two input channels, that is,
stereo signals, is suppressed. An effect that makes the announcer's voice easy to hear is obtained.
[0054]
Further, since noise suppression and gain adjustment of signals of the common component and
the input channel are performed based on the correlation between the common component and
the signal of each input channel, noise contained in the common component is made more
natural. Or the sound image of the signal of each input channel can be kept more natural, and the
effect of being easy to hear can be obtained.
For example, in the case of 2-channel input, it is possible to obtain an effect that it becomes easy
to hear the announcer's voice described in a sports broadcast program or the like.
[0055]
The present invention is not limited to the above embodiment as it is, and at the implementation
stage, the constituent elements can be modified and embodied without departing from the scope
of the invention.
For example, the idea described in the above equations (1) to (7) can be applied not only to the
time domain but also to the frequency domain.
In this case, the calculation for each sample can be realized by replacing the calculation for each
spectrum or the calculation for each frequency band.
08-05-2019
15
[0056]
FIG. 1 is a circuit block diagram showing a configuration of a first embodiment of a noise
suppression device according to the present invention. The circuit block diagram which shows
the structural example at the time of making an input signal into 2 systems by the noise
suppression apparatus shown in FIG. The circuit block diagram which shows the structure of 2nd
Embodiment of the noise suppression apparatus concerning this invention. The circuit block
diagram which shows the structural example at the time of making an input signal into 2 systems
by the noise suppression apparatus shown in FIG. The circuit block diagram which shows the
structure of 3rd Embodiment of the noise suppression apparatus concerning this invention. The
circuit block diagram which shows the structural example at the time of making an input signal
into 2 systems by the noise suppression apparatus shown in FIG.
Explanation of sign
[0057]
110, 111, 120, 121, 130, 131 ... common component extraction unit, 210, 211, 220, 221 ... noncommon component extraction unit, 310, 320, 330 ... common component noise suppression
unit, 311, 321, 331 ... monaural Component noise suppression unit, 410, 420 ... non-common
component processing unit, 411, 421 ... stereo component processing unit, 510, 511, 520, 521,
530, 531 ... multiple channel generation unit, 511a, 511b, 521a, 521b, 531a , 531b ... channel
generation unit, 630a 631a ... gain adjustment unit.
08-05-2019
16
Документ
Категория
Без категории
Просмотров
0
Размер файла
27 Кб
Теги
jp2008283385
1/--страниц
Пожаловаться на содержимое документа