close

Вход

Забыли?

вход по аккаунту

?

DESCRIPTION JP2005348173

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 JP2005348173
PROBLEM TO BE SOLVED: To provide a noise reduction device for reducing "smearing" of speech
and sufficiently suppressing noise. SOLUTION: A frequency band dividing unit 22 which converts
an input speech signal into a signal of a frequency domain and divides it into a plurality of
frequency bands, and an input speech signal which calculates an input speech signal power for
each frequency from a frequency band signal of the input speech signal. Power calculator 24,
noise power estimator 51 for estimating noise power for each frequency from input voice signal
power for each frequency, element for calculating an element gain factor from input voice signal
power for each frequency and noise power for each frequency Gain factor calculation unit 61,
gain factor smoothing unit 62 for smoothing an element gain factor, gain factor insertion for
calculating a frequency band signal of a noise reduction signal by superposing a smoothing gain
factor on a frequency band signal of an input speech signal And a time domain conversion unit
29 which reversely converts the frequency band signal of the noise reduction signal into the time
domain and outputs the result. Noise reduction device. [Selected figure] Figure 1
Noise reduction method, apparatus implementing the method, program and recording medium
therefor
[0001]
The present invention relates to a noise reduction method, an apparatus for carrying out the
method, a program and a recording medium therefor, and in particular, reduces voice signals
superimposed on target voice and collected by the microphone in voice communication using the
microphone. The present invention relates to a noise reduction method for improving quality, an
apparatus for implementing the method, a program, and a recording medium therefor.
10-04-2019
1
[0002]
A conventional example of reducing a noise signal from an input voice signal in which a noise
signal is superimposed on a target voice signal will be described with reference to FIG.
In FIG. 2, the input voice signal X (n) = S (n) + N (n) obtained by mixing unnecessary noise into
the target voice picked up by the microphone 11 is analog / digital (A / D) conversion The signal
is digitized in the unit 21 and input to the frequency band dividing unit 22. Here, S (n) represents
a target audio signal, and N (n) is an unnecessary noise signal mixed. Here, the input speech
signal X (n) after A / D conversion input to the frequency band division unit 22 is converted into
a signal in the frequency domain and then divided into a plurality of frequency bands. Each
frequency band signal divided by the frequency band division unit 22 is input to the input speech
signal power calculation unit 24 and the gain factor insertion unit 28. Hereinafter, the flow of
processing for the frequency band signal Xk (n) will be described by taking the k-th frequency
band signal of the input voice signal X (n) as a representative.
[0003]
The input speech signal power calculation unit 24 calculates the power level of the input
frequency band signal Xk (n), and inputs it to the S / N ratio estimation unit 27 and the noise
power estimation unit 51. The noise power estimation unit 51 estimates the noise power PNk (n)
using the input input audio signal power PXk (n). In the S / N ratio estimation unit 27, the input
voice signal power PXk (n), the estimated noise power PNk (n) and a signal Y′k (n) obtained by
reducing the noise obtained by the gain factor insertion unit 28 one processing frame before To
estimate the target speech signal to noise signal ratio (S / N ratio) SNRk (n). The S / N ratio SNRk
(n) estimated by the S / N ratio estimation unit 27 is input to the gain factor calculation unit 30
and the input speech signal addition rate determination unit 52. In gain factor calculation unit
30, gain factor G (SNR k (n)) is determined using S / N ratio SNR k (n) input from S / N ratio
estimation unit 27. Here, specific calculation of the gain factor G (SNRk (n)) is performed based
on spectral subtraction, Wiener filter, ML estimation method, MMSE method, and other short
time spectral amplitude (STSA) estimation. The gain factor G (SNR k (n)) estimated by the gain
factor calculation unit 30 is input to the gain factor insertion unit 28. The gain factor insertion
unit 28 performs noise reduction using the gain factor G (SNR k (n)). Specifically, calculation is
performed to superimpose (multiply) a gain factor G (SNR k (n)) in the frequency domain on the
frequency band signal X k (n) input from the frequency band division unit 22. Then, the noise
reduced signal Y′k (n) with noise reduced thus obtained is input to the input speech signal
adder 53 and the S / N ratio estimator 27. In parallel with this, the input speech signal addition
10-04-2019
2
rate determination unit 52 determines the input speech signal addition rate α based on the S / N
ratio using the input S / N ratio SNRk (n), and the input speech The signal is input to the signal
addition unit 53.
[0004]
In the input audio signal addition unit 53, the frequency band signal Xk (n) is added (added) to
the noise reduction signal Y'k (n) at a ratio according to the input audio signal addition rate α,
and the following frequency band signal Yk (N) is output (see Patent Document 1). Yk (n) = αXk
(n) + (1-α) Y'k (n) The frequency band signal Yk (n) is input to the time domain conversion unit
29, and all bands are combined and inverted to the time domain signal. It is converted and input
to a digital / analog (D / A) converter 34. Here, the noise-reduced output signal Y (n) converted
to an analog signal in the digital / analog converter 34 is output. Patent No. 3454402
[0005]
In the above conventional example, the gain factor G (SNR k (n)) is calculated using the S / N
ratio. The S / N ratio corresponds to the power ratio of the target audio signal and the noise
signal, but the target audio signal S (n) and the noise signal N (n) are mixed in the input audio
signal X (n). Therefore, each power can not be measured independently. Therefore, in Patent
Document 1, although a method of estimating each power and S / N ratio is adopted, estimation
error occurs in the estimated power and S / N ratio. Under the influence of this estimation error,
the calculated gain factor is not an ideal value based on the true S / N ratio. That is, since the
gain factors adjacent to each other in the frequency domain are intermittent (jumped) compared
to the ideal value based on the true S / N ratio, distortion is caused to the processed output audio
signal Y (n). Cause the problem.
[0006]
From the above, the present invention is a noise reduction method for suppressing the
occurrence of distortion of the output audio signal after processing and achieving sufficient noise
reduction even if there is an estimation error in the estimated power and S / N ratio, An
apparatus, a program, and a recording medium for implementing the method are provided.
[0007]
Claim 1: In a noise reduction method for reducing a noise signal from an input speech signal, the
10-04-2019
3
input speech signal is converted to a signal in the frequency domain, and the input speech signal
power for each frequency is calculated from the frequency band signal of the input speech signal
Noise power for each frequency is estimated from each input voice signal power, element gain
factor is calculated from input voice signal power for each frequency and noise power for each
frequency, element gain factor is smoothed, and smoothing gain factor is input A noise reduction
method is constructed in which the frequency band signal of the noise reduction signal is
calculated by being superimposed on the frequency band signal of the voice signal, and the
frequency band signal of the noise reduction signal is inversely converted into a time domain
signal and output.
[0008]
In a second aspect of the present invention, there is provided the noise reduction method
according to the first aspect of the present invention, in which smoothing processing is
performed to perform weighted averaging on the elementary gain factors.
Claim 3: In the noise reduction method according to any one of claims 1 and 2, the smoothing
gain factor is emphasized, and the emphasis gain factor is superimposed on the frequency band
signal of the input speech signal to generate noise. We constructed a noise reduction method to
calculate frequency band signal of reduced signal.
In a fourth aspect of the present invention, there is provided the noise reduction method
according to the third aspect, wherein the smoothing gain factor is enhanced to approach 0 or 1
depending on the magnitude of the value of the smoothing gain factor.
[0009]
Claim 5: In the noise reduction device for reducing noise signals from an input speech signal, the
frequency band division unit 22 converts the input speech signal into a signal in the frequency
domain and divides it into a plurality of frequency bands, and a frequency band signal of the
input speech signal Input voice signal power calculator 24 for calculating the input voice signal
power for each frequency from the noise, noise power estimation unit 51 for estimating noise
power for each frequency from the input voice signal power for each frequency, and input voice
signal power for each frequency Element gain factor calculating unit 61 that calculates an
element gain factor from noise power for each frequency, a gain factor smoothing unit 62 that
smoothes the element gain factor, and a smoothing gain factor superimposed on the frequency
band signal of the input voice signal Factor insertion unit 28 for calculating the frequency band
10-04-2019
4
signal of the noise reduction signal, and the frequency band signal of the noise reduction signal
To constitute a noise reduction apparatus comprising a time domain conversion unit 29 to
output the inverse transformation in the region.
[0010]
A sixth aspect of the present invention is the noise reduction apparatus according to the fifth
aspect of the present invention.
The noise reduction apparatus according to any one of claims 5 and 6, further comprising a gain
factor emphasizing unit 63 for emphasizing the smoothing gain factor, wherein the emphasis
gain factor is an input speech signal. The noise reduction apparatus is configured to calculate the
frequency band signal of the noise reduction signal by superimposing on the frequency band
signal of An eighth aspect of the present invention is the noise reduction apparatus according to
the seventh aspect, wherein the smoothing gain factor is enhanced to be close to 0 or 1
depending on the magnitude of the value of the smoothing gain factor.
[0011]
In a ninth aspect of the present invention, there is provided a noise reduction program in which
the noise reduction method according to any one of the first to fourth aspects is described by a
code written in and read out from a computer. Claim 10: A recording medium on which the noise
reduction program recited in claim 9 is recorded is configured.
[0012]
According to the present invention, by performing gain factor smoothing, intermittence in the
frequency domain of the gain factor is reduced and distortion of the output voice signal is
suppressed, and by emphasizing the gain factor after smoothing, By avoiding loss of part of the
frequency components of the processed voice, the effect of reducing "stagnation" of the voice and
suppressing noise sufficiently can be achieved.
[0013]
10-04-2019
5
The best mode for carrying out the present invention will be described with reference to FIG.
FIG. 1 is a diagram for explaining an embodiment of a noise reduction device, which is compared
with the noise reduction device described with reference to FIG. The other configurations are the
same except for the 30 configurations. From the above, the calculation of each value in the part
other than the gain factor calculation unit 30 in this embodiment can be executed according to
the method described in this patent document 1. An input voice signal X (n) = S (n) + N (n)
obtained by mixing unnecessary noise into the target voice collected by the microphone 11 is
digitalized in an analog / digital (A / D) converter 21. The signal is converted into a signal and
input to the frequency band dividing unit 22. Here, S (n) represents a target audio signal, and N
(n) is an unnecessary noise signal mixed. In the frequency band dividing unit 22, the input voice
signal X (n) is converted into a signal in the frequency domain and divided into a plurality of
frequency bands. The divided frequency band signals are input to the input speech signal power
calculation unit 24 and the gain factor insertion unit 28. Hereinafter, the flow of processing for
the k-th frequency band signal Xk (n) will be described with the k-th frequency band signal of the
input speech signal as a representative. The input speech signal power calculation unit 24
calculates the power level from the input frequency band signal Xk (n), and the input speech
signal power PXk (n) that is the calculation result is converted to the S / N ratio estimation unit
27, noise power Input to the estimation unit 51. The noise power estimation unit 51 estimates
the noise power PNk (n) using the input input audio signal power PXk (n). In the S / N ratio
estimation unit 27, a frequency band signal Y'k obtained by reducing the input speech signal
power PXk (n), the estimated noise power PNk (n) and the noise obtained by the gain factor
insertion unit 28 one processing frame before The signal-to-noise ratio SNRk (n) is estimated
using (n). The S / N ratio SNRk (n) estimated by the S / N ratio estimation unit 27 is input to the
gain factor calculation unit 30 and the input speech signal addition rate determination unit 52.
[0014]
Here, in the embodiment of the present invention, the gain factor calculation unit 30 comprises
an element gain factor calculation unit 61, a gain factor smoothing unit 62, and a gain factor
emphasis unit 63. In the gain factor calculation unit 30, first, the element gain factor calculation
unit 61 uses the S / N ratio SNRk (n) input from the S / N ratio estimation unit 27 to obtain an
element gain factor G (SNRk (n). ) Is calculated. The specific calculation of the element gain factor
is performed based on spectral subtraction, Wiener filter, ML estimation method, MMSE method
or other short time spectral amplitude (STSA) estimation. The gain factor is calculated based on
the S / N ratio, but specifically means the ratio of each frequency domain in the frequency
domain of the target voice to the input voice signal, and it is calculated as it is based on the S / N
ratio The gain factor of is referred to as the prime gain factor.
10-04-2019
6
[0015]
Next, the gain factor smoothing unit 62 performs smoothing with weighting given to the element
gain factor G (SNR k (n)). The element gain factor G (SNR k (n)) calculated by the element gain
factor calculation unit 61 is ideal due to the influence of the S / N ratio estimation error as
described in the section of [Problems to be solved by the invention]. Deviations from the value of
occur, causing discontinuities in each element gain factor in the frequency domain. To this end,
in this embodiment, smoothing is performed to smooth the value of each gain factor by reducing
the intermittence of element gain factors adjacent to each other on the frequency axis.
Hereinafter, the method of smoothing will be specifically described.
[0016]
Assuming that the element gain factor G (SNR k (n)) of the kth frequency band is replaced with G
(k), and the smoothing gain factor Ge (k) after smoothing, an example of the smoothing process is
It can be expressed by the following equation. Ge (k) =. SIGMA.i, ja (i) .times.G (j) /. SIGMA.ia (i)
This equation is an average value of a plurality of element gain factors G (j) adjacent to the kth
frequency band indicated by index j To obtain a smoothing gain factor Ge (k) of the kth
frequency band. The total number of i and j in summing is equal, and the total number is at most
equal to or less than the number of frequency analysis points. The weighting factor a (i) controls
the influence of each element gain factor when calculating the average value, that is, the rate of
alleviating intermittence. After the above processing, the smoothing gain factor Ge (k), that is, Ge
(SNRk (n)) is output.
[0017]
Next, in the gain factor emphasizing unit 63, the smoothing gain factor Ge (k) that has been
smoothed in the gain factor smoothing unit 62 is emphasized. Although the smoothing gain
factor Ge (k) has no discontinuities due to the smoothing, it has the negative effect of "rounding"
in the trade-off. Specifically, when the gain factor of the kth frequency band is 1, the gain factor
of the k-1st frequency band is 0.92, and the gain factor of the k + 2th frequency band is 0.93,
these three The smoothing gain coefficient of the kth frequency band from the average value in
the frequency band is 0.95. In this example, the weight coefficients a (i) are all one.
10-04-2019
7
[0018]
As described above, the gain factor is calculated based on the S / N ratio, and more specifically, is
the ratio of each frequency domain in the frequency domain of the target voice to the input voice
signal. That is, when the calculated gain factor is close to 1, it means that the noise is small in the
input speech signal and the proportion of the target speech is high, and when close to 0, the
target speech in the input speech signal is small and the proportion of the noise is small. It means
that there are many states. As explained above, when the smoothing gain factor is "rounded", its
value will be away from 0 and 1. When the signal is separated from 1, for example, when
smoothing to 0.95 as in the previous example, a 5% loss occurs in the target speech in order to
make the frequency domain component in which only the target speech originally exists is 95%.
When it is away from 0, for example, 0.05, when only noise is present and should be reduced by
100%, it is reduced by 95% and noise remains to adversely affect the speech quality. Therefore,
enhancement processing of the smoothed and "smoothed" smoothing gain factor Ge (k) shown
below is performed. Assuming that the gain factor is Gg (k) by emphasizing the k-th in the
frequency domain, the emphasizing process is a process in which each gain factor approaches 0
or 1 depending on the magnitude of the value of the smoothing gain factor Ge (k). is there. That
is, when the smoothing gain factor Ge (k) is close to one, the target voice is made to be closer to
one to make it easier to pass the target voice. When the smoothing gain factor Ge (k) is close to
zero, it is more zero. Emphasizing the smoothing gain factor Ge (k) to bring it closer and reduce
the noise more. One specific example of this emphasizing process is shown by the following
equation. When Ge (k) is larger than th1: Gg (k) = th1 × (Ge (k) / th1) <v1> When Ge (k) is
smaller than th2: Gg (k) = 1- (1-th2) {(1-Ge (k)) / (1-th2)} <v2> Here, v1 (k) and v2 (k) are integers
of 1 or more. Further, th1 and th2 are integers of 0 or more and 1 or less which satisfy the
relationship of th1 ≧ th2. Since Ge (k) has a value in the range of 0 to 1, this equation realizes
processing to approach 1 more if it is greater than th1 and approach 0 if it is less than th2. After
the above processing, an enhanced gain factor Gg (k), that is, Gg (SNRk (n)) is output.
[0019]
The enhanced gain factor G (SNR k (n)) calculated by the gain factor emphasizing unit 63 in the
gain factor calculating unit 30 is input to the gain factor inserting unit 28. The gain factor
insertion unit 28 performs noise reduction using the enhancement gain factor Gg (SNR k (n)).
Specifically, calculation is performed to superimpose (multiply) the emphasis gain factor Gg (SNR
k (n)) in the frequency domain on the frequency band signal X k (n) input from the frequency
band division unit 22. Then, the signal Y ′ k (n) with reduced noise is input to the input speech
signal addition unit 53 and the S / N ratio estimation unit 27. In parallel with this, the input
10-04-2019
8
speech signal addition rate determination unit 52 determines the input speech signal addition
rate α based on the S / N ratio using the input S / N ratio SNRk (n), and the input speech The
signal is input to the signal addition unit 53.
[0020]
In the input voice signal addition unit 53, the frequency band signal Xk (n) is added (added) to
the noise reduced signal Y'k (n) at a ratio according to the input voice signal addition rate α The
following frequency band signal Yk (n) as described above is output. Yk (n) = αXk (n) + (1-α) Y'k
(n) The frequency band signal Yk (n) is input to the time domain conversion unit 29, and all the
bands are combined and a signal in the time domain Is converted back to The inversely
converted time domain signal is input to a digital / analog converter 34, converted to an analog
signal, and output as a noise reduced output signal Y (n).
[0021]
By the way, the noise reduction device of the present invention can be configured by a DSP
(Digital Signal Processor). Alternatively, the program may be made to function by being executed
by a computer. In this case, the program is executed by loading the program recorded on a CDROM, a floppy (registered trademark) disk, a magnetic disk or the like into a program memory in
the computer. The program memory may be downloaded via communication.
[0022]
The figure explaining an Example. The figure explaining a prior art example.
Explanation of sign
[0023]
Reference Signs List 11 microphone 21 analog / digital converter 22 frequency band division
unit 24 input audio signal power calculation unit 27 S / N ratio estimation unit 28 gain factor
insertion unit 29 time domain conversion unit 30 gain factor calculation unit 34 digital / analog
converter 51 noise Power estimation unit 52 input speech signal addition rate determination unit
10-04-2019
9
53 input speech signal addition unit 61 elementary gain factor calculation unit 62 gain factor
smoothing unit 63 gain factor emphasis unit
10-04-2019
10
Документ
Категория
Без категории
Просмотров
0
Размер файла
20 Кб
Теги
description, jp2005348173
1/--страниц
Пожаловаться на содержимое документа