close

Вход

Забыли?

вход по аккаунту

?

JP2010156742

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 JP2010156742
A signal processing apparatus and method capable of adaptively correcting an input signal when
a signal acquisition situation changes over a wide range and suppressing other signals mixed in a
desired signal with a small amount of calculation. Do. A main signal acquisition unit mainly
acquires a target signal, reference signal acquisition units 2-1 to 2-n mainly acquire signals other
than a target, and a main signal acquired by the main signal acquisition unit Signal suppression
means for generating a noise suppression signal in which signal components other than the
target contained in the main signal are suppressed using the signal and the reference signal
acquired by the reference signal acquisition means, and correction means for correcting each
reference signal And the correction means 3 updates the correction content based on the noise
suppression signal. [Selected figure] Figure 1
Signal processing apparatus and method
[0001]
The present invention relates to a signal processing apparatus and method, and to a signal
processing apparatus and method for suppressing other signals mixed in with a desired signal or
enhancing a desired signal.
[0002]
A voice signal input from a microphone, a handset or the like is subjected to voice coding and
voice recognition processing.
04-05-2019
1
Background noise signals mixed in such speech signals pose a serious problem when performing
speech coding and speech recognition in a narrowband speech coding apparatus, speech
recognition apparatus or the like with a high degree of information compression. As a signal
processing method aiming at suppression of such an acoustically superimposed unnecessary
component, for example, Patent Document 1 describes a method of suppressing an unnecessary
component by an adaptive filter with a plurality of input signals as inputs. There is.
[0003]
FIG. 4 is a diagram showing a typical configuration of a conventional signal processing
apparatus. This signal processing apparatus includes a main signal acquisition unit 1, a plurality
of reference signal acquisition units 2-1 to 2-n, adaptive filters 45-1 to 45-n, a subtraction unit 6,
and an output unit 7. Have. Here, n is a value indicating the number of reference signal
acquisition means.
[0004]
At time k, the main signal acquisition means 1 receives a signal x (k) that has been A / electrically
converted by a microphone or the like placed in the vicinity of the speaker, and the reference
signal acquisition means 2-1 to 2-n Signals y1 (k) to yn (k) acoustoelectrically converted are
input by a microphone or the like placed near the speaker.
[0005]
The adaptive filters 45-1 to 45-n perform filter operations with the input signals y1 (k) to yn (k)
as inputs, and output pseudo noise signals r1 (k) to rn (k) as operation results.
[0006]
The subtraction means 46 subtracts the pseudo noise signals r1 (k) to rn (k) from the main signal
x (k) to generate a difference signal e (k) and outputs it as an output signal to the output means 7
as well as adaptively The adaptive filters 45-1 to 45-n are supplied as error signals for updating
the coefficients of the filters 45-1 to 45-n.
[0007]
04-05-2019
2
The adaptive filters 45-1 to 45-n use the coefficient update algorithm to update the filter
coefficients based on the input error signals.
Here, assuming that the LMS algorithm described in Non-Patent Document 1 is used as the
coefficient updating algorithm of the adaptive filter, and the j-th coefficient wij (k) of the adaptive
filter 45-i at time k, the adaptive filter 45-i The pseudo noise signal ri (k) to be output is
expressed by equation (1).
[0008]
[0009]
Here, Tr (i) is a constant called the number of taps of the adaptive filter 45-i, which is a
parameter indicating the size of the filter.
Also, updating of the coefficient is performed according to equation (2).
[0010]
[0011]
Here, μ r (i) is a constant called step size, which is a parameter for determining the convergence
time of the coefficient and the residual error after convergence.
[0012]
Further, in such a method using a plurality of input signals, it is important to optimize the input
signal in accordance with the difference in characteristics between the input terminals.
As a method of correcting such a plurality of input signals, for example, Patent Document 2
describes a method of eliminating variations in input signals caused by variations in manufacture
of microphone elements, wiring at the time of attachment, and the like.
04-05-2019
3
[0013]
FIG. 5 is a view showing a typical configuration of a conventional input signal correction
apparatus.
The signal correction apparatus includes a plurality of signal acquisition units 2-1 to 2-n,
correction units 53-1 to 53-n, a signal processing unit 55, and an output unit 7.
[0014]
At time k, the signals y1 (k) to yn (k) acoustoelectrically converted by the microphone or the like
are input to the signal acquisition units 2-1 to 2-n.
[0015]
The correction means 53-1 to 53-n perform correction calculations with the input signals y1 (k)
to yn (k) as inputs, and the correction input signals y'1 (k) to y'n (k) as calculation results. Output
[0016]
The signal processing unit 55 receives the corrected input signals y'1 (k) to y'n (k), performs
signal processing operation, generates an output signal e (k) as the operation result, and outputs
it to the output means 7 Do.
JP-A-11-18194 “Microphone array device” JP-A-2002-99297 “Microphone device” Jin-ichi
Nagumo and Atsuhiko Noda, “A Learning Method for System Identification,” IEEE Transactions
on Automatic Control, VOL. 12, No. 3, 1967, pp. 282-287
[0017]
However, in the conventional signal processing method using a plurality of adaptive filters, it is
necessary to perform the arithmetic processing of the adaptive filter as shown in equations (1)
and (2) a number of times according to the number of input signals. The problem is that the
04-05-2019
4
amount of computation increases with the number of
[0018]
Also, the conventional input signal correction method acts to match the characteristics of each
signal based on a predetermined value, and adaptively according to the position and volume of
the sound source that changes widely. There is a problem that correction such as weighting can
not be performed on a specific input signal.
[0019]
Therefore, it is an object of the present invention to be able to perform correction of an input
signal adaptively when the signal acquisition situation changes over a wide range, and to perform
signal processing that can suppress other signals mixed in a desired signal with a small amount
of calculation. An apparatus and method are provided.
[0020]
In order to solve the above problems, the present invention is configured as follows.
The invention according to claim 1 is a main signal acquisition means for acquiring a target
signal and outputting it as a main signal, and at least two or more reference signal acquisition
means for acquiring a signal other than the target and outputting it as a reference signal. And
first correction means for correcting the reference signal to output a correction reference signal,
and using the main signal and the correction reference signal to suppress noise other than the
target signal component included in the main signal. A signal processing apparatus comprising:
signal suppression means for outputting a signal, wherein the first correction means updates the
correction content based on the noise suppression signal.
Further, in the invention according to claim 2, the signal suppression means is an adaptive filter
which receives the correction reference signal and outputs a pseudo noise signal, and a
difference between the main signal and the pseudo noise signal as the noise suppression signal.
And subtracting means for outputting as.
The invention according to claim 3 is characterized in that the first correction means corrects the
amplitude of the reference signal.
04-05-2019
5
The invention according to claim 4 is characterized in that the first correction means corrects the
phase of the reference signal.
In the invention according to claim 5, at least two main signal acquisition means for acquiring a
target signal and outputting it as a main signal, and at least two for acquiring a signal other than
the target and outputting it as a reference signal One or more reference signal acquisition means,
a first correction means for correcting the reference signal to output a correction reference
signal, a second correction means for correcting the main signal to output a correction main
signal, and First signal suppressing means for outputting a noise suppression signal in which a
signal component other than the target included in the correction main signal is suppressed
using the correction main signal and the correction reference signal, the correction reference
signal and the correction main signal And second signal suppression means for outputting a
target suppression signal obtained by suppressing a target signal component included in the
correction reference signal using the first correction means is the noise suppression signal. To
Zui updates the correction contents, the second correction means is characterized in updating the
correction contents on the basis of the target reduction signal.
Also, in the invention according to claim 6, the first signal suppression means is a first adaptive
filter which receives the target suppression signal and outputs a pseudo noise signal, and the
correction main signal and the pseudo noise signal. A first adaptive means for outputting a
difference as the noise suppression signal, wherein the second signal suppression means is a
second adaptive filter for receiving the noise suppression signal and outputting a pseudo target
signal, and the correction reference And second subtraction means for outputting a difference
between the signal and the pseudo target signal as the target suppression signal. The invention
according to claim 7 further comprises signal-to-noise ratio estimation means for calculating a
signal-to-noise ratio from the noise suppression signal and the target suppression signal, wherein
the first correction means and the second correction means It is characterized in that the means
determines whether to update the correction content based on the signal-to-noise ratio. The
invention according to claim 8 is characterized in that the first correction means corrects the
amplitude of the reference signal, and the second correction means corrects the amplitude of the
main signal. It is. The invention according to claim 9 is characterized in that the first correction
means corrects the phase of the reference signal, and the second correction means corrects the
phase of the main signal. It is. The invention according to claim 10 is a main signal acquisition
means for acquiring a target signal and outputting it as a main signal, and at least two or more
reference signals for acquiring a signal other than the target and outputting it as a reference
signal. The signal component other than the target included in the main signal is suppressed
using the acquiring unit, the first correcting unit that corrects the reference signal and outputs a
04-05-2019
6
correction reference signal, and the main signal and the correction reference signal. A signal
processing apparatus comprising: a signal suppression unit that outputs a noise suppression
signal, wherein the first correction unit updates the correction content based on the noise
suppression signal. The signal suppression means may be an adaptive filter that receives the
correction reference signal and outputs a pseudo noise signal, and a difference between the main
signal and the pseudo noise signal as the noise suppression signal. And subtracting means for
outputting as. The invention according to claim 12 is characterized in that the first correction
means corrects the amplitude of the reference signal. The invention according to claim 13 is
characterized in that the first correction means corrects the phase of the reference signal. The
invention according to claim 14 is at least two main signal acquisition means for acquiring a
target signal and outputting it as a main signal, and at least two for acquiring a signal other than
the target and outputting it as a reference signal. One or more reference signal acquisition
means, a first correction means for correcting the reference signal to output a correction
reference signal, a second correction means for correcting the main signal to output a correction
main signal, and First signal suppressing means for outputting a noise suppression signal in
which a signal component other than the target included in the correction main signal is
suppressed using the correction main signal and the correction reference signal, the correction
reference signal and the correction main signal And second signal suppression means for
outputting a target suppression signal obtained by suppressing a target signal component
included in the correction reference signal using the first correction means is the noise
suppression signal. Based update the correction contents, the second correction means is
characterized in updating the correction contents on the basis of the target reduction signal.
The invention as set forth in claim 15 is characterized in that the first signal suppression means
is a first adaptive filter which receives the target suppression signal and outputs a pseudo noise
signal, and the correction main signal and the pseudo noise signal. A first adaptive means for
outputting a difference as the noise suppression signal, wherein the second signal suppression
means is a second adaptive filter for receiving the noise suppression signal and outputting a
pseudo target signal, and the correction reference And second subtraction means for outputting a
difference between the signal and the pseudo target signal as the target suppression signal. The
invention according to claim 16 is at least two main signal acquisition means for acquiring a
target signal and outputting it as a main signal, and at least two for acquiring a signal other than
the target and outputting it as a reference signal. Using one or more reference signal acquisition
means, a first correction means for correcting the reference signal and outputting a correction
reference signal, and using the main signal and the correction reference signal, other than the
target included in the main signal A first signal suppression means for outputting a noise
suppression signal in which a signal component is suppressed, a second correction means for
correcting the main signal and outputting a correction main signal, the reference signal and the
correction main signal A second signal suppression means for outputting a target suppression
signal obtained by suppressing a target signal component included in the reference signal; and a
04-05-2019
7
signal-to-noise ratio estimation means for calculating a signal-to-noise ratio from the noise
suppression signal and the target suppression signal. In the signal processing apparatus, the first
correction unit and the second correction unit determine whether to update the correction
content based on the signal-to-noise ratio. . The invention according to claim 17 is characterized
in that the first correction means corrects the amplitude of the reference signal, and the second
correction means corrects the amplitude of the main signal. It is. The invention according to
claim 18 is characterized in that the first correction means corrects the phase of the reference
signal, and the second correction means corrects the phase of the main signal. It is.
[0021]
According to the present invention, the correction means determines the correction content of
the input signal acquired by the signal acquisition means so that the noise suppression amount of
the output signal is maximized, thereby making it possible to intervene between the signal
acquisition means with respect to the input signal. It becomes possible to perform correction so
that the amount of noise suppression is maximized according to the position of the sound source
and the volume changing widely, regardless of the characteristic difference, and noise
components can be suppressed with high accuracy. Also, by performing only correction on the
input signal, it is possible to realize signal processing with a small amount of calculation.
[0022]
[0023]
A first embodiment of the present invention will be described in detail with reference to the
drawings.
Hereinafter, a signal processing method according to the present invention will be described by
way of an example realized as a method of processing an acoustic signal. However, the signal
processing methods of the following embodiments can be used for various signals other than the
acoustic signal without changing the configuration. For example, the various signals include an
antenna signal, an image signal, a motor drive control signal, and the like. DESCRIPTION OF A
STRUCTURE FIG. 1: is a block diagram which shows the structure of the signal processing
apparatus of the 1st Embodiment of this invention. This signal processing apparatus comprises a
main signal acquisition unit 1, n reference signal acquisition units 2-1 to 2-n, a correction unit 3,
04-05-2019
8
an averaging unit 4, an adaptive filter 5, a subtraction unit 6, and an output. And at least a means
7. [Description of operation]
[0024]
Next, the operation of the present invention at time k will be described. The main signal
acquisition unit 1 mainly acquires an acoustic signal coming from a target sound source, and
outputs an acoustic-electrically converted main signal x (k). The main signal acquisition unit 1 is,
for example, a voice input device such as a microphone.
[0025]
The reference signal acquisition means 2-1 to 2-n mainly acquire the acoustic signal coming
from the noise source, and output the reference signals y1 (k) to yn (k) subjected to the acousticelectrical conversion. The reference signal acquisition units 2-1 to 2-n are, for example, voice
input devices such as a microphone. Here, n is an integer of 2 or more, and the reference signal
acquisition units 2-1 to 2-n are configured by at least two voice input devices.
[0026]
The correction means 3 receives the reference signals y1 (k) to yn (k), performs a correction
operation on each reference signal, and outputs a correction reference signal y'1 (k) to y'n (k) as
a calculation result. Output.
[0027]
In this embodiment, as an example of the correction operation, as represented by equation (11),
the reference signals y (k) to yn (k) multiplied by the correction coefficients α1 (k) to αn (k) Are
set as the correction reference signals y'1 (k) to y'n (k).
The correction coefficients α1 (k) to αn (k) are coefficients to be updated, and the details will be
described later.
04-05-2019
9
[0028]
<img class = "EMIRef" id = "204236817-000006" />
[0029]
The averaging means 4 receives the correction reference signals y'1 (k) to y'n (k), averages the
correction reference signals, and outputs an average reference signal Y (k) as a calculation result.
In the present embodiment, as an example of averaging, the sum of the correction reference
signals y'1 (k) to y'n (k) is divided by the number n of reference signals as represented by
equation (12). A signal is taken as an average reference signal Y (k).
[0030]
[0031]
The adaptive filter 5 receives the average reference signal Y (k), performs a filter operation, and
outputs a pseudo noise signal r (k) as the operation result.
[0032]
The subtracting means 6 receives the main signal x (k) and the pseudo noise signal r (k),
subtracts the pseudo noise signal r (k) from the main signal x (k), and generates an error main as
a calculation result. Output the signal es (k).
Here, the adaptive filter 5 and the subtraction means 6 constitute a first signal suppression
means.
[0033]
[0034]
04-05-2019
10
The output terminal 7 outputs the error main signal es (k) output from the first signal
suppression unit to the outside as a noise suppression signal.
[0035]
The adaptive filter 5 receives the error main signal es (k) as an input and updates the coefficients
of the filter using a coefficient update algorithm.
In this embodiment, the LMS algorithm is assumed as an example of the coefficient updating
algorithm of the adaptive filter, and the pseudo noise signal r output from the adaptive filter 5 is
assumed to be the j-th coefficient wj (k) of the adaptive filter 5 at time k. (k) is expressed by
equation (14).
[0036]
[0037]
Here, Tr is a constant called the number of taps of the adaptive filter 5 and is a parameter
indicating the size of the filter.
Also, updating of the coefficient is performed according to equation (15).
[0038]
[0039]
Here, μ r is a constant called step size, which is a parameter for determining the convergence
time of the coefficient and the residual error after convergence.
[0040]
04-05-2019
11
The correction means 3 receives the error main signal es (k) which is the output of the first
signal suppression means and updates the correction coefficients α1 (k) to αn (k) using a
coefficient update algorithm.
In this embodiment, the LMS algorithm is assumed as an example of the correction coefficient
update algorithm, and the correction coefficient αi (k) for the i-th reference signal yi (k) at time k
is updated according to equation (16) It will be.
[0041]
[0042]
Here, λα is a step size for updating the correction coefficient.
In the equation (16), instead of the error main signal e (k), the instantaneous error main signal
ess (k) represented by the equation (17) may be used.
[0043]
[0044]
In the present embodiment, the correction means 3 is configured to correct the amplitude by
handling only the real value of each signal, but it is also possible to correct the phase by the
same configuration by performing complex number conversion and treating it as a complex
number. It is also possible.
Also, any coefficient update algorithm may be configured as long as it uses an algorithm that
performs coefficient update adaptively, and is not limited to the LMS algorithm.
Furthermore, the updating of the coefficients of the adaptive filter in the adaptive filter 5 and the
updating of the correction coefficient in the correction means 3 are not performed every
04-05-2019
12
acquisition cycle of the signal, but are configured to perform update processing every specific
cycle. The configuration may be such that the update process is performed only when the
number is exceeded.
[0045]
The signal processing apparatus according to the first embodiment of the present invention
updates the correction coefficient for each reference signal such that the suppression amount of
the noise component for the main signal is maximized by the configuration and the operation as
described above. This makes it possible to realize highly accurate noise suppression regardless of
variations in microphone elements, wiring, etc., and changes in conditions such as the position of
the noise source and the volume.
Further, since only one correction coefficient is provided for one reference signal, it is possible to
reduce the amount of calculation compared to the case of performing multiple coefficient
updating and filter operation of the adaptive filter for one reference signal. .
[0046]
Next, a second embodiment of the present invention will be described in detail with reference to
the drawings.
[Description of configuration]
[0047]
FIG. 2 is a block diagram showing the configuration of a signal processing apparatus according to
a second embodiment of the present invention. This signal processing apparatus includes m main
signal acquisition units 21-1 to 21-m instead of the main signal acquisition unit 1 as compared
with the signal processing apparatus according to the first embodiment shown in FIG. ing. Here,
m is an integer of 2 or more. Furthermore, the second correction means 23, the second averaging
means 24, the second adaptive filter 25, and the subtraction means 26 are added.
04-05-2019
13
[0048]
In this figure, elements which are the same as or equivalent to the elements shown in FIG.
Therefore, in the following, the description of the same elements as those in FIG. 1 which
perform the same operation is omitted, and the added components and the different elements of
the operation will be described. [Description of operation]
[0049]
At time k, the main signal acquisition units 21-1 to 21-m acquire an acoustic signal that mainly
comes from the target sound source by using, for example, a microphone and main signal x1 (k)
to xm (k) obtained by acoustic-electrical conversion. Output
[0050]
The correction unit 23 receives the main signals x1 (k) to xm (k), performs correction calculation
of each main signal, and calculates a corrected main signal x'1 (k) to x'n (k) as a calculation
result. Output.
In the present embodiment, as an example of the correction operation, as represented by
equation (21), the main signals x1 (k) to xm (k) multiplied by the correction coefficients β1 (k) to
βm (k) And the corrected main signal x'1 (k) to x'm (k).
[0051]
<img class = "EMIRef" id = "204236817-000014" />
[0052]
The averaging means 24 receives the corrected main signals x'1 (k) to x'm (k), averages the
respective corrected main signals, and outputs an average main signal X (k) as a calculation
result.
04-05-2019
14
In the present embodiment, the sum of the corrected main signals x'1 (k) to x'm (k) is divided by
the number m of main signals as an example of averaging, as expressed by equation (22). A
signal is taken as an average reference signal X (k).
[0053]
[0054]
The adaptive filter 25 receives the error main signal es (k), performs a filter operation, and
outputs a pseudo target signal s (k) as the operation result.
[0055]
The subtraction means 26 receives the average reference signal Y (k) and the pseudo target
signal s (k), and the pseudo target signal is calculated from the average reference signal Y (k) as
expressed by equation (23). s (k) is subtracted, and an error reference signal er (k) (target
suppression signal) is output as a calculation result.
Here, the adaptive filter 25 and the subtracting means 26 constitute a second signal suppressing
means.
[0056]
[0057]
The adaptive filter 25 receives the error reference signal er (k) output from the second signal
suppression unit as an input, and uses the coefficient update algorithm to update the filter
coefficients.
In this embodiment, the LMS algorithm is assumed as an example of the coefficient updating
algorithm of the adaptive filter, and assuming that the j-th coefficient vj (k) of the adaptive filter
25 at time k, the pseudo target signal s output by the adaptive filter 25 (k) is expressed by
equation (24).
04-05-2019
15
[0058]
[0059]
Here, Ts is a constant called the number of taps of the adaptive filter 25 and is a parameter
indicating the size of the filter.
Also, updating of the coefficient is performed according to equation (25).
[0060]
[0061]
Here, μs is a constant called step size, which is a parameter for determining the convergence
time of the coefficient and the residual error after convergence.
[0062]
The adaptive filter 5 receives the error reference signal er (k) instead of the average reference
signal Y (k), performs a filter operation, and outputs a pseudo noise signal r (k) as a calculation
result.
At this time, the pseudo noise signal r (k) output from the adaptive filter 5 is expressed by
Expression (26) instead of Expression (14).
[0063]
[0064]
Also, updating of the coefficient is performed according to equation (27).
04-05-2019
16
[0065]
[0066]
The correction unit 23 receives the error reference signal er (k) (target suppression signal) and
updates the correction coefficients β1 (k) to βm (k).
In this embodiment, an LMS algorithm is assumed as an example of the correction coefficient
update algorithm, and the correction coefficient βi (k) for the i-th main signal xi (k) at time k is
updated according to equation (28) It will be.
[0067]
[0068]
Here, λβ is a step size for updating the correction coefficient.
The operation of the other components is the same as that of the first embodiment shown in FIG.
[0069]
The signal processing apparatus according to the second embodiment of the present invention
has each configuration such that the amount of suppression of the noise component with respect
to the main signal and the amount of suppression of the target component with respect to the
reference signal become maximum. It is possible to update the correction factor for the signal.
Therefore, highly accurate noise suppression can be realized regardless of variations in the
microphone elements, the wiring, etc., the noise source, the position of the target sound source,
the change in conditions such as the volume, etc.
04-05-2019
17
Also, there is only one correction factor for one input signal.
Therefore, it is possible to reduce the amount of calculation compared to the case of performing
a plurality of coefficient updating of the adaptive filter and the filter operation for one input
signal.
[0070]
Next, a third embodiment of the present invention will be described in detail with reference to
the drawings.
[Description of configuration]
[0071]
FIG. 3 is a block diagram showing the configuration of a signal processing apparatus according to
the third embodiment of the present invention. As compared with the signal processing device
according to the second embodiment shown in FIG. 2, this signal processing device has a
configuration in which a signal to noise ratio estimation means 31 is added. In this figure,
elements which are the same as or equivalent to the elements shown in FIG. Therefore, in the
following, the description of the same elements as in FIG. 2 and performing the same operations
is omitted, and the added components and the different elements of the operations will be
described. [Description of operation]
[0072]
At time k, the signal-to-noise ratio estimating means 31 receives the error main signal es (k) and
the error reference signal er (k), estimates the signal-to-noise ratio of the input signal, and
estimates the signal as an estimation result The noise to noise ratio SNR (k) is output. In the
present embodiment, as an example of a method of estimating the signal-to-noise ratio, as
represented by the equations (31) to (33), the error main signal es (k) and the error reference
signal er (k) Let the ratio of the average power from time kL to time k be the estimated signal-to-
04-05-2019
18
noise ratio SNR (k).
[0073]
[0074]
[0075]
[0076]
Note that the signal-to-noise ratio may be operated to determine the absolute amplitude ratio of
the signal instead of the power ratio.
[0077]
The correction means 3 receives the error main signal es (k) and the estimated signal-to-noise
ratio SNR (k) as input, and corrects the correction coefficient α1 (s) when the estimated signalto-noise ratio SNR (k) falls below the threshold value τr. k) Update ~ α n (k).
The coefficient update content is determined in the same manner as in the first embodiment.
[0078]
The correction means 23 receives the error reference signal e (k) and the estimated signal-tonoise ratio SNR (k) as input, and corrects the correction coefficient β 1 (s) when the estimated
signal-to-noise ratio SNR (k) exceeds the threshold value τs. k) Update .beta.m (k).
The coefficient update content is determined in the same manner as in the second embodiment.
[0079]
The operation of the other components is the same as that of the second embodiment shown in
04-05-2019
19
FIG.
[0080]
The signal processing apparatus according to the third embodiment of the present invention
updates the correction coefficient for the reference signal when the signal-to-noise ratio is small,
that is, when the noise signal is dominant, by the configuration and operation as described above.
And if the signal-to-noise ratio is large, i.e. if the target signal is dominant, update the correction
factor for the main signal.
Thereby, the amount of suppression of the noise component with respect to the main signal and
the amount of suppression of the target component with respect to the reference signal are
maximized while avoiding that the correction coefficient is updated for the unnecessary signal. It
becomes possible to update the correction coefficient.
As a result, highly accurate noise suppression can be realized regardless of variations in the
microphone elements, the wiring, etc., the noise source, the position of the target sound source,
the change in conditions such as the volume, etc.
Also, there is only one correction factor for one input signal. Therefore, it is possible to reduce
the amount of calculation compared to the case of performing a plurality of coefficient updating
of the adaptive filter and the filter operation for one input signal.
[0081]
As a configuration in which the step size of the adaptive filter of the present invention is made
variable according to the estimated signal-to-noise ratio or a configuration in which the pseudo
noise signal is output by another method such as fixed filter or independent component analysis
instead of adaptive filter. It is also good.
[0082]
It is a block diagram which shows the structure of the signal processing apparatus of the 1st
Embodiment of this invention.
04-05-2019
20
It is a block diagram which shows the structure of the signal processing apparatus of the 2nd
Embodiment of this invention. It is a block diagram which shows the structure of the signal
processing apparatus of the 3rd Embodiment of this invention. It is a block diagram which shows
the structure of the conventional signal processing apparatus. It is a block diagram which shows
the structure of the conventional signal correction apparatus.
Explanation of sign
[0083]
DESCRIPTION OF SYMBOLS 1 main signal acquisition means 2-1 to 2-n reference signal
acquisition means 3 correction means 4 averaging means 5 adaptive filter 6 subtraction means 7
output terminal 21-1 to 21-m main signal acquisition means 23 correction means 24 averaging
means 25 adaptive Filter 26 Subtraction means 31 Signal-to-noise ratio estimation means 45-1
to 45-n Adaptive filter 46 Subtraction means 53-1 to 53-n Correction means 55 Signal
processing unit
04-05-2019
21
Документ
Категория
Без категории
Просмотров
0
Размер файла
31 Кб
Теги
jp2010156742
1/--страниц
Пожаловаться на содержимое документа