close

Вход

Забыли?

вход по аккаунту

?

JPH06292292

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 JPH06292292
[0001]
BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an
adaptive signal processing apparatus suitable for use in, for example, an audio pickup apparatus
of a camera integrated type VTR.
[0002]
2. Description of the Related Art Conventionally, an adaptive noise canceller as shown in FIG. 6
has been known as this type of adaptive signal processing apparatus.
[0003]
FIG. 6 shows a basic configuration of this adaptive noise canceller, in which 1 is a main input
terminal and 2 is a reference input terminal.
The main input signal input through the main input terminal 1 is supplied to the synthesis circuit
4 through the delay circuit 3. Further, the reference input signal input through the reference
input terminal 2 is supplied to the synthesis circuit 4 through the adaptive filter circuit 5 and is
subtracted from the signal from the delay circuit 3. The output of the synthesis circuit 4 is fed
back to the adaptive filter circuit 5 and is led to the output terminal 6.
03-05-2019
1
[0004]
In this adaptive noise canceller, the main input signal is the sum of the desired signal s and the
noise signal n0 uncorrelated with it. On the other hand, the noise signal n1 is input as the
reference input signal. The noise signal n1 of the reference input is uncorrelated with the desired
signal s, but is correlated with the noise signal n0.
[0005]
The adaptive filter circuit 5 filters the reference input noise signal n1 and outputs a signal y
which approximates the noise signal n0. In this case, the adaptive filter circuit 5 updates the
filtering coefficient of the reference input noise signal n1 by the predetermined adaptive
algorithm so that the subtraction output (residual output) e of the combining circuit 4 is
minimized. Go on.
[0006]
As the output signal y of the adaptive filter circuit 5, it is also possible to obtain a signal having
the same amplitude as that of the noise signal n0. The delay circuit 3 compensates for the time
delay required for the arithmetic processing in the adaptive filter circuit 5, the propagation time
in the adaptive filter, and the like, and is used to time align with the signal to be subtracted.
[0007]
The principle of the adaptive noise canceller will be described below.
[0008]
Now, assuming that the desired signal s, the noise n0, the noise n1, and the output signal y are
statistically stationary and the average value is zero, the residual output e becomes e = s + n0−y.
The expected value of this squared is E [e2] = E [s2] + E [(n0-y) 2] + 2E [s (n0-y) because the
03-05-2019
2
desired signal s has no correlation with the noise n0 and the output y. ]] = E [s2] + E [(n0-y) 2]
[0009]
Assuming that the adaptive filter circuit 5 converges, the adaptive filter circuit 5 updates the
filter coefficients such that E [e 2] is minimized. At this time, E [s2] is not affected, so Emin [e2] =
E [s2] + Emin [(n0-y) 2].
[0010]
That is, E [(n0-y) 2] is minimized by minimizing E [e2], and the output y of the adaptive filter
circuit 5 becomes an estimate of the noise signal n0. The expected value of the output from the
combining circuit 4 is only the desired signal s. That is, adjusting the adaptive filter circuit 5 to
minimize the total output power is equal to the subtraction output e being the least squares
estimated value of the desired speech signal s.
[0011]
Although the output e generally has some noise remaining in the signal s, since the output noise
is given by (n 0 −y), minimizing E [(n 0 −y) 2] is an output It is equivalent to maximizing the
signal to noise ratio.
[0012]
In some cases, the synthesis circuit 4 may be acoustic synthesis means.
That is, the adaptive filter circuit 5 forms a noise cancellation voice signal -y having the same
phase as noise and the opposite phase of noise and supplies this to a speaker or the like to
acoustically add to the main voice to reduce noise. Do. The residual e in this case is picked up by
the residual detection microphone.
[0013]
03-05-2019
3
The adaptive filter circuit 5 can be realized either by an analog signal processing circuit or a
digital signal processing circuit, but in general, a digital processing circuit using a DSP (digital
signal processor) It is made up of An example of the configuration of the adaptive filter circuit 5
in the case of the configuration of the digital processing circuit is shown in FIG.
[0014]
In this example, the adaptive filter circuit 5 comprises an FIR filter type adaptive linear combiner
100 and a filter coefficient update computing means 110. The adaptive filter circuit 5 can be
configured as software by a DSP on which a microcomputer is mounted. In this example, the
algorithm for updating the filter coefficient will be described as using LMS (Least Mean Squares),
which is frequently used because the amount of calculation is small and practical.
[0015]
The LMS method will be described with reference to FIG. As shown in FIG. 7, the adaptive linear
combiner 100 includes a plurality of delay circuits DL1, DL2,..., DLm (m is a positive integer)
each having a delay time Z-1 of unit sampling time, and input noise n1 and weighting circuits
MX0, MX1, MX2,... MXm for multiplying the output signals of delay circuits DL1, DL2,... DLm by
weighting factors (filter coefficients), and the outputs of weighting circuits MX0 to MXm The
adder circuit 101 is provided. The output of the adder circuit 101 is the signal y described in FIG.
[0016]
The weighting coefficients to be supplied to the weighting circuits MX0 to MXm are formed by
the filter coefficient arithmetic circuit 110 based on the residual signal e from the synthesizing
circuit 4 and the reference input n1 by the LMS algorithm. The algorithm executed by the filter
coefficient calculation circuit 110 is as follows.
[0017]
Now, let the input vector Xk at time k be Xk = [x0k x1k x2k... Xmk] T, as also shown in FIG. 7,
output yk, weight coefficient wjk (j = 0,1,2,. Assuming that m), the input / output relationship is
03-05-2019
4
as shown in the following equation 1:
[0019]
Then, if the weight vector Wk at time k is defined as Wk = [w0k w1k w2k... Wmk] T, the input /
output relation is given by yk = Xk T · Wk (1).
Here, assuming that the desired response is dk, the residual ek is ek = dk −yk = dk −Xk T · Wk
(2)
[0020]
In the LMS method, updating of the weight vector is sequentially performed according to the
following equation (3): Wk + 1 = Wk + 2 μ ・ ek ・ Xk (3) The initial value of the weighting
factor is set to a constant value or a random value. Here, μ is a gain factor (step gain) that
determines the speed and stability of adaptation.
[0021]
In the above equation (3), the vector for correcting the coefficient vector Wk at a certain time k is
the second term of the right side of equation (3), but the gain factor μ and the instantaneous
error ek are both scalar values, Directly influence the value. Since the reference input vector Xk
also works in the form of a product, this also influences the correction value. The average
convergence time constant τa is represented by τa = (n + 1) / 4μ · trE [Xi Xj T]. Here, n is the
order of the reference input vector (corresponding to the number of taps of the FIR filter), and
trE [Xi Xj T] is the average power of the reference input. That is, the larger the number of taps of
the FIR filter, the slower the convergence speed, and the larger the gain factor μ, the faster the
convergence speed. When the noise is a stationary signal, if the convergence speed is fast, the
final residual noise level is large, and if the convergence is slow, the final noise level is small.
[0022]
03-05-2019
5
SUMMARY OF THE INVENTION In the above-mentioned adaptive noise canceller, in order to deal
with noise in the full range of audio frequencies, it takes some time for the FIR filter of the
adaptive filter circuit to converge. The For this reason, when noise is a stationary signal, noise
reduction can be performed effectively, but the target noise signal is a nonstationary signal, for
example, a frequency characteristic changes like momentarily like voice. In the case of such a
signal, the nature of the signal changes before convergence, so that the processing can not be
applied, resulting in a disadvantage that the unnecessary signal can not be reduced.
[0023]
As mentioned above, if the step gain μ is increased, the convergence speed of the adaptation can
be increased, but it is generally difficult to completely decorrelate the desired speech in the main
input and the reference input. Therefore, if the step gain μ is increased, the adaptive processing
affects not only the noise but also the desired signal, resulting in a disadvantage that the desired
signal as an output signal is distorted.
[0024]
In view of the above, it is an object of the present invention to provide an adaptive signal
processing device capable of increasing the amount of reduction for non-stationary signals such
as voice without causing distortion in a desired signal. .
[0025]
SUMMARY OF THE INVENTION In order to solve the above problems, the adaptive signal
processing device according to the present invention corresponds to the main input signal
including the desired voice and the noise signal when the reference numerals of the
embodiments described later correspond. The noise signal correlated with the noise signal in the
main input signal is supplied to the first adaptive filter means 23 as a reference input signal and
supplied to the first combining means 13 to approximate the noise signal in the main input
signal. A device in which the output signal of the first adaptive filter means 23 is subtracted from
the main input signal in the first combining means 13 to reduce and eliminate noise signals in
the main input signal. And band limiting means 31 and 33 for extracting only components of
specific frequency bands of the main input signal and the reference input signal, and a specific
frequency of the main input signal from the band limiting means 31. A second combining means
32 for supplying a component of the band, a second adaptive filter means 34 for receiving a
component of a specific frequency band of the reference input signal from the band limiting
means 33, and a second combining means Control means 35 for adaptively updating the
weighting factors of the second filter means 34 so as to minimize the output power of 32. The
weighting factors determined by the control means 35 It is characterized in that it is also used as
a weighting factor.
03-05-2019
6
[0026]
According to the present invention of the above configuration, the second adaptive filter means is
configured to form a weighting factor (filter factor) so as to form a signal approximating noise in
the main input signal from the band-limited reference input signal. Is updated.
The weighting factor determined by the adaptive processing in the second adaptive filter means
is copied as the weighting factor of the first adaptive filter means.
In the first combining means, the output of the first adaptive filter means is subtracted from the
main input signal.
An output signal of the combining means is derived as a system output.
[0027]
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT An embodiment of an adaptive
noise reduction system according to the invention will now be described with reference to FIGS.
1 and 2. FIG.
[0028]
In FIG. 1, reference numeral 11 denotes a main input microphone for picking up desired voice,
and 21 denotes a reference input microphone for picking up unnecessary voice and ambient
noise in a direction to be removed as noise.
In this example, the arrival direction of the desired voice is mainly from the upper side to the
lower side in the figure, as indicated by an arrow AR in FIG. 2, and is the front direction in the
camera integrated VTR. And, this example is an example of realizing a device for preventing
voices from the direction opposite to this direction (rear direction) from being picked up as noise.
03-05-2019
7
[0029]
In the case of this example, the main input microphone 11 is configured by a nondirectional
microphone as shown in FIG. On the other hand, as shown in FIG. 2, the reference input
microphone 21 is composed of a unidirectional microphone having a low sensitivity in the
desired voice incoming direction and a high sensitivity in the back direction. Moreover, in the
case of this example, these microphones 11 and 21 are closely arranged along the direction of
the arrow AR.
[0030]
Then, an audio signal picked up by the main input microphone 11 and converted into an electric
signal is obtained and supplied to the A / D converter 12 for the entire audio frequency band to
be converted into a digital signal. Further, an audio signal collected by the reference input
microphone 21 and converted into an electric signal is obtained and supplied to the A / D
converter 22 for the entire audio frequency band to be converted into a digital signal. In this
case, the sampling frequency (sampling frequency) at the A / D converters 12 and 22 is set to 48
kHz so as to cover the entire audio frequency band.
[0031]
Then, the output digital signal of the A / D converter 12 is supplied to the subtraction circuit 13.
Further, the output digital signal of the A / D converter 22 is supplied to an adaptive linear
combiner (FIR filter) 23 shown in FIG. The weighting factor (filter factor) of the adaptive linear
combiner 23 is supplied from the filter factor computing circuit 35, and the value of the filter
factor is determined as described later.
[0032]
The output signal of the adaptive linear coupler 23 is a signal approximated to an unnecessary
signal as a noise signal in the output signal of the A / D converter 12, and this signal is supplied
to the subtraction circuit 13, and the output signal of the A / D converter 12 Is subtracted from
The output signal of the subtraction circuit 13 is supplied to a D / A converter 14 with a
03-05-2019
8
sampling frequency of 48 kHz, converted back to an analog signal, and derived at the output
terminal 15 as a system output.
[0033]
The filter coefficients of the adaptive linear combiner 23 are determined in the filter coefficient
calculation circuit 35 as follows. That is, the digital signal of the main input signal from the A / D
converter 12 is supplied to the digital low pass filter 31 having a cutoff frequency of 250 Hz, for
example, band-limited, and supplied to the subtraction circuit 32. The digital signal of the
reference input signal from the A / D converter 22 is also supplied to the digital low pass filter
33 having a cutoff frequency of 250 Hz and band-limited to the adaptive linear combiner 34 and
the filter coefficient calculation circuit 35. Supplied.
[0034]
The low pass filters 31 and 33 can be realized by appropriately thinning samples from the output
digital signals of the A / D converters 12 and 22 in accordance with the cutoff frequency. The
sampling frequency (sampling frequency) of the output signals of the low pass filters 31 and 33
is a frequency capable of reproducing a band up to 250 Hz, that is, twice or more of a cutoff
frequency (highest frequency of limited band). The frequency is, for example, 600 Hz.
[0035]
The adaptive linear combiner 34 forms from the output signal of the low pass filter 33 a signal
that approximates the unwanted signal in the output of the low pass filter 31. Then, the
subtraction circuit 32 subtracts the output signal of the adaptive linear coupler 34 from the
signal from the low pass filter 31, and the subtraction output is supplied to the filter coefficient
calculation circuit 35.
[0036]
Then, as described above, in the filter coefficient calculation circuit 35, the output power of the
residual signal is minimized from the output signal of the low pass filter 33 and the subtraction
03-05-2019
9
output (residual signal) from the subtraction circuit 32. The updated values of the filter
coefficients to be supplied to the adaptive linear combiner 34 are determined. The determined
filter coefficient is also supplied to the adaptive linear combiner 23 as the filter coefficient.
[0037]
In this case, since the filter coefficient obtained by the filter coefficient calculation circuit 35 is
supplied to the adaptive linear combiner 34, it is possible to use a digital signal with a low
sampling frequency of 600 Hz and a small number of samples per unit time. It corresponds.
Therefore, as it is, the sampling frequency is 48 kHz and does not conform to the handling signal
in the adaptive linear combiner 23 with a large number of samples per unit time. For this reason,
the filter coefficient supplied to the adaptive linear combiner 23 is subjected to interpolation
processing (for example, pre-value hold) on the coefficient value obtained as described above,
and corresponds to the sampling frequency of 48 kHz. It is being done.
[0038]
In FIG. 1, the portions of the adaptive linear couplers 23 and 34 and the filter coefficient
calculation circuit 35 which are shown surrounded by a solid line can be constituted by a DSP
(digital signal processor) including a microcomputer.
[0039]
With such a configuration, the component of the desired sound is hardly included in the collected
voice from the directivity characteristic of the microphone 21. Therefore, the reference input
signal from the microphone 21 is the output of the microphone 11 (main input) The desired
sound in the signal is uncorrelated to the unnecessary signal (noise component) in the main
input.
[0040]
Therefore, the signal obtained from the adaptive linear coupler 23 is a signal that approximates
the unnecessary signal in the main input, and this is subtracted from the main input in the
subtraction circuit 13, so only the desired signal component from the subtraction circuit 13 is Is
obtained.
03-05-2019
10
That is, the unnecessary signal component contained in the main input is adaptively canceled,
and only the desired sound signal is obtained at the output terminal 15.
[0041]
As described above, by using the adaptive noise canceller, it is possible to realize a superdirective
microphone device having a sharp directivity in the direction of the sensitivity minimum of the
unidirectional microphone 21 for reference input.
[0042]
And in the above example, the signal handled by the adaptive noise canceller including the
adaptive linear combiner 34 is a more stationary signal compared to the signal of the entire
audio frequency band by band-limiting the main input and the reference input. It can be handled.
Moreover, in general, in the voice signal, power is concentrated on the low frequency side.
Therefore, if the main input and the reference input are band-limited by the low pass filters 31
and 33 and the adaptive processing is performed on the component having the large power, as
described above, the adaptive processing is performed on the signal having the large power.
Since the system works effectively and quickly, the system of the adaptive noise canceller
including the adaptive linear combiner 34 has a high speed of convergence of the adaptation and
a large amount of noise reduction.
[0043]
Since the filter coefficient of this system is also used as it is as the filter coefficient of the
adaptive type linear combiner 23, the noise reduction amount of the system output obtained at
the output terminal 15 also becomes large.
[0044]
In addition, since the adaptive linear combiner 34 and the filter coefficient calculation circuit 35
may correspond to a limited band, the number of samples to be handled is reduced, which can be
03-05-2019
11
realized by a low speed DSP, and a general-purpose microprocessor It becomes possible to use.
Since only the filter coefficients are copied to the adaptive linear combiner 23, the adaptive linear
combiner 23 can be similarly configured, which simplifies the overall configuration of the system
and reduces the hardware size. It can be smaller and cost can be reduced.
[0045]
As described above, according to the present invention, the adaptive operation on non-stationary
signals such as voice becomes faster than the conventional adaptive noise reduction system, and
as a result, the amount of noise reduction becomes large.
[0046]
FIG. 3 is a diagram showing the results of experiments to confirm the effects of the present
invention, in which a female voice comes from the front direction as a desired signal and a male
voice from a side as an unnecessary voice to be removed as noise. Fig. 6 shows a comparison of
the system output of the conventional system and the system of the invention described above.
[0047]
FIG. 3A is a waveform of male voice only from the side, FIG. 3B is an output waveform when the
reduction removal of male voice from this side is performed by the conventional adaptive noise
reduction system, and FIG. 3C is this In each of the output waveforms when the reduction and
removal of the male voice from the side are performed in the device of the invention, the rising
portions of the voice are cut out.
From this FIG. 3, according to the device of the present invention, it is possible to observe how
adaptation is reduced and unnecessary sound is reduced more quickly than in the conventional
system.
Incidentally, the reduction by the conventional system of male voice in FIG. 3A is −13.0 dB, and
the reduction by the device of the present invention is −16.3 dB.
[0048]
03-05-2019
12
In the example of FIG. 1, although the low pass filter for band limiting was comprised with the
digital processing circuit (for example, thinning-out processing circuit), a band limiting circuit
can also be made into an analog circuit. FIG. 4 is a block diagram of one embodiment in that case.
[0049]
That is, in this example, the output signal of the main input microphone 11 is supplied to the A /
D converter 42 through the low pass filter 41 whose cutoff frequency is 250 Hz. Then, in this A /
D converter 42, sampling is performed at a sampling frequency of 600 Hz, and the sampled value
is converted into a digital signal and supplied to the subtraction circuit 32.
[0050]
The output signal of the reference input microphone 21 is also supplied to the A / D converter
44 via the low-pass filter 43 of 250 Hz, sampled at a sampling frequency of 600 Hz, and the
sampled value is converted to a digital signal. Then, the output digital signal of the A / D
converter 44 is supplied to the adaptive linear combiner 34 and the filter coefficient calculation
circuit 35. The other configuration is completely the same as the example of FIG.
[0051]
The D / A converter 14 may be omitted, and the output signal of the subtraction circuit 13 may
be derived to the output terminal 15 as it is as a digital signal.
[0052]
FIG. 5 is another embodiment of the present invention.
In this example, the adaptive processing is performed with band limitation and the filter
coefficient of the adaptive linear combiner 34 is determined by the filter coefficient computing
circuit 35 in the same manner as in the example of FIG. 1 or FIG. Although corresponding to the
03-05-2019
13
example of FIG. 4), signal processing of the entire band is performed as an analog signal.
[0053]
That is, the main input signal from the microphone 11 is supplied to the low pass filter 41 for
band limitation and is also supplied to the subtraction circuit 16 without being converted into a
digital signal. Further, the reference input signal from the microphone 21 is supplied to the low
pass filter 43 for band limitation and also supplied to the gain adjustment circuit 24 of an analog
configuration without being converted into a digital signal. Then, the output signal of the gain
adjustment circuit 24 is supplied to the subtraction circuit 16, and the output terminal 17 is
derived from the subtraction circuit 16 so that a system output can be obtained at the output
terminal 17.
[0054]
In the case of this example, the value of the weighting factor of each tap of the adaptive linear
combiner 34 for the low-pass component of the main and reference input signals, and the value
of the gain of the gain adjustment circuit 24 are pre-specified The filter coefficient calculation
circuit 35 controls the gain value of the gain adjustment circuit 24 according to the obtained
weighting coefficient. In the filter coefficient calculation circuit 35, for example, a
correspondence table of the weight coefficient values of the respective taps of the adaptive linear
combiner 34 and the gain values of the gain adjustment circuit 24 is prepared.
[0055]
In this example, the same effect as described above is obtained, and the signals of all the main
and reference bands are analog signal processing, and a gain adjustment circuit is used instead of
the digital filter to further enhance the hardware. Scale reduction and cost reduction will be
possible.
[0056]
Although the low pass filter is used to band-limit the main input signal and the reference input
signal in the above example, it is also possible to use a band pass filter to exclude 100 Hz or less
in the above example, for example. it can.
03-05-2019
14
The point is that, for example, components of a specific band where the power mainly exists in
the target signal of adaptive signal processing is extracted and subjected to adaptive processing,
and the weighting factor obtained at that time is used as the adaptive filter for the entire band.
The weighting factor may be copied, or the gain of the gain adjustment circuit may be controlled
in accordance with the weighting factor.
[0057]
The above example is the case where the present invention is applied to a microphone device,
but the present invention is not limited to the above example, but is applicable to all adaptive
signal processing that handles non-stationary signals. It's too late.
[0058]
Further, the algorithm of updating of the weighting factor in the adaptive processing is not
limited to the LMS method, and it is a matter of course that, for example, a learning identification
method or another algorithm can be used.
[0059]
As described above, according to the present invention, adaptive signal processing is performed
by band-limiting the main input signal and the reference input signal, and the weighting factor
obtained at that time is used for the adaptation for the entire band. Since the weighting factor of
the filter is copied or the gain of the gain adjusting circuit is controlled according to the
weighting factor, the adaptation operation can be quickened even for non-stationary signals such
as voice, resulting in noise reduction Will also grow.
[0060]
In addition, since only the low frequency components of the main input and the reference input
can be realized by adaptive arithmetic processing, the sampling frequency at A / D conversion
can be lowered, thereby reducing the size and cost of hardware. be able to.
In addition, since low-speed DSPs can perform adaptive arithmetic processing, low-cost general
purpose DSPs can be used.
03-05-2019
15
[0061]
Furthermore, as in the example of FIG. 5, the use of a gain adjustment circuit, which is an analog
circuit, for processing of signals in the entire band can further reduce the size and cost of
hardware.
03-05-2019
16
Документ
Категория
Без категории
Просмотров
0
Размер файла
27 Кб
Теги
jph06292292
1/--страниц
Пожаловаться на содержимое документа