close

Вход

Забыли?

вход по аккаунту

JP2002367298

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 JP2002367298
[0001]
TECHNICAL FIELD The present invention relates to a noise canceller device and a noise
cancellation method.
[0002]
2. Description of the Related Art For example, in an electronic device having a playback device
and a recording / reproducing device internally provided with a drive component such as a
camera integrated type VTR, the rotational noise of a rotating drum and drum motor, magnetic
tape and rotating magnetic magnet Mechanical noise (mechanical noise) generated from driving
parts such as beating sound between heads collects audio of an audio signal to be recorded on a
recording medium together with or independently of a video signal in a reproducing apparatus or
a recording / reproducing apparatus There is a problem that noise is mixed into the built-in
microphone as noise sound, and the noise signal is recorded on the recording medium together
with the original audio signal.
[0003]
In particular, with the recent miniaturization of electronic devices, it has become increasingly
difficult to isolate noise sources from microphones.
[0004]
08-05-2019
1
Next, with reference to FIG. 1, a configuration of a general active noise canceller (ANC) for
removing noise mixed in an audio signal will be described.
The main input microphone 1 picks up the required voice (indicated by a straight solid line
arrow), but at the same time also picks up noise (indicated by a wavy solid line arrow).
Therefore, the main voice input signal S mixed with the noise signal is obtained from the
microphone 1, and the main voice input signal S is input to the delay unit 3.
[0005]
On the other hand, the reference input signal X highly correlated with the above-mentioned noise
signal is input to the adaptive signal processing circuit 5 according to the LMS method. When an
audio signal based on an audio as indicated by a broken-line broken arrow is mixed into the
reference input signal X, the original audio signal is also canceled in the adder 6 described later.
In the inventions described in JP-A-176113, JP-A-11-232802 and JP-A-11-232803, a drum
rotation reference signal serving as a base frequency for noise generation is obtained as a
reference input signal X from a microcomputer or the like. I use it.
[0006]
The main input signal S is input to the positive terminal of the adder 6 through the delay unit 3.
On the other hand, the reference input signal X is subjected to adaptive processing by the
adaptive signal processing circuit 5, and a pseudo noise signal similar to the noise signal
contained in the main input signal S is output as the adaptive output signal Y, and the adaptive
output signal Y is an adder. The signal is input to the negative terminal of 6 and subtracted from
the delayed main input signal S.
[0007]
Therefore, from the adder 6, an audio signal in which the noise signal is canceled is obtained, and
this audio signal is output from the output terminal 7. The voice signal from which the noise
signal output from the adder 6 is canceled is fed back to the adaptive signal processing circuit 5
08-05-2019
2
as a residual signal E and used for adaptive processing.
[0008]
The delay amount of the delay device 3 described above, including the correction of the delay
required for the processing time by the adaptive signal processing circuit 5, is set to the delay
time at which the noise cancellation effect is highest.
[0009]
Next, a specific circuit of the adaptive signal processing circuit 5 in the active noise canceller of
FIG. 1 will be described with reference to FIG.
Although various methods have been proposed as this adaptive signal processing algorithm,
generally, a Least Mean Square (LMS) method having a relatively high convergence speed and a
small arithmetic circuit size is often used. The case of the LMS method will be described. The
LMS method can be processed entirely by DSP (digital signal processor), hardware by digital LSI,
or software by microcomputer.
[0010]
The reference input signal X corresponding to the reference input signal X in FIG. 1 is input to
the adaptive filter 10 and the LMS arithmetic processing circuit 14 surrounded by a broken line.
The adaptive filter 10 is generally composed of an FIR (finite impulse response) digital filter with
several hundred taps and the adaptive filter coefficient W at each tap is adaptively updated
according to the LMS algorithm. Go.
[0011]
The adaptive filter 10 in this case is a FIR filter of (m + 1) taps. 111, 112, ..., 11m are delay
devices for unit sampling time. X0, X1,..., Xm indicate the signals subjected to the respective
delays. 120, 121, ..., 12m are multipliers for coefficient multiplication, and W0, W1, ..., Wm are
coefficients of the respective multipliers. The outputs from the multipliers 120, 121,..., 12m are
added by the adder 13 and the added output is output as the adaptive output signal Y.
08-05-2019
3
Accordingly, the adaptive output signal Y is expressed by the following equation (1).
[0013]
In the equation (1), j represents 0, 1,..., M.
[0014]
Further, in the LMS arithmetic processing circuit 14, each of the adaptive filter coefficients W0,
W1,..., Wm is sampled from the reference input signal X and the residual signal E described
above according to Equation 2 shown below. I will update it.
[0015]
[Equation 2] Wk + 1 = Wk + 2 μ · Ek · Xk
[0016]
In the equation (2), each small letter k suffix indicates the passage of time, and for example, let k
be a unit sampling number, and let (k + 1) th sampling Wk + 1 be a current adaptive filter
coefficient. , Wk denote the adaptive filter coefficients at the k-th sampling point, that is, one
sampling past.
Also, μ is called step gain or step size and is a parameter that determines the convergence speed
in the LMS algorithm. The larger the value, the faster the convergence, but the lower the
accuracy after convergence, conversely, the smaller the value, the convergence Is slower and
more accurate after convergence, so it is optimized and set according to the adaptive system
conditions used.
Also, the residual signal E corresponds to the residual signal E of FIG.
[0017]
As described above, the LMS arithmetic processing circuit 14 uses the equation 2 to minimize the
adaptive filter coefficient W in the adaptive filter 10 with respect to the reference input signal X
included in the residual signal E at all times. By making the reference input signal X in FIG. 1 a
noise signal for updating, noise components included in the main input signal S can be
08-05-2019
4
minimized.
[0018]
Next, with reference to FIG. 3, a conventional adaptive control noise canceller (refer to JP-A-11232802) will be described.
FIG. 3 shows the case of the adaptive control noise canceller in the VTR integrated camera.
[0019]
In FIG. 3, reference numerals 20 and 21 denote an Lch (left channel) microphone and an Rch
(right channel) microphone which constitute a built-in stereo microphone, respectively.
The type and number of microphones are arbitrary. The microphones 20 and 21 are disposed
close to a helical scan type magnetic recording and reproducing apparatus of a VTR integrated
camera.
[0020]
In the magnetic recording and reproducing apparatus, reference numeral 41 denotes a rotary
drum, and a pair of rotary magnetic heads 41 and 44 are provided on the main surface thereof,
for example, at an angle of 180 °. Then, the magnetic tape 43 is guided so as to be obliquely
wound around the rotating drum 41 and a fixed drum (not shown). Then, noise noise and
vibration from the magnetic recording and reproducing apparatus are mixed with normal sound
as so-called mechanical noise and collected by the microphones 20 and 21. The mechanical noise
is caused by the tapping noise when the magnetic tape 43 and the rotary magnetic head 44
contact, the electromagnetic noise of the drum motor 42 driving the rotary drum 41 and the like,
and the frequency of the mechanical noise generated is the drum motor 42. Harmonics of the
rotational frequency of
[0021]
08-05-2019
5
In order to control a predetermined frequency and phase of rotation of the drum motor 42, a
servo signal 45 from a microcomputer (microcomputer) 46 is supplied to the drum motor 42,
whereby the drum motor 42 is controlled so as to always keep constant rotation. Be done.
Incidentally, the recording / reproducing apparatus is not limited to the VTR apparatus as shown
in the figure, but is applicable to a disk drive apparatus such as a hard disk drive apparatus or an
optical disk drive apparatus having a drive apparatus controlled by a servo signal. It is possible.
[0022]
In FIG. 3, the mixed signal of the voice signal and the mechanical noise signal from the
microphones 20 and 21 is pre-amplified by the amplifiers (AMPs) 22 and 23 and then supplied
to the A / D converters 24 and 25 for analog-to-analog conversion. The digital signal obtained by
digital conversion is input to the mechanical noise reduction processing circuit 40.
[0023]
The mechanical noise reduction processing circuit 40 performs independent processing with two
stereo channels, and is configured to be able to perform optimal mechanical noise reduction
processing even if the noise components differ between Lch and Rch, for example.
If the noise components of the monaural signal, the Lch signal and the Rch signal are almost the
same, the mechanical noise reduction processing circuit 40 can be configured as a circuit for one
channel.
[0024]
In the mechanical noise reduction processing circuit 40, digital signals from the A / D converters
24 and 25 are respectively input to the positive terminals of the adders 26 and 27 and are input
to the negative terminals of the adders 24 and 25 as described later. The pseudo noise signal
generated by the adaptive signal processing is subtracted, and output as Lch and Rch output
signals subjected to noise cancellation from the output terminals 36 and 37, respectively.
Furthermore, the Lch and Rch output signals are input to the limiters 34 and 35, respectively,
and Lch and Rch are input to the adaptive processing circuit to input only noise signal
components that are normally at a low level with respect to the dynamic range of voice. When
08-05-2019
6
the level of the output signal is a large level, processing for applying a limiter is performed and
input to the step gain control circuit (SG) 32, 33.
[0025]
The step gain control circuits (SG) 32 and 33 perform processing of multiplying the outputs of
the limiters 34 and 35 by the step gain μ in the above equation (2). The step gain control signal
48 from the microcomputer 46 is supplied to the step gain control circuits (SG) 32 and 33 so
that the step gain μ can be varied by the step gain control signal 48. Here, the common step
gain control signal 48 is supplied to the step gain control circuits (SG) 32 and 33 to control the
step gain, but the step gain is controlled independently for each of Lch and Rch. You may
[0026]
Further, the signals from the step gain control circuits (SG) 32 and 33 are input to the adaptive
signal processing circuits 28 and 29 as the residual signal E in FIG. Here, the adaptive signal
processing circuits 28 and 29 are configured as in FIG. 2 and processed by the LMS algorithm.
The drum reference signal 47 similar to the reference input signal X in FIG. 1 is input from the
microcomputer 46 to the adaptive signal processing circuits 28 and 29. The drum reference
signal 47 is a servo signal of the drum motor 42. Because of the 45 base frequency signals, in the
adaptive signal processing circuits 28 and 29, all signal components correlated with this
reference signal, that is, the drum rotation frequency and its harmonic components are targets
for adaptive processing. Further, output signals corresponding to the adaptive output signal Y in
FIG. 1 from the adaptive signal processing circuits 28 and 29 are input to the negative terminals
of the adders 26 and 27 via the switches (SW) 30 and 31, respectively. Here, the switches (SW)
30, 31 control whether or not subtraction is performed in the adders 26, 27, by the cancel on /
off control signal 49 from the microcomputer 46, and can be canceled by turning on only when
necessary. It is like that.
[0027]
In the conventional adaptive control noise canceller (mechanical noise reduction circuit)
configured as shown in FIG. 3, the mode transition is appropriately performed by controlling the
step gain control signal 48 and the cancel on / off control signal 49 from the microcomputer
according to the mode transition of the VTR. I was doing mechanical noise cancellation at the
time.
08-05-2019
7
[0028]
Next, noise generated at the time of mode transition of the VTR and conventional parameter
control will be described with reference to FIG.
First, FIG. 4A conceptually shows the noise generation amount when the VTR mode changes from
recording to recording pause (temporary stop) to recording. First, at the time of recording, the
rotating drum 41 (FIG. 3) is rotated at a predetermined rotation speed, and the rotating magnetic
head 44 and the magnetic tape 43 attached to the rotating drum 41 contact at high speed. Both
the sound and the tapping sound between the magnetic tape 43 and the rotary magnetic head 44
are generated.
[0029]
Next, after transition to the pause mode after the mode transition period, the tape tension
pressure weakens to protect the magnetic tape, and the noise generated is only the drum rotation
sound and the noise amount is reduced. Further, when transitioning to the recording mode after
passing through the mode transition period, the noise generated due to the magnetic tape
coming into contact with the head again becomes the drum rotation sound and the beating sound
between the rotary magnetic heads 44 of the magnetic tape 43, The amount of noise increases.
[0030]
Next, examples of parameter control and noise amount when this noise is processed by the
conventional adaptive control noise canceller shown in FIG. 3 are shown in FIGS. 4B to 4D, which
will be described below. In the example of FIG. 4, since mechanical noise always occurs, the
switches (SW) 30, 31 in FIG. 3 are always in the cancel on state by the cancel on / off control
signal 49.
[0031]
Here, the influence of the step gain μ value on the mechanical noise cancellation operation will
08-05-2019
8
be described. When the step gain μ value is large, the coefficient update amount of the adaptive
filter increases, and for example, the followability when the noise amount changes is improved,
but conversely, the stability after the coefficients converge is deteriorated. Also, when the step
gain μ value is small, the coefficient update amount of the adaptive filter decreases, and the
followability when the noise amount changes is deteriorated, but conversely, the stability after
the coefficient converges is good.
[0032]
Therefore, generally, as shown in FIG. 4B, the step gain μ (shown by a broken line) is fixed to a
predetermined value, and in this case, the step gain μ is determined to perform stable noise
cancellation in each VTR mode. Therefore, if the amount of noise or the band of noise changes
due to mode transition, a problem occurs in which the noise increases until it follows. Therefore,
in Japanese Patent Application Laid-Open No. 11-232802, as shown in FIG. 4C, the value of the
step gain μ is increased only at the time of mode transition to accelerate the follow-up and
suppress the noise amount.
[0033]
Furthermore, in FIG. 4D, the step gain μ is fixed to a predetermined value only at the time of
recording, and is zero in the other modes, so that the second term of the right side in the above
equation 2 becomes zero, and the first term The coefficient of the adaptive filter is held so that
coefficient updating is not performed. In this way, noise cancellation is always performed for
recording, that is, for an audio signal recorded on a magnetic tape.
[0034]
Next, with reference to FIG. 5, the VTR mode transition and the conventional cancel on / off
control will be described. FIG. 5A conceptually shows the noise generation amount when the VTR
mode changes from recording to recording standby (stop) to recording. First, at the time of
recording, the rotating drum 41 is rotated at a predetermined rotational speed, and the rotating
magnetic head 44 attached to the rotating drum 41 and the magnetic tape 43 contact at high
speed. Both the hitting sound between the tape 43 and the rotary magnetic head 44 is generated.
Next, after transition to the recording standby mode after the mode transition period, the rotation
of the drum is stopped and noise generation is eliminated. Furthermore, when transitioning to
08-05-2019
9
the recording mode after the mode transition period, the magnetic tape 43 comes into contact
with the rotating magnetic head 44 again, so the noise generated is the drum rotating sound and
the tapping sound between the magnetic tape 43 and the rotating drum 41. And the amount of
noise increases.
[0035]
Next, examples of parameter control and noise amount when this noise is processed by the
conventional adaptive control noise canceller shown in FIG. 3 are shown in FIGS. 5B to 5D, which
will be described below. First, FIG. 5B shows the case where the step gain μ (indicated by a
broken line) is fixed to a predetermined value, and the cancel on / off control (indicated by a
solid line) is always in the cancel on state. At this time, first, when the recording mode is changed
to the recording standby mode, both the drum rotational noise that has been canceled and the
tapping noise between the magnetic tape 43 and the rotary drum 41 disappear rapidly. , The
pseudo noise component of the same level remains, and it tracks to the noise zero state in the
recording standby mode. Also, when transitioning from the recording standby mode to the
recording mode, the drum rotational sound and the tapping sound between the magnetic tape 43
and the rotary drum 41 are generated suddenly from the noise zero state during the mode
transition, so The amount of noise increases, and noise is generated for a while even if the mode
is shifted to the recording mode.
[0036]
Therefore, according to the invention of Japanese Patent Application Laid-Open No. 11-232802,
as shown in FIG. 5C, the follow-on property at the time of recording standby is accelerated by
setting the cancellation on / off control to the cancellation off state at the time of recording
standby It is like that. However, in this example, even if the switches (SWs) 30 and 31 in FIG. 3
are turned off in the cancel-off state, the adaptive signal processing circuits 28 and 29 continue
updating the adaptive filter coefficients, so that adaptive operation can be performed. It remains
unstable and there is a possibility that the adaptive filter coefficients may malfunction.
Furthermore, in FIG. 5D, by setting the step gain μ to zero except when recording, and holding
the adaptive filter coefficient at the time of recording, noise cancellation is always performed at
the time of recording as in FIG. To be done.
[0037]
The present applicant has previously relied on drum rotation for adaptive signal processing by
08-05-2019
10
the least mean square (LMS) method in order to cancel mechanical noise in a camera integrated
VTR in Japanese Patent Laid-Open No. 11-176113. We have proposed a mechanical noise
canceller that inputs the base frequency signal as a reference signal and efficiently reduces only
mechanical noise without affecting the audio signal.
[0038]
This mechanical noise canceller cancels only mechanical noise very efficiently even if the audio
level changes in the normal input signal level range where the audio level changes without
changing the mechanical noise level, but, for example, recording AGC (Automatic Gain Control)
(automatic gain control) circuit operates to compress the dynamic range (hereinafter referred to
as D range) of the audio signal to be recorded in accordance with the D range recordable on the
recording medium, or When the MGC (Manual Gain Control) (manual gain control) circuit is
operated to allow the user to arbitrarily change the audio level, the mechanical noise level is also
changed at the same time as the audio level. Step gain (also called step size) which is an internal
parameter that determines the pull-in speed Only in the setting of that), it can not follow the
change in the level of this mechanical noise, if the mechanical noise is not canceled has occurred.
[0039]
Further, in Japanese Patent Application Laid-Open No. 11-232803 proposed by the applicant of
the present invention, when the noise level is sharply reduced due to the AGC operation etc., the
internal feedback loop of LMS adaptive signal processing is cut off and the noise cancellation off
mode In this case, disturbance of the adaptive filter coefficient occurs in the noise cancellation
off mode, and the followability is inferior.
[0040]
SUMMARY OF THE INVENTION The present invention has been made in view of these problems,
and the conventional optimization of the step gain only leads the past data to the rapid noise
level change. In order to improve the poor followability, the present invention introduces a
forgetting factor that determines the weight to past data, and adaptively controls the noise to
improve the followability to noise changes. It is intended to propose a canceller device and a
noise cancellation method.
[0041]
Also, conventionally, when the noise level decreases sharply due to AGC operation etc., the
internal feedback loop of LMS adaptive signal processing is cut off to realize the noise
cancellation off mode, so the disturbance of the adaptive filter coefficient in the noise
08-05-2019
11
cancellation off mode In the present invention, the noise cancel off mode is realized by
controlling the step gain and the forgetting factor in order to improve the poor followability, and
the disturbance of the adaptive filter factor does not occur. It is an object of the present invention
to propose a noise canceller device and a noise cancellation method capable of improving the
followability to noise change.
[0042]
According to a first aspect of the present invention, an information signal is mixed in an
information signal based on an energy wave from an energy wave generating means for
generating an energy wave driven by a drive signal from a drive signal source. In a noise
canceller device in which noise components are canceled, pseudo noise having a correlation with
noise components mixed in information signals based on level control means for controlling the
level of information signals and energy waves from energy wave generation means An adaptive
filter for generating a signal, a subtraction means for obtaining an information signal in which a
noise component is canceled by subtracting a pseudo noise signal from an information signal
output from the level control means, and an information signal in which a noise component is
canceled Limiter means for supplying the adaptive filter after controlling the threshold level
below the Is controlled by that control unit, a noise canceller device having a variable gain means
for varying the step gain and the forgetting factor in the noise cancellation device.
[0043]
According to the first aspect of the invention, the noise component mixed in the information
signal is canceled based on the energy wave from the energy wave generating means generating
the energy wave driven by the drive signal from the drive signal source. The level control means
controls the level of the information signal, and the adaptive filter generates a pseudo noise
signal correlated with the noise component mixed in the information signal based on the energy
wave from the energy wave generation means, and subtracting means The pseudo noise signal is
subtracted from the information signal output from the level control means to obtain an
information signal in which the noise component is offset, and the information signal in which
the noise component is offset by the limiter means is less than a predetermined threshold level.
Can be controlled by the control means that supplies the adaptive filter and controls the drive
signal source. The gain unit varies the step gain and the forgetting factor in the noise
cancellation device.
[0044]
According to a second invention, in the noise canceller device of the first invention, the level
control means comprises a multiplier for multiplying an information signal mixed with noise
components by a predetermined gain coefficient, a detection means for detecting an information
08-05-2019
12
signal, and detection thereof Gain generating means for generating a gain coefficient based on
the detection output from the means, and a time constant added to the gain coefficient from the
gain generating means, and the gain coefficient to which the time constant is added is supplied to
the control means Means, and switching means for switching between the gain coefficient to
which the time constant from the time constant adding means is added and the manual gain
coefficient from the control means by the switching control signal from the control means and
supplying it to the multiplier In addition, the variable gain means is controlled by the control
means to add a time constant added gain coefficient (or a manual gain coefficient) and a time
constant added gain coefficient. The value obtained by performing logarithm operation on the
ratio to the one delayed for a predetermined time (or the manual gain coefficient delayed for the
predetermined time), the gain is variable by the parameter control value obtained from the
predetermined conversion map It is a noise canceller device designed to
[0045]
According to a third invention, in the noise canceller device of the second invention, when the
gain coefficient changes, the step gain in the variable gain means in the vicinity of the change
period is a predetermined step gain in other cases. This is a noise canceller device that is
different from the above and increases the noise cancellation pull-in speed.
[0046]
According to a fourth aspect of the present invention, in the noise canceller device of the second
aspect, when the gain coefficient changes, the step gain and the forgetting factor in the variable
gain means in the vicinity of the changing period are predetermined respectively in the other
cases. This noise canceller device is configured to increase the noise cancellation pull-in speed by
making it different from the step gain and the predetermined forgetting factor.
[0047]
According to a fifth invention, in the noise canceller device of the second invention, when the
gain coefficient is near zero, the step gain and the forgetting coefficient in the variable gain
means in the vicinity of that period are predetermined respectively in other cases. Different from
the step gain and the predetermined forgetting factor, so that the variable gain means does not
perform noise cancellation.
[0048]
A sixth aspect of the invention is a noise that cancels the noise component mixed in the
information signal based on the energy wave from the energy wave generating means that
generates the energy wave driven by the drive signal from the drive signal source. In the
08-05-2019
13
cancellation method, the level of the information signal is controlled, and based on the energy
wave from the energy wave generation means, a pseudo noise signal having a correlation with
the noise component mixed in the information signal is generated using an adaptive filter. The
pseudo noise signal is subtracted from the information signal output from the above to obtain an
information signal in which the noise component is canceled, and the information signal in which
the noise component is canceled is suppressed to a predetermined threshold level or less and
supplied to the adaptive filter. Step control gain in the noise cancellation method by the control
of the control means for controlling the drive signal source. This is a noise cancellation method in
which the forgetting factor is made variable.
[0049]
A seventh invention is the noise cancellation method according to the sixth invention, wherein
the information signal level is obtained by multiplying the information signal mixed with the
noise component by a predetermined gain coefficient, detecting the information signal, and
detecting the detection output of the information signal. A gain coefficient is generated based on
this, a time constant is added to the generated gain coefficient, the gain coefficient to which the
time constant is added is supplied to the control means, the gain coefficient to which the time
constant is added, and the manual from the control means The gain factor is switched by the
switching control signal from the control means and supplied to the multiplier, and the step gain
and the forgetting factor are varied by the gain factor (or manual gain factor) to which the time
constant is added by the control means. The ratio of the time constant-added gain factor to the
delayed for a predetermined time (or the manual gain factor delayed for a predetermined time) is
logarithmically calculated. Te is the noise canceling method to be variable gain at the parameter
control value determined from a predetermined conversion map.
[0050]
In an eighth invention according to the noise cancellation method of the seventh invention, when
the gain coefficient changes, the step gain in the vicinity of the change period is made different
from the predetermined step gain in the other cases, and noise is caused. This is a noise
cancellation method in which the pull-in speed of cancellation is increased.
[0051]
According to a ninth aspect of the present invention, in the noise canceling method of the
seventh aspect, when the gain coefficient changes, the step gain and the forgetting factor in the
vicinity of the changing period are given predetermined step gains and predetermined in other
cases. This is a noise cancellation method in which the drawing speed of noise cancellation is
increased, unlike the oblivion factor of.
08-05-2019
14
[0052]
According to a tenth aspect of the present invention, in the noise canceling method of the
seventh aspect, when the gain coefficient is near zero, the step gain and the forgetting factor in
the vicinity of that period are respectively given predetermined step gain and predetermined in
other cases. This is a noise cancellation method in which noise cancellation by variable gain is
not performed unlike the oblivion factor of.
[0053]
BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, with reference to FIG. 6, an
example of a noise canceller apparatus and a noise cancellation method according to an
embodiment of the present invention will be described in detail.
In FIG. 6, the parts corresponding to those in FIG.
In FIG. 6, reference numerals 20 and 21 denote an Lch (left channel) microphone and an Rch
(right channel) microphone which constitute a built-in stereo microphone, respectively.
The type and number of microphones are arbitrary.
The microphones 20 and 21 are disposed close to a helical scan type magnetic recording and
reproducing apparatus of a VTR integrated camera.
[0054]
In the magnetic recording and reproducing apparatus, reference numeral 41 denotes a rotary
drum, and a pair of rotary magnetic heads 44 are provided on the main surface thereof at an
angle of, for example, 180 °.
Then, the magnetic tape 43 is guided so as to be obliquely wound around the rotating drum 41
and a fixed drum (not shown).
08-05-2019
15
Then, noise noise and vibration from the magnetic recording and reproducing apparatus are
mixed with normal sound as so-called mechanical noise and collected by the microphones 20 and
21.
The mechanical noise is caused by the tapping noise when the magnetic tape 43 and the rotary
magnetic head 44 contact, the electromagnetic noise of the drum motor 42 driving the rotary
drum 41 and the like, and the frequency of the mechanical noise generated is the drum motor
42. Harmonics of the rotational frequency of
[0055]
In order to control a predetermined frequency and phase of the rotation of the drum motor 42,
the servo signal 45 from the microcomputer 46 is supplied to the drum mode 42, whereby the
drum mode 42 is controlled so as to always keep a constant rotation. Be done.
Incidentally, the recording / reproducing apparatus is not limited to the VTR apparatus as shown
in the figure, but is applicable to a disk drive apparatus such as a hard disk drive apparatus or an
optical disk drive apparatus having a drive apparatus controlled by a servo signal. It is possible.
[0056]
In FIG. 6, the mixed signal of the audio signal and the mechanical noise signal from the
microphones 20 and 21 is pre-amplified by the amplifiers (AMPs) 22 and 23 and then supplied
to the A / D converters 24 and 25 for analog-to-analog conversion. The digital signal is converted
into a digital signal, which is input to the level control means 55.
[0057]
The level control means 55 normally compresses the D range of the microphone input (generally
about 120 dB) to the proper D range (about 90 dB in one example) according to the signal
processing in the latter stage and the specifications of the recording / reproducing apparatus.
Inserted into
08-05-2019
16
In this case, an AGC (Automatic Gain Control) (automatic gain control) operation described later
is performed.
The level control means 55 is also configured to allow the user etc. to set the gain for setting to
an arbitrary microphone input level. In this case, an MGC (Manual Gain Control) (manual gain
control) operation described later is performed. And can be switched with the AGC operation.
[0058]
Furthermore, the level control means 55 outputs the level detection signal 50 to the
microcomputer 46, and the control signal 51 from the microcomputer 46 is inputted to the level
control means 55, and the user etc. An operation switch (operation SW) 52 for setting the gain is
connected to the microcomputer 46.
[0059]
In the example of FIG. 6, the level control means 55 is provided downstream of the A / D
converters 24 and 25 and is constituted by digital level control means for performing level
control of digital signals. The level control means 55 may be provided upstream of the A / D
converters 24 and 25, and the level control means 55 may be constituted by analog level control
means.
[0060]
A digital signal from the A / D converters 24 and 25 level-controlled by the level control means
55 is supplied to the mechanical noise reduction processing circuit 40.
The mechanical noise reduction processing circuit 40 performs independent processing with two
stereo channels, and is configured to be able to perform optimal mechanical noise reduction
processing even if, for example, noise components differ between Lch and Rch.
If the noise components of the monaural signal, Lch and Rch signals are almost the same, one
channel can be configured.
08-05-2019
17
[0061]
The mechanical noise reduction processing circuit 40 inputs the digital signals from the A / D
converters 24 and 25 level-controlled by the level control means 55 to the positive terminals of
the adders 26 and 27, respectively. The pseudo noise signal generated by adaptive signal
processing to be described later, which is input to the negative side terminal, is subtracted from
the pseudo noise signal, and is output as Lch and Rch output signals subjected to noise
cancellation from the output terminals 36 and 37, respectively.
Furthermore, the Lch and Rch output signals are input to the limiters 34 and 35, respectively,
and Lch and Rch are input to the adaptive processing circuit to input only noise signal
components that are normally at a low level with respect to the dynamic range of voice. When
the level of the output signal is a large level, processing for applying a limiter is performed and
input to the step gain control circuit (SG) 32, 33.
[0062]
The step gain control circuits (SG) 32 and 33 perform processing of multiplying the outputs of
the limiters 34 and 35 by the step gain μ in the above equation (2).
The step gain control signal 48 from the microcomputer 46 is supplied to the step gain control
circuits (SG) 32 and 33 so that the step gain μ can be varied by the step gain control signal 48.
Here, control is performed using the step gain control signal 48 common to the step gain control
circuits (SG) 32 and 33. However, control may be performed independently for each of Lch and
Rch.
[0063]
Further, the signals from the step gain control circuits (SG) 32 and 33 are input to the adaptive
signal processing circuits 28 and 29 as the residual signal E in FIG.
08-05-2019
18
Here, the adaptive signal processing circuits 28 and 29 are configured as in FIG. 2 and processed
by the LMS algorithm.
The drum reference signal 47 similar to the reference input signal X in FIG. 1 is input from the
microcomputer 46 to the adaptive signal processing circuits 28 and 29. The drum reference
signal 47 is a servo signal of the drum motor 42. Because of the 45 base frequency signals, in the
adaptive signal processing circuits 28 and 29, all signal components correlated with this
reference signal, that is, the drum rotation frequency and its harmonic components are targets
for adaptive processing.
Furthermore, the output signals corresponding to the adaptive output signal Y in FIG. 1 from the
adaptive signal processing circuits 28 and 29 are input to the negative terminals of the adders
26 and 27, respectively.
[0064]
The adaptive noise canceller shown in FIG. 6 is characterized in that a forgetting factor λ is
newly introduced and the forgetting factor λ is variably controlled by the microcomputer 46.
More specifically, the following two points are shown in FIG. This is different from the
conventional example of
[0065]
First, in the adaptive signal processing circuits 28 and 29, the adaptive filter coefficients W0,
W1,..., Wm shown in FIG. 2 are updated according to the following equation 3 respectively.
[0066]
Wk + 1 = λ · Wk + 2μ · Ek · Xk
[0067]
Here, assuming that the adaptive filter coefficient in the current unit sampling is Wk + 1,
equation (3) multiplies Wk, which is the adaptive filter coefficient in the past of one unit
sampling, by the coefficient λ.
08-05-2019
19
This factor λ is called the forgetting factor, and is usually set to a value slightly smaller than 1 to
reduce the weight on past data and perform adaptive processing to discard past data as it gets
farther from the present. It can be operated.
For example, if the residual signal Ek or the reference input Xk suddenly becomes zero in the
equation (3) when the adaptation is sufficiently advanced, for example, the drum motor performs
mode transition from the rotation state to the stop state to generate mechanical noise. If the
second term of the right-hand side in the equation 3 becomes zero when there is no such thing,
etc., Wk of the first term will remain forever without the forgetting factor λ, and the sense of
incongruity may occur in hearing, but forgetting The response of such a case is improved over
that of the prior art because Wk gradually decays and finally becomes zero when there is a factor
λ.
However, conversely, when the drum motor makes mode transition from the stop state to the
rotation state, the responsiveness of the adaptive processing deteriorates, so the forgetting factor
λ is set to 1 so that there is no influence in equation (3). Is good.
[0068]
Next, the second point is stopping the on / off switching of the switches (SW) 30 and 31 by the
cancel on / off control signal 49 of FIG. 3. The merit by stopping this is that the feedback loop is
not cut off at the off The continuity of processing is to be maintained.
That is, when the switches (SWs) 30 and 31 are controlled as on → off → on, for example, the
mode transition of the drum motor from the rotation state to the stop state and the mode
transition to the rotation state again is considered. Since there is no occurrence of mechanical
noise that is the target of adaptive processing in the above, it is necessary to turn off the switches
(SW) 30 and 31 and the adaptive processing loop is also cut at the same time. If the coefficient
changes to a large value, the response to stabilization becomes worse when the switches (SW)
30, 31 are turned on again.
[0069]
Therefore, in the adaptive control noise canceller shown in FIG. 6, the forgetting factor λ and the
step gain μ are controlled by the microcomputer 46 to realize cancellation on / off without
breaking the feedback loop.
08-05-2019
20
[0070]
Accordingly, in FIG. 6, the forgetting factor λ control signal 49 is input from the microcomputer
46 to the adaptive signal processing circuits 28 and 29 in the mechanical noise reduction
processing circuit 40, and the adaptive signal processing circuits 28 and 29 follow the equation
The factor Wk of each adaptive filter is multiplied by the forgetting factor λ for each unit
sampling.
[0071]
In the above-mentioned JP-A-11-232802 and JP-A-11-232803, ON / OFF of mechanical noise
cancellation is performed by the cancel ON / OFF control signal, but in this case, the feedback
loop is cut off at the time of OFF. Therefore, there is a problem that the continuity of the adaptive
processing can not be maintained.
That is, the coefficient of the adaptive filter is not determined during this OFF period, and if the
coefficient changes significantly during this period, the responsiveness to stability when the
switch is turned ON is deteriorated.
[0072]
Therefore, in the adaptive control noise canceller device of FIG. 6, the forgetting factor λ and the
step gain μ are controlled by the microcomputer 46, and cancellation on / off is realized without
breaking the feedback loop.
[0073]
Next, the specific configuration of the level control means 55 in the adaptive control noise
canceller device of FIG. 6 will be described with reference to FIG.
The Lch digital signal and the Rch digital signal from the A / D converters 24 and 25 of FIG. 6 are
input to the input terminals 60 and 61, respectively.
08-05-2019
21
The Lch digital signal is supplied to AGC (automatic gain control) means 62 for Lch and MGC
(manual gain control) means 63 for Lch. Also, the Rch digital signal is supplied to Rch AGC means
64 and Rch MGC means 65.
[0074]
The level detection signal 50 obtained from the AGC means 62, 64 is supplied to the
microcomputer 46 of FIG. Further, the control signal 51 from the microcomputer 46 is supplied
to the MGC means 63, 65 through the terminal 73, and is also supplied as a switching control
signal to changeover switches (SW) 66, 67 described later.
[0075]
The level-controlled digital signals from Lch AGC means 62 and MGC means 63 are supplied to
switch 66 and switched, and then supplied to amplifier (AMP) 68 for amplification, and the
amplified output is output to Lch output. As the output terminal 70. Also, after the levelcontrolled digital signals from the AGC means 64 and the MGC means 65 of Rch are supplied to
the switch 67 and switched, they are supplied to the amplifier (AMP) 69 and amplified, and the
amplified output is amplified. It is output from the output terminal 71 as an Rch output.
[0076]
The control signal 51 from the terminal 73 may be controlled to operate in the same state for
each of the Lch and Rch circuits, or may be controlled to operate in each different state.
[0077]
Next, a specific configuration example (one channel) of the AGC means 62 and 64, the MGC
means 63 and 65, and the switches 66 and 67 in FIG. 7 will be described with reference to FIG.
The digital audio signal from the input terminal 80 is supplied to one input terminal of the
multiplier 82 through the delay unit 81 and to the detection means 83.
08-05-2019
22
[0078]
The detection means 83 detects the peak level or the average value level of the digital audio
signal at predetermined time intervals, and the detected signal is supplied to the gain generation
means 84 to generate a gain coefficient. The gain generation means 84 sets the AGC operation
level by the AGC operation level setting signal from the input terminal 93. That is, for example,
the detection signal level from the detection means 83 and the AGC operation level are
compared, and when the detection signal level is below the AGC operation level, a predetermined
gain coefficient is output and the detection signal level exceeds the AGC operation level. In this
case, the digital audio signal is suppressed to a predetermined level in a multiplier 82 described
later by reducing the gain coefficient by a predetermined amount than a predetermined value.
[0079]
The gain coefficient generated in this way is supplied to, for example, a time constant addition
circuit 85 configured of an LPF (low pass filter), and a time constant is added such that the gain
coefficient changes smoothly in human hearing. The A coefficient 92 for the AGC operation is
supplied to one input terminal of the changeover switch (SW) 86 and is output from the output
terminal 89 as the level detection signal 50. Further, among the control signals 51 from the input
terminal 88, the other input terminal of the changeover switch 86 is supplied with an M
coefficient 90 which is a gain coefficient for the MGC operation. Furthermore, the control signal
51 from the microcomputer 46 is supplied as the switching control signal 91 to the switch 86 to
control the switching of the switch 86, and one of the A coefficient 92 and the M coefficient 90 is
the other of the multiplier 82. Input to the terminal.
[0080]
The multiplier 82 multiplies the digital audio signal by the selected gain factor. Assuming that
the gain coefficient is, for example, an arbitrary coefficient of 0 or more and 1 or less, the digital
audio signal output outputted from the output terminal 87 is a digital audio signal of zero level
when the gain coefficient is 0, and is 1 when Is a digital audio signal at the same level as the
input digital audio signal, and when the gain coefficient is an intermediate value, the digital audio
signal has a level corresponding to the coefficient. Here, the delay unit 81 is for delaying the
input digital audio signal so that there is no time shift between the audio signal input to the
multiplier 82 and the gain coefficient. The switch 86 corresponds to the switches 66 and 67 in
FIG.
08-05-2019
23
[0081]
Next, referring to FIG. 9, in microcomputer 46 of FIG. 6, the step gain μ control signal is
obtained from the gain coefficient in level detection signal 50 in the case of AGC operation or in
the case of MGC operation in the case of MGC operation. An example of a functional block for
generating the parameters 48 and the parameter of the forgetting factor λ control signal 49 will
be described.
[0082]
The gain coefficient in the AGC / MGC operation from the input terminal 100 is input as the
signal S to the input terminal 107 of the log (LOG) arithmetic unit 102, and the gain coefficient
thereof is a predetermined unit sample time delay (D-1) The signal is supplied to 101 and
delayed, and the delayed signal is input to the input terminal 106 of the LOG operator 102 as a
signal P.
[0083]
The LOG operator 102 performs LOG (common logarithm) operation of the following equation 4
and outputs it as a signal Q from a terminal 108.
[0084]
Q = log (S / P)
[0085]
Here, assuming that the gain coefficient input to the input terminal 100 is an arbitrary real
number coefficient of 0 or more and 1 or less as described above, the output signal Q is the
current sample signal S with respect to the signal P of unit sample past. If it is large, it becomes a
positive value, if it is small, it becomes a negative value, and if the same, it becomes zero.
That is, the change amount of the gain coefficient to be input is determined by the LOG
computing unit 102, and the determination result is input to the parameter generation unit 103
in the subsequent stage.
08-05-2019
24
[0086]
The parameter generation means 103 is a functional block that generates the step gain μ
control signal 48 and the forgetting factor λ control signal 49 of the above-mentioned
parameters from the signal Q to be input, and shows a conversion table of the Q signal and the
above-mentioned parameter values. When the conversion map 104 is provided, for example,
when the gain coefficient in FIG. 12A described later changes from the past to the present due to
attack / recovery or gain down / gain up, the Q value is adjusted according to the amount of
change. The parameter value corresponding to the Q value is selected from the conversion map
104, and the parameter output is obtained from the output terminal 105.
Therefore, according to the functional block example of FIG. 9, since the parameter is adaptively
selected by the change of the gain coefficient, the optimal parameter control can always be
performed.
[0087]
In the example of FIG. 9, the LOG operator 102 is separately prepared, but in order to simplify
the process, for example, Q = S / P is calculated by providing a log (LOG) conversion table in the
conversion map 104. , And may be selected from the LOG conversion map to output parameters.
[0088]
Next, referring to FIG. 10, in the adaptive control noise canceller device of FIG. 6, the AGC
operation of the level control means 55 when mechanical noise signals are mixed in the audio
signals from the microphones 20 and 21 is mechanical noise cancellation processing. Explain the
impact on
[0089]
FIG. 10A shows the case where the input voice signal is below the AGC operation level, FIG. 10B
shows the case where the input voice signal is at a high level as if it exceeds the AGC operation
level, (a) is the microphone input, (b) 11 (c) shows examples of signal waveforms after
mechanical noise cancellation processing after AGC processing of the level control means 55.
FIG.
[0090]
First, when the digital audio signal changes in level stepwise at time point T, the mechanical
08-05-2019
25
noise signal is superimposed on the audio signal and changes at time point T because the
frequency of the mechanical noise signal is higher than the level change frequency of the audio
signal. Is amplitude modulated by the audio signal to be
[0091]
In the case of (a) of FIG. 10A, since the microphone input is below the AGC operation level
indicated by the broken line, the gain coefficient of AGC processing becomes a constant value,
and the amplitude of the mechanical noise signal after AGC processing of (b) is Even if the audio
signal level changes, it does not change.
Therefore, after the mechanical noise cancellation processing of (c), since the mechanical noise
signal level is constant, the noise signal can be canceled stably.
[0092]
Next, (a) of FIG. 10B is a case where the microphone input changes to the AGC operation level
shown by the broken line at time T or more, and in this case after AGC processing of (b) Although
the gain factor is reduced and the audio level is suppressed to below the AGC operation level, the
amplitude of the mechanical noise signal is also reduced at this time.
Therefore, after the mechanical noise cancellation process in (c), the mechanical noise signal with
a constant amplitude is stably canceled until time point T, but the amplitude of the mechanical
noise signal decreases after time point T, and the period until mechanical noise cancellation
process follows In this case, the cancellation operation is overcompensated, and the mechanical
noise signal is not canceled but remains (untilly pulled in) until it follows it.
[0093]
Also, in the AGC operation for the audio signal, the dynamic range in the subsequent stage is not
broken due to the rapid level increase of the audio signal at the so-called attack time, as
described in FIG. The gain coefficient is lowered in response to the level change in about a
second, but in the subsequent recovery (also called release), the gain coefficient is usually
08-05-2019
26
gradually taken over several seconds because there is no concern such as breakdown of the
dynamic range in the latter stage. Recover.
Therefore, in most cases, the attack operation as shown in FIG. 10 is the problem in tracking the
mechanical noise cancellation.
[0094]
Next, referring to FIG. 11, in the adaptive control noise canceller device of FIG. 6, the MGC
operation of the level control means 55 when mechanical noise signals are mixed in the audio
signals from the microphones 20 and 21 is mechanical noise cancellation processing. Explain the
impact on
[0095]
FIG. 11A shows the case where the level of the input audio signal is raised, FIG. 11B shows the
case where the level of the input audio signal is lowered, (a) shows the microphone input, and (b)
shows the MGC processing of the level control means 55. (C) shows an example of each signal
waveform after mechanical noise cancellation processing.
[0096]
Since the level of the audio signal is small as shown in (a) of FIG. 11A, the user operates the
operation switch (SW) 52 in FIG. 6 to set the gain coefficient at time T by the MGC means 63 and
65 of the level control means 55. When the level of the audio signal is increased, the amplitude
of the mechanical noise signal mixed with the audio signal also increases as in the MGC
processing of (b).
Furthermore, after (c) mechanical noise cancellation processing, mechanical noise was stably
canceled before time T, but after time T, the cancel operation is insufficient and mechanical noise
is not canceled and remains until it follows Resulting in.
[0097]
In FIG. 11B, the user operates the operation switch (SW) 52 in FIG. 6 because the voice level of
08-05-2019
27
the microphone input is large as shown in (a), and MGC means 63 and 65 of the level control
means 55 If the gain factor is reduced and the voice level is lowered, the amplitude of the
mechanical noise signal mixed with the voice signal as in the MGC processing of (b) is also
reduced to a very small level, (c) After the mechanical noise cancellation processing of, the
mechanical noise is stably canceled before time T, and the cancellation operation becomes
overcompensated after time T, and the mechanical noise is not canceled but remains until it
follows A problem occurs.
Therefore, the present invention solves the problem of mechanical noise cancellation in the AGC
/ MGC operation as described above.
[0098]
Next, referring to FIG. 12, an example of parameter control in the AGC / MGC operation of the
present invention will be described.
First, FIG. 12A shows the time change of the gain coefficient of the AGC / MGC operation in FIG.
6 and the amount of generated noise at that time.
The gain coefficient is a coefficient that is output from the switch (SW) 86 in FIG. 8 and that is
multiplied by the input speech signal in the multiplier 82.
Further, the amount of generated noise is the amount of noise when mechanical noise
cancellation is not performed. First, the AGC operation is started from time point T0, and in the
switches (SW) 66 and 67 of FIG. 7, it is assumed that the AGC means 62 and 64 are selected to
operate with a predetermined gain coefficient.
[0099]
Next, at time T1, if a voice (not shown) of a relatively high level equal to or higher than the AGC
operation level is input to the microphone, the gain is increased in the attack operation to reduce
the audio level to near the AGC operation level by the AGC operation. Decrease the coefficient.
Accordingly, the amount of generated noise also decreases, and thereafter, when the voice
08-05-2019
28
disappears at time T2, the gain coefficient gradually recovers by the recovery operation, and the
amount of noise also increases accordingly.
[0100]
Next, after completion of recovery, the switches (SW) 66 and 67 in FIG. 7 are selected on the
MGC means 63 and 65 side. Thereafter, at time T3, when the gain coefficient is reduced by a
predetermined amount to reduce the gain, the amount of noise also decreases. Thereafter, when
the gain is decreased again at time T4, and then the gain coefficient is operated to zero at time
T5, the noise amount also decreases according to the gain coefficient, and the noise amount also
becomes zero when the gain coefficient becomes zero.
[0101]
Next, when the gain coefficient is increased by a predetermined amount at time T6, the amount
of noise also increases, and then the gain is increased again at time T7, and then the gain
coefficient is further increased at time T8. It shows that the amount of noise also increases
according to the gain coefficient when the operation is performed to return to the initial value.
Thus, the amount of generated noise fluctuates according to the change of the gain coefficient.
[0102]
Next, FIGS. 12B to 12D show an example in which mechanical noise cancellation processing is
performed on these noise fluctuations. FIG. 12B shows the case where the parameters of the step
gain μ and the forgetting factor λ are fixed to predetermined values in the above equation 3,
and the amount of noise after mechanical noise cancellation at this time is first from time T0 to
time T1. Assuming that noise cancellation is stable, if the amount of generated noise decreases in
the AGC attack operation at time T1, the mechanical noise cancellation operation can not follow
this attack time constant, and the cancellation operation becomes overcompensated and noise
The amount increases and gradually converges at a following speed determined by a
predetermined step gain μ, and the amount of noise decreases, but when the amount of
generated noise gradually increases in the AGC recovery operation, this corresponds However,
the amount of noise also decreases.
08-05-2019
29
[0103]
Next, when the amount of generated noise decreases at time T3 of the MGC operation, the
mechanical noise cancellation operation can not follow again, the cancellation operation becomes
overcompensated and the noise amount increases, and then the tracking determined by the
predetermined step gain μ The speed gradually converges and the amount of noise decreases,
but the amount of noise increases again at time T4, and the amount of noise also increases at
time T5.
[0104]
Furthermore, when the amount of generated noise increases at time T6, the mechanical noise
canceling operation can not follow again, and the amount of noise increases by that amount, and
then gradually converges at the following speed determined by the predetermined step gain μ
Although the amount decreases, the noise amount increases again at time T7, and the noise
amount also increases at time T9.
[0105]
Next, FIG. 12C shows a case where the step gain μ is shown by a broken line and is increased
when the gain coefficient changes to improve pull-in, and at this time, the forgetting factor λ is
fixed to a predetermined value.
Here, since the follow-up speed increases after the amount of generated noise changes, the
amount of noise can be generally suppressed as compared with FIG. 12B.
[0106]
Next, FIG. 12D shows the case where the forgetting factor λ is also controlled as shown by the
solid line together with the control of the step gain μ, and the amount of noise generated first
decreases at time T1 in the AGC operation. When the step gain μ is increased and the forgetting
factor λ is smaller than a predetermined amount, the weight of the past coefficient can be
reduced by the adaptive filter coefficient, and the weight of the current coefficient can be
increased. Updating is further promoted, and by optimizing parameter control of the step gain μ
and the forgetting factor λ, an increase in the amount of noise can be almost suppressed.
[0107]
Also, in the MGC operation, when the amount of noise similarly increases between time point T3
08-05-2019
30
and time point T8, the step gain μ is increased, and when the forgetting factor λ is smaller than
a predetermined amount, the adaptive filter coefficient becomes the past coefficient Since the
weight can be reduced and the weight of the current coefficient can be increased, coefficient
updating in the adaptive filter is further promoted, and by optimizing parameter control, almost
all increase in noise amount can be suppressed.
[0108]
As described above, in the present invention, by setting the optimum parameters in the steady
state and controlling the parameters only when the amount of generated noise changes, the
mechanical noise signal level is not deteriorated without the stability in the steady state. Even
when it fluctuates, the follow-up property of mechanical noise cancellation can be improved, and
mechanical noise cancellation can be performed stably.
[0109]
Next, an example of VTR mode transition and parameter control will be described with reference
to FIG.
FIG. 13 shows an example of control at the time of transition of the VTR mode as in FIG.
In the present invention, by newly introducing the forgetting factor λ, the weight of the past
adaptive filter coefficient can be reduced and the weight of the current adaptive filter coefficient
can be increased, so that the followability when noise changes is improved. Be done.
Therefore, the adaptive filter coefficient is multiplied by the predetermined forgetting factor λ
which is slightly smaller than coefficient 1 in FIG. 13B with respect to the mode transition and
the amount of generated noise (same as FIG. 4A) in FIG. 13A. The amount of noise at the time of
mode transition is reduced as compared with.
[0110]
Next, FIG. 13C shows the case where the step gain μ value and the forgetting factor λ value are
controlled at the time of mode transition, and the step gain μ is increased at the time of mode
08-05-2019
31
transition to improve followability, and conversely, the forgetting factor λ is decreased. In this
way, it is easy to accelerate the forgetting of the past coefficient and follow the current
coefficient.
As a result, the noise amount hardly deteriorates even at the time of mode transition, and can
follow changes in noise. Further, FIG. 13D shows a parameter control method in the case where
the adaptive filter coefficient in the recording mode is always held and used as shown in FIG. 4D.
In this case, the step gain μ is made zero in the case other than the recording mode. The
updating of the adaptive filter coefficient is stopped, the forgetting coefficient λ is set to 1, and
the adaptive filter coefficient in the recording mode is held.
[0111]
Next, another example of VTR mode transition and parameter control will be described with
reference to FIG. FIG. 14 shows an example of control at the time of VTR mode transition similar
to FIG. First, with respect to the VTR mode transition shown in FIG. 14A and the amount of
generated noise (same as FIG. 4A), FIG. 14B sets the step gain μ to zero when the VTR mode
changes from recording to recording standby (stop). By stopping the updating of the adaptive
filter coefficient and reducing the forgetting coefficient λ to accelerate the forgetting of the
coefficient, the drum rotation noise and the tapping noise between the magnetic tape 43 and the
rotation magnetic head 44 are abrupt. In the next transition from recording standby to recording,
the step gain μ is made larger than a predetermined value to improve the followability of the
adaptive filter coefficient, and the forgetting factor λ is 1 By stopping the oblivion of the
coefficient, the drum rotation noise and the tapping noise between the magnetic tape 43 and the
rotation magnetic head 44 are rapidly generated from the noise zero state during the mode
transition. It is improved trackability even if the amount's increases, the amount of noise is
reduced as compared with the conventional example of FIG. 4B.
[0112]
Next, FIG. 14C shows the case where the cancellation is controlled to be turned off as the noise
generation disappears in the recording standby mode. First, when the VTR mode shifts from
recording to recording standby (stop), By stopping the adaptive operation by setting both the
step gain μ and the forgetting factor λ to zero, the adaptive outputs of the adaptive signal
processing circuits 29 and 30 in FIG. It is possible to follow the change in noise caused by the
rapid disappearance of the drum rotation noise and the tapping noise between the magnetic tape
08-05-2019
32
43 and the rotation magnetic head 44. Next, when transition from recording standby to
recording is performed, the step gain μ is made larger than a predetermined value to improve
the followability of the adaptive filter coefficient, and the forgetting coefficient λ is made 1 to
stop forgetting the coefficient, During the mode transition, the drum rotation sound and the head
striking sound are generated suddenly from the noise zero state, and even if the noise amount
increases, the noise amount can be followed even if the mode transition is performed, and the
noise amount is suppressed. In the recording standby mode, the amount of noise can be made
zero, and furthermore, there is no problem due to the fact that the feedback loop in the
conventional cancel on / off control is broken.
[0113]
Next, FIG. 14D shows a control method for holding noise in the recording mode as in the
conventional example. First, when the VTR mode transitions from recording to recording standby
(stop), recording is performed at the start of mode transition. In order to hold the noise of the
mode, the step gain μ is made zero, and at the same time the forgetting factor λ is made 1, the
adaptive filter coefficient in the recording mode is held. At the start of recording, the step gain μ
and forgetting factor λ are canceled to predetermined recording values to output an adaptive
output with the adaptive filter coefficients held until then, and in the recording mode, constant
noise cancellation is always performed. Is made.
[0114]
As described above, according to the present invention, in addition to the control of the
conventional step gain μ, the cancellation on / off can be performed without breaking the
feedback loop by controlling the forgetting factor λ, thereby following the adaptive processing
more than the conventional example. It becomes possible to improve the sex.
[0115]
The mechanical noise cancel on / off operation of the present invention can also be applied to
the AGC / MGC operation described with reference to FIG.
For example, when an audio signal of a very large level is input in AGC operation, an attack
operation occurs in order to optimize the level of this audio signal, and mechanical noise is
reduced to a level almost inaudible. Rather, it is better to cancel out by setting the step gain μ
and the forgetting factor λ to zero as described above.
08-05-2019
33
Also, in the MGC operation, when the gain coefficient is zero (between time point T5 and time
point T6 in FIG. 12A) or near zero, the mechanical noise is reduced to a level hardly audible on
the aural sense. As described above, it is better to cancel out by setting the step gain μ and the
forgetting factor λ to zero.
[0116]
Even when the microphone and the recording and reproducing apparatus are arranged relatively
close to each other as in a camera-integrated type VTR in recent years, according to the abovedescribed noise canceller and noise canceling method, mechanical noise generated from the
recording and reproducing apparatus is When mixing into the microphone, even if the
mechanical noise level fluctuates with the audio signal in the AGC operation or the MGC
operation, it is possible to efficiently cancel only the mechanical noise.
[0117]
According to the noise canceller and noise cancellation method described above, the change
amount (slope of change) of the gain coefficient in AGC operation or MGC operation is extracted
by LOG operation, and further, the step gain By generating control parameters of and the
forgetting factor λ, parameter control in accordance with changes in gain factor becomes
possible, and furthermore, optimal noise cancellation is possible.
[0118]
According to the noise canceller apparatus and the noise cancellation method described above,
the step gain and the cancellation on / off control example of the adaptive processing in JP-A-11232802 and JP-A-11-232803 are newly forgotten. By introducing the coefficient λ, it is possible
to weight the current adaptive filter coefficient with respect to the past, and the followability to
the noise change is improved.
Furthermore, since the number of parameters to be controlled does not change either,
microcomputer processing does not become complicated.
[0119]
The cancellation on / off control in the conventional example cuts off the feedback loop in the
adaptive processing to put it into the off mode, so that the coefficient of the adaptive filter tends
08-05-2019
34
to fly out during off, and becomes stable when the adaptive processing starts again. Although it
took time, according to the noise canceller and noise cancellation method described above, the
adaptive processing can be turned off without breaking the feedback loop by controlling the
forgetting factor, and the followability of the adaptive processing is improved. Be done.
[0120]
According to the above-described noise canceller and noise cancellation method, all digital
processing is possible, so that hardware such as DSP (digital signal processor) or LSI and
software such as microcomputer are easy to realize, and semiconductor miniaturization in the
future With the increase in density and memory, the increase in circuit scale can be realized with
almost no problems.
[0121]
In the above example, although the present invention is applied to the noise cancellation
processing of the built-in microphone in the video camera, the present invention is generally
applicable to adaptive signal processing employing an adaptive filter, and further, LMS algorithm
It can also be used in other RLS (Recursive Least Square) algorithms and the like.
[0122]
The present invention can be applied not only to a VTR integrated camera such as a video
camera but also to an apparatus having a disk recording and reproducing apparatus such as a
hard disk recording and reproducing apparatus and an optical disk recording and reproducing
apparatus.
[0123]
According to the first aspect of the invention, the noise component mixed in the information
signal based on the energy wave from the energy wave generating means for generating the
energy wave driven by the drive signal from the drive signal source. In the noise canceller
apparatus for canceling the noise, the pseudo noise signal correlated with the noise component
mixed in the information signal based on the level control means for controlling the level of the
information signal and the energy wave from the energy wave generating means An adaptive
filter to be generated, a subtraction means for obtaining an information signal in which a noise
component is canceled by subtracting a pseudo noise signal from an information signal output
from the level control means, a predetermined threshold for the information signal in which the
noise component is canceled Limiter means for supplying to the adaptive filter after suppressing
to below the hold level, and control for controlling the drive signal source Since it has step
control in the noise cancellation apparatus controlled by the stage and variable gain means for
08-05-2019
35
changing the forgetting factor, a feedback loop is introduced by introducing a forgetting factor to
determine the weight for the past data and adaptively controlling it. The adaptive processing can
be turned off without turning off, the followability of the adaptive processing can be improved,
and the followability to the noise conversion can be enhanced, and the noise cancellation can be
performed by controlling the step gain and the forgetting factor. By realizing the mode, the
disturbance of the adaptive filter coefficient does not occur, and from this point as well, it is
possible to improve the followability to the noise change, and in spite of the fluctuation of the
noise component level due to the level control of the information signal. A noise canceller
capable of efficiently canceling noise components It is possible.
[0124]
According to the second invention, in the noise canceller device of the first invention, the level
control means includes a multiplier for multiplying the information signal mixed with the noise
component by a predetermined gain coefficient, a detection means for detecting the information
signal, When a gain generation means for generating a gain coefficient based on the detection
output from the detection means and a gain constant from the gain generation means are added
with a time constant, and the gain coefficient added with the time constant is supplied to the
control means A constant adding means, and a switching means for switching over the gain
coefficient to which the time constant from the time constant adding means is added and the
manual gain coefficient from the control means by the switching control signal from the control
means and supplying the multiplier The variable gain means is controlled by the control means
to add a time constant added gain coefficient (or a manual gain coefficient) and a time constant
added gain coefficient. The value obtained by performing a logarithmic operation on the ratio of
the input coefficient to that delayed for a predetermined time (or the manual gain coefficient
delayed for that predetermined time), the parameter control value obtained from the
predetermined conversion map Since the gain can be varied, in addition to the effect of the first
invention, the step gain and the forgetting factor can be varied according to the level change
amount of the noise component by the level control of the information signal automatically and
manually. Since the gain of the variable gain means is varied, it is possible to obtain a noise
canceller device capable of further improving the followability of noise cancellation.
[0125]
According to the third invention, in the noise canceller device of the second invention, when the
gain coefficient changes, the step gain in the variable gain means in the vicinity of the change
period is a predetermined step gain in other cases. In addition to the effect of the second
invention, a noise canceller device capable of improving the ability to follow noise changes when
the gain coefficient changes in addition to the effects of the second invention. You can get it.
08-05-2019
36
[0126]
According to the fourth invention, in the noise canceller device of the second invention, when the
gain coefficient changes, the step gain and the forgetting coefficient in the variable gain means in
the vicinity of the change period are respectively obtained in the other cases. Because the pull-in
speed of noise cancellation is made different from the predetermined step gain and the
predetermined forgetting factor, in addition to the effects of the second invention, by controlling
the forgetting factor, adaptation before noise change is realized. From the state, it is possible to
increase the drawing speed of the new noise to the adaptation state, and to obtain the noise
canceller device capable of improving the followability to the noise change.
[0127]
According to the fifth invention, in the noise canceller device of the second invention, when the
gain coefficient is near zero, the step gain and the forgetting factor in the variable gain means
near that period are obtained in the other cases. In addition to the effects of the second
invention, since the variable gain means does not perform noise cancellation different from the
predetermined step gain and the predetermined forgetting factor, the level of the noise
component is also close to zero. By not cutting the adaptive loop, it is possible to obtain a noise
canceller device capable of performing stable noise cancellation operation when adaptive control
is restarted.
[0128]
According to the sixth aspect of the invention, the noise component mixed in the information
signal is canceled based on the energy wave from the energy wave generating means generating
the energy wave driven by the drive signal from the drive signal source. In the above noise
cancellation method, the level of the information signal is controlled, and based on the energy
wave from the energy wave generating means, a pseudo noise signal having a correlation with
the noise component mixed in the information signal is generated using an adaptive filter. The
pseudo noise signal is subtracted from the information signal output from the control means to
obtain an information signal in which the noise component is offset, and the information signal in
which the noise component is offset is suppressed to a predetermined threshold level or less.
Step control in the noise cancellation method by the control of the control means that Since the
variable and forgetting factor are made variable, by introducing a forgetting factor that
determines the weight to the past data and adaptively controlling it, adaptive processing can be
turned off without breaking the feedback loop, The adaptability of the adaptive processing is
improved, and the adaptability to noise conversion can be enhanced, and by controlling the step
gain and the forgetting factor, the noise cancel off mode is realized, and the disturbance of the
adaptive filter factor occurs. Noise cancellation that can improve the ability to follow noise
changes from this point as well, and can efficiently cancel noise components despite fluctuations
in noise component levels due to level control of information signals You can get the way.
08-05-2019
37
[0129]
According to the seventh invention, in the noise cancellation method of the sixth invention, the
level of the information signal is that the information signal mixed with the noise component is
multiplied by a predetermined gain coefficient to detect the information signal, and the
information signal is detected A gain coefficient is generated based on the output, a time
constant is added to the generated gain coefficient, the gain coefficient added with the time
constant is supplied to the control means, and the gain coefficient added with the time constant
and the control means The manual gain factor is switched by the switching control signal from
the control means and supplied to the multiplier, and the variable of the step gain and the
forgetting factor is a gain factor (or manual gain factor) added with a time constant by the
control means. Obtained by performing a logarithmic operation on the ratio of the time constant
to the time-delayed gain coefficient (or the time-delayed manual gain coefficient) In addition to
the effects of the sixth aspect of the invention, noise components due to automatic and manual
level control of the level of the information signal are obtained because the gain is varied
according to the parameter control value obtained from the predetermined conversion map.
Since the gain of the variable gain means for varying the step gain and the forgetting factor is
varied in accordance with the level change amount of {circle around (1)}, it is possible to obtain a
noise cancellation method capable of further improving the followability of noise cancellation.
[0130]
According to the eighth invention, in the noise cancellation method of the seventh invention,
when the gain coefficient changes, the step gain in the vicinity of the change period is made
different from the predetermined step gain in other cases. Since the pull-in speed of noise
cancellation is increased, in addition to the effect of the seventh invention, it is possible to obtain
a noise cancellation method capable of improving the followability to the noise change when the
gain coefficient changes. .
[0131]
According to the ninth invention, in the noise cancellation method of the seventh invention, when
the gain coefficient changes, the step gain and the forgetting coefficient in the vicinity of the
change period are respectively given predetermined step gains in the other cases. And the
predetermined forgetting factor is different, and the drawing speed of noise cancellation is
increased. Therefore, in addition to the effect of the seventh invention, by controlling the
forgetting factor, the adaptation state before the noise change is newly changed. It is possible to
increase the drawing speed of the noise into the adaptation state, and to obtain this noise
cancellation method capable of improving the followability to noise change.
08-05-2019
38
[0132]
According to the tenth invention, in the noise canceling method of the seventh invention, when
the gain coefficient is near zero, the step gain and the forgetting factor near that period are
respectively given predetermined step gains in the other cases. In addition to the effects of the
seventh invention, the adaptive loop is not cut even when the level of the noise component is
close to zero, because noise cancellation is not performed by variable gain unlike the
predetermined forgetting factor. By doing this, it is possible to obtain a noise cancellation method
capable of performing stable noise cancellation operation when adaptive control is resumed.
08-05-2019
39
1/--страниц
Пожаловаться на содержимое документа