close

Вход

Забыли?

вход по аккаунту

?

DESCRIPTION JP2005159877

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 JP2005159877
PROBLEM TO BE SOLVED: To extract and reduce a noise waveform when noise generated by the
operation is mixed in a microphone from a recording and reproducing apparatus having an
operation with periodicity such as, for example, a rotating drum. SOLUTION: Data obtained from
microphones 6, 7 are converted into digital data by A / D converters 11, 12, and supplied to a
control microcomputer 5 and adders 15, 16. The control microcomputer 5 has a memory for
storing a plurality of input data in mechanical noise generation cycle units, and generates data of
substantially only noise components by calculating an average value of data of the same phase of
the cycle. And the result is stored in the memory 13, 14. The adders 15 and 16 subtract the inphase data stored in the memories 13 and 14 from the input current data to output data from
which noise components have been removed. [Selected figure] Figure 1
Electronic device and video camera device
[0001]
The present invention relates to a technology for processing acoustic data obtained from sound
wave collecting means such as a microphone.
[0002]
As an apparatus of this type, for example, a camera integrated video camera is known.
In the case of such an apparatus, a magnetic tape is used as a recording medium, and in order to
10-04-2019
1
increase the recording density on the recording medium, a mechanism of a helical scan system
using a rotating drum is used. However, with this mechanism, noise called head hitting occurs
when the magnetic tape and the rotating recording head come in contact with each other.
Moreover, in order to switch the energization of the coil of the magnetic motor to drive the
rotary drum, the electromagnetic noise is also generated. These noises were mixed in the built-in
microphone, and were recorded along with the sound to be originally recorded, and it became
unpleasant.
[0003]
As a solution to this point, a noise reduction method is disclosed by examining the noise
spectrum mixed in the microphone in advance and subtracting the component of the band
according to the size of the noise (for example, a patent) Reference 1).
[0004]
According to this proposal, by providing means for extracting the components of the noise
spectrum, by measuring the noise level in a state in which the voice to be originally recorded is
blocked, control is performed so as to have the same magnitude as the mixed noise. The noise
spectrum was only reduced by the level of the spectrum, and when the voice entered, it was less
likely to affect the voice.
JP 7-177596 A
[0005]
However, in such a conventional method, it is necessary to examine in advance the component to
be the noise, and it is necessary to prepare means for extracting noise as many as the number of
components of the noise spectrum, for example, a band pass filter. As a workaround, it is
conceivable to reduce the number of band pass filters by extracting frequencies close in
spectrum together. For example, when the noise spectrum is two 1 kHz and 2 kHz, if the center
frequency is 1.5 kHz and the bandwidth is 1 kHz, 1 kHz and 2 kHz of the noise component can
be extracted. However, in this case, if the levels of 1 kHz and 2 kHz are different, the two average
values are taken to match or not, and they do not have the same size as the noise component,
and the removal effect is diminished. Furthermore, if there is a 1.5 kHz component in the audio
signal to be recorded, that component is also attenuated, which affects the sound quality. In
10-04-2019
2
addition, there is a drawback that it is necessary to re-adjust the level to cancel if the change in
the level matches the change with time, or it becomes difficult to cope with the change in the
noise spectrum. is there.
[0006]
The present invention is intended to provide a technique capable of removing mechanical noise
generated by a device which has been known in advance and capable of following up the time
change of the mechanical noise.
[0007]
In order to achieve this task, for example, the electronic device of the present invention has the
following configuration.
That is, a relationship between sound wave collecting means for converting sound waves into
electric signals, driving means for performing periodic operations, and a plurality of acoustic data
output from the sound wave collecting means over a plurality of cycles of the drive means
respectively having the same phase. And a combining means for combining the average value of
the sound data output from the calculation means in the same phase as the sound data output
from the sound wave collection means.
[0008]
According to the present invention, the noise waveform is extracted from the device having a
periodic operation such as a rotating drum, even if the noise generated by the operation is mixed
in the microphone. As well as being able to reduce, it becomes possible to follow time change.
[0009]
Hereinafter, embodiments according to the present invention will be described in detail with
reference to the attached drawings.
[0010]
FIG. 1 is a diagram showing a main configuration of an audio recording and reproducing portion
(helical scan type recording and reproducing portion) in a digital video camera apparatus to
which the embodiment is applied.
10-04-2019
3
[0011]
In the figure, 1 is a magnetic tape of a recording medium (a magnetic tape cassette is not shown),
2 is a rotating drum, 3 and 4 are magnetic heads mounted on the rotating drum and recording
audio and video data, 5 is an average value calculation unit 5a. Control microcomputers having L,
6 and 7 are L and R channel microphones, 8 and 9 are microphone amplifiers, 10 is AGC
(automatic level control circuit), 11 and 12 are A / D converters, 13 and 14 are memories , 15,
16 indicate adders.
[0012]
In this configuration, audio signals converted to electric signals by the microphones 6 and 7 are
amplified by the microphone amplifiers 8 and 9, respectively, and are amplified to an appropriate
size by the automatic level control circuit 10, and the A / D converter 11 , 12 are converted to
digital data.
[0013]
A recording signal is recorded on the magnetic tape 1 in a helical scan through the magnetic
heads 3 and 4 attached to the rotating drum 2 rotationally controlled by the microcomputer 5 in
a positional relationship of 180 degrees.
Further, the rotary drum 2 outputs a reference signal indicating the phase of rotation, that is, a
head switching pulse to the control microcomputer 5 to notify the contact state of the magnetic
heads 3 and 4 with the magnetic tape 1 to record the recording signal. Control the flow timing to
4
At this time, noise (mechanical vibration) is generated when the magnetic heads 3 and 4 come in
contact with or separate from the magnetic tape 1 wound around the rotary drum 2 and mixed
with the above-described audio signal to enter the microphones 6 and 7. It will
[0014]
FIG. 2 shows the temporal relationship between the reference signal, the contact timing of the
magnetic heads 3 and 4 and the magnetic tape 1, and the generation of noise.
10-04-2019
4
[0015]
In the figure, the recording signal is applied to the magnetic head 3 when the head switching
pulse is at the high level, and the recording signal is applied to the magnetic head 4 when the
head switching pulse is at the low level.
One cycle of this head switching pulse represents one rotation of the rotary drum 2.
Since each magnetic head needs to be in contact with the magnetic tape 1 in a stable state before
the recording signal is applied, the magnetic head contacts a little before switching of the
recording signal, and the application of the recording signal is completed. It is also in contact for
a very short time.
This is shown in the drawing in which the magnetic head 3 is in contact and the magnetic head 4
are in contact, with the high level indicating contact and the low level indicating non-contact.
Below that, the timing of noise generation occurring at this time is shown.
[0016]
Further, this phenomenon is shown by a noise waveform as shown in FIG. 3, with respect to the
time T of one cycle of the head switching pulse shown at the top in the figure, the continuous
time is T1, T2, T3,. , Tn and one cycle period, and arranged. As can be seen from the figure, it is
found that the noise generated in each rotation is almost the same each time. The relationship
between the operation of the rotary head and the magnetic tape is that there is almost no
changing factor. The lowermost portion shows a waveform obtained by taking the average Ave.
[0017]
On the other hand, FIG. 4 shows the state of an audio signal entering the microphones 6 and 7 at
time T of one cycle of head switching pulse. That is, the continuous speech waveform is divided
10-04-2019
5
into T1 and T2,..., Tn and arranged for each period of time. Since the change of the audio signal
for one cycle is irrelevant to the head switching pulse, the probability that they synchronize is
small. Therefore, increasing the number of times to take the average will approach infinitely zero.
[0018]
Therefore, when the voice data is stored for an appropriate period and data of the same phase in
time T of one cycle of the head switching pulse is added, only the noise component shown in FIG.
3 remains. In the present embodiment, the difference between the characteristics shown in FIGS.
3 and 4 is used.
[0019]
First, as the operation of the embodiment, the above-described operation of extracting the
average value Ave will be described.
[0020]
The outputs from the A / D converters 11 and 12 are input to the average value calculator 5 a of
the control microcomputer 5.
Data are aligned based on the rising edge of the head switching pulse described above as the
digital data sequence input here. Assuming that the names of time intervals of one period of the
head switching pulse are F1, F2, F3,..., The relationship is as shown in FIG. For example, in the
case of a digital video camera, assuming that one cycle of this head switching pulse is 1/150 sec
and the sampling frequency of the A / D converter of the audio signal is 48 kHz, one cycle of the
head switching pulse as shown in FIG. There will be 320 digital data per channel of the
microphone. Based on the rise of the head switching pulse, the first data is t0, t1, t2... T319, and
a heading for each cycle is added, and the data of the first cycle is F1t0, F1t1, F1t2, The data of
the next cycle defined as F1t319 is F2t0, F2t1, F2t2,. For example, in the case of taking an
average of data for a time of 1 sec, data of the final cycle is F149t0, F149t1, F149t2,. If the data
for 150 cycles is considered as a matrix arrangement of the number of data and the number of
cycles for one cycle, it becomes 320 × 150. In order to obtain an average waveform of one cycle
from these data groups, the codes tn indicating the same time (phase) are added together and
divided by the number, so if the code indicating the average value is A, then t0 An average At0 of
10-04-2019
6
time sample values is expressed by the following equation.
[0021]
[0022]
If the number of sample data input successively exceeds 150, that is, the data following the t319
data on the F149th line is again input to the t0 on the F0 line, and the control microcomputer 5
is rewritten The memories 13 and 14 of each channel are controlled.
Thus, the data in the memories 13 and 14 is updated each time data of the same phase is input
from the switching pulse. As described above, when the sampling frequency of data is 48 kHz,
320 occur in one cycle. At each sampling frequency of 48 kHz, the average value of the input
phase data is calculated, and the average value is stored in the memories 13 and 14 of each
channel.
[0023]
In order to realize the above process, the average value calculation unit 5a can be realized, for
example, by the configuration shown in FIG. The figure shows the configuration for calculating
the average value stored in the memory 13, but the configuration for calculating the average
value stored in the memory 14 is the same.
[0024]
In FIG. 11, 110-1 to 110-149 are FIFO memories capable of storing 320 pieces of data from the
A / D converter 11, respectively, and 111 is data from the A / D converter 12 and each FIFO
(149 Is an adder that adds the data from the above (a), and a divider 112 that divides the result
of the addition by "150". Each of these circuits will operate in synchronization with the sampling
frequency (48 kHz in the embodiment).
[0025]
10-04-2019
7
Since each FIFO memory can store 320 data (one cycle of the drum 2), the data output from the
FIFO memory 110-1 is the data from the A / D converter 11 at that time. It will be the data (data
of the same phase) just 1/150 seconds before the data. Further, since the output data of the FIFO
memory 110-1 is input to the FIFO memory 110-2 of the next stage, the data output from the
FIFO memory 110-2 is 1 more than that of the FIFO memory 110-1. It becomes data before /
150 seconds. Since there are 149 FIFO memories as shown, a total of 150 in-phase data
including the data from the A / D converter 11 are supplied to the adder 111. The adder 111
adds the input 150 data and outputs the result to the divider 112. Then, the divider 112 divides
the addition result by the divisor "150". Therefore, the output of the divider 112 is the output of
the value of the equation (1) shown above. Further, since the memories 13 and 14 only need to
be able to substantially hold one piece of data, they can be configured by registers.
[0026]
Next, the operation of noise reduction will be described. The processing for each channel is the
same, so we will only discuss one channel here.
[0027]
The microphone signal from the microphone 6 is a signal mixed with the above-mentioned noise,
and it is digitized by the A / D converter 11. One of the data is input to the adder 15 and is a
positive phase input. In the other adder input, the data from the memory 13 described above is
input in the reverse phase (the sign is minus), and takes the form of subtraction. The data
sequence based on the switching pulse is data successively stored in the memory 13, that is, the
time from an arbitrary switching pulse of an arbitrary cycle Fm (m: integer) of the positive phase
input signal of the adder 15, The average value Atn of noise is to be subtracted from tn (n:
integer), and the output signal OUTtn of the adder 15 is expressed by the following equation:
OUTtn = Fmtn-Atn (2) to remove noise generated by rotation of the drum Will be successful.
[0028]
The attenuation amount of the noise at this time is determined by the rotational accuracy of the
rotary drum 2 and the variation width of the generated noise level. It is assumed that the
10-04-2019
8
rotational unevenness has an accuracy of changing the phase at that frequency by, for example,
5 degrees with respect to the highest frequency of the noise component to be reduced, and it is
assumed that the fluctuation range is 2 dB with respect to the above average value. FIG. 7 shows
the amount of attenuation when two signals having different phases and levels are subtracted.
Under the above conditions, the attenuation of 15 dB or more is obtained from FIG. 7, and
sufficient effects can be obtained. Become.
[0029]
As described above, according to the present embodiment, even if the periodic mechanical noise
emitted by the device is superimposed on the electrical signal obtained through the microphone,
the data for canceling the mechanical noise is stored in memory 13, By storing in 14, it becomes
possible to obtain voice data from which the noise has been removed. In addition, since data
corresponding to the latest periodic mechanical noise emitted by the device is stored as data
stored in the memories 13 and 14, it is also possible to follow the time change and remove the
noise. .
[0030]
Second Embodiment FIG. 8 shows a configuration relating to noise removal in the second
embodiment. In the second embodiment, in addition to the noise reduction processing of the
helical scan type recording / reproducing apparatus using the magnetic tape of the above
embodiment (first embodiment) as a recording medium, the electromagnetic noise emitted from
the capstan is also reduced. The purpose is to
[0031]
Since 1 to 16 in FIG. 8 are the same components as those described in FIG. 1, the description
thereof will be omitted.
[0032]
Reference numeral 17 denotes a pinch roller provided for moving the magnetic tape 1 as a
recording medium, 18 denotes a capstan for giving the transfer of the magnetic tape 1, 19 and
20 denote memories, and 21 and 22 denote adders. .
10-04-2019
9
In this configuration, descriptions will be added to components added to the first embodiment.
[0033]
First, the reference signal from the rotary drum 2 is used as a first reference signal. In one
rotation of the capstan which gives the transfer of the magnetic tape 1, a second reference signal
based on the position of the capstan 18 is output from the capstan 18, and the reference signal is
input to the control microcomputer 5. Each data of the A / D converters 11 and 12 is input to the
data processing unit of the control microcomputer 5 based on the second reference signal. The
digital data sequence input here is aligned data based on the rising of the second reference
signal. For example, assuming that the period of one rotation of the capstan is 3 Hz, 12000
sample values are generated when the sampling frequency is 48 kHz in one period. The names of
the time intervals of one period of the second reference signal are C1, C2, C3..., And from the first
data to t0, t1, t2. If headings for each cycle are added and data of the first cycle is defined as
C1t0, C1t1, C1t2,... C1t11999, data of the next cycle becomes C2t0, C2t1, C2t2,. For example, in
the case of taking an average of data for 12 seconds, the data of the final cycle is C39t0, C39t1,
C39t2, ... C39t11999. When considered as a matrix arrangement of the number of data and the
number of periods for one period, it becomes 12000 × 40. In order to obtain an average
waveform of one cycle from these data groups, the codes tn indicating the same phase time are
added together and divided by the number, so if the code indicating the average value is A, the
time t0 The average Act0 of the sample values is given by the following equation (3)
[0034]
[0035]
Is represented by
If the number of sample data input successively exceeds 40, that is, the data following the
t11999 data on the C39th line is again input to t0 on the C0 line and the control microcomputer
5 rewrites it. , Controls the memories 19 and 20 of the respective channels. Since the
configuration at that time may be substantially the same as that shown in FIG. 11 described
above, the description thereof will be omitted.
10-04-2019
10
[0036]
In this way, 40 pieces of data at the same time are always stored every time from the second
reference signal. As described above, when the sampling frequency of data is 48 kHz, 12000
occur in one cycle. The average value is calculated for each period of the second reference signal,
and the average value of 12000 is stored in the memories 19 and 20 of the respective channels.
[0037]
Next, the operation of noise reduction will be described. Since the processing of each channel is
the same, one channel will be described. The microphone signal from the microphone 6 is a
signal in which the noise generated by the contact between the rotating drum and the magnetic
tape mentioned above and the electromagnetic noise at the time of the rotation of the capstan
mentioned above mix, and it is digitized by the A / D converter 11. It is done. As described in the
first embodiment, the average value of noise synchronized with the rotation of the rotary drum is
stored in the memories 13 and 14, and the adders 15 and 16 remove the noise and add the
adder 21. , 22. In the adders 21 and 22, the output signals of the adders 15 and 16 are positivephase input, and the other adder input is the data from the memories 19 and 20 described above
being input in reverse phase, It takes the form of subtraction. As described above, since the
average value of noise synchronized with the rotation of the capstan is stored in the memories
19 and 20, the average value is subtracted by the adders 21 and 22, and the noise generated by
the capstan is generated here. Is also reduced here. More specifically, assuming that the time
series of the rotating drum noise is DN, the data series based on the switching pulse is DNt0,
DNt1, DNt2, ... DNt319, and one cycle of noise from capstans having different cycles is CN Then,
CNt'0, CNt'1, CNt'2, ... CNt'12000 are obtained. Since tn representing this time and t'm have
different cycles as described above, the final output signal OUTtk at a certain time of this
configuration is expressed by the following equation (4) where Mtk represents an audio signal
string, that is, OUTtk = Mtk + DNtn + CNt'm-Atn-Act'm (4) Since DNtn At Atn CNt'm Act Act'm, the
respective noises are eliminated and OUTtk M Mtk.
[0038]
Third Embodiment In the first embodiment, in order to take an average of 150 pieces of data, it
was necessary to store all the necessary data. Assuming that the sampling frequency is 48 kHz
10-04-2019
11
and the averaging time is 1 second to calculate the average value, the sum of the number of data
to be stored for calculation and the number of data for storing the average value is 1 When M
represents M = 48000 + 320 1 data in 16 bits per channel, the total number TM for two
channels is TM = 2M × 16 = 1546240 bytes, requiring a memory with an unignorable capacity.
[0039]
In the third embodiment, therefore, a technique for reducing such memory capacity is proposed.
FIG. 9 shows a configuration for saving memory capacity in the third embodiment.
[0040]
As shown in FIG. 9, in the case of calculating the average value of k (k is an integer of 1 or more)
data, the data to be input this time has been weighted by 1 / k with respect to the previous
average value. Equivalent to.
[0041]
Therefore, as shown in FIG. 9, the newly input data is first multiplied by 1 / k, and the data from
the memory storing the previous average value is “1-1 / k” (= (k -1) / k) may be weighted and
added, and the memory may be updated with the result.
[0042]
For example, as in the first embodiment, assuming that k is 150 to correspond to the average
value of 150 pieces of data, and the input signal is Idt0, and the signal Mdt0 from the memory,
the calculated average value Adt0 is Adt0 = 1/150 × Idt0 + 149/150 × Mdt0.
Since there are 320 pieces of data for one frame (one drum revolution), it suffices to have 320
pieces of memory.
[0043]
As a result, the total number of memories (for two channels) described above can be realized
with a small configuration of TM = 320 × 2 × 16 = 10240 bits, where one data is 16 bits.
10-04-2019
12
[0044]
According to the above configuration, when k = 150, the correct average value equivalent to the
past 150 frames can not be calculated unless 150 frames of data are input, but this period is
about one second, In addition, taking into consideration the case where it is applied to a digital
video camera, recording and recording are not immediately performed after the power is turned
on, and after confirming that the object is in the field of view, recording and recording
instructions are given. There is no problem above.
[0045]
Other Embodiments In the first to third embodiments, it is determined whether the signal level of
the input signal is a level much higher than the noise data (a level exceeding a preset threshold).
A circuit may be provided, and if the input signal level is sufficiently large, it may be considered
as data of an audio signal other than noise, and control may be added so as to be excluded as a
calculated value of average value calculation.
By this, when calculating the average value of the noise, it is possible to prevent the data of the
average value of only the generated noise from being broken by calculating the data which is not
the noise value higher than the noise value. The noise waveform can be accurately extracted, and
the noise reduction degree is improved.
[0046]
FIG. 10 changes the control method by adding memories 23 and 24 in the paths to the A / D
converters 11 and 12 and the control microcomputer 5 in FIG.
[0047]
In this configuration, when one channel is described, data output through A / D converter 11 is
taken into memory 23.
This memory 23 is fetched for a period of one cycle of the switching pulse, and whether or not
there is data having an absolute value exceeding a preset threshold among the data for one cycle
by the control microcomputer 5 To judge.
10-04-2019
13
If at least one of the data exceeds the threshold, processing is performed such that all data of the
period of the switching pulse is not treated as average calculation data, and all the data is less
than the threshold In some cases, processing is performed as data of average value.
[0048]
This process makes it possible to more accurately determine whether there is a voice louder than
the noise, and destroys the average value data of the noise generated only by calculating the data
that is not the noise level higher than the noise value. Can be prevented, a more accurate noise
waveform can be extracted, and the noise reduction degree is improved.
[0049]
In the embodiment, examples of noises of mechanical noise sources include noise generated
when a head mounted on a drum for recording on a magnetic tape comes into contact with the
tape without contact, and electromagnetic noise during rotation of the capstan. However, the
present invention is not limited by these.
In short, any periodic mechanical noise may be mixed into the microphone as long as it is mixed.
[0050]
As described above, according to the present embodiment, when the recording and reproducing
apparatus has a periodic operation such as a rotating drum, noise generated by the operation
mixes into the microphone, It becomes possible to extract and reduce noise waveforms.
[0051]
In addition, it is possible to extract the noise waveform at each independent cycle of the noise
emitted from the operation unit having periodicity such as rotation of a rotating drum and a
capstan motor for sending a tape, which the recording and reproducing apparatus has, for
example. Thus, it is possible to provide a device having a high noise reduction effect.
[0052]
Furthermore, it is not necessary to individually adjust the noise mixed in the microphone, and
10-04-2019
14
also when the mixed noise varies in size and tone depending on the device, and the same means
is provided even when it changes with time. It is possible to cope with the noise specific to the
device by the same process.
[0053]
Further, as in the prior art, the filter does not remove even the components of the sound
originally recorded, and only the unnecessary noise waveform is extracted, so that the sound
quality can be reduced without any deterioration.
[0054]
It is a block block diagram in 1st Embodiment.
FIG. 8 is a diagram showing the timing of noise generation due to the rotation of the rotating
drum and the contact state of the magnetic head and the magnetic tape in the first embodiment.
It is a figure which shows the waveform of the periodic noise in 1st Embodiment.
It is a figure which shows the asynchronous speech waveform in 1st Embodiment.
It is a figure showing arrangement of data in a 1st embodiment.
It is a figure for demonstrating the process of the data sequence in 1st Embodiment. It is a figure
which shows the reduction amount of the noise in 1st Embodiment. It is a block block diagram in
2nd Embodiment. It is a block block diagram in 3rd Embodiment. It is a block block diagram in
other embodiment. It is a figure which shows the concrete structure of the average value
calculation part in 1st Embodiment.
10-04-2019
15
Документ
Категория
Без категории
Просмотров
0
Размер файла
28 Кб
Теги
description, jp2005159877
1/--страниц
Пожаловаться на содержимое документа