JP2017046234

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 JP2017046234
Abstract: The present invention provides an audio signal extraction device for extracting an audio
signal as a target sound emitted by a user from an object which the user is not aware of or which
the user contacts for another purpose. A sound signal collecting apparatus 10 generates a first
vibration signal kb including a noise vibration component from vibration signals S1 (t) to Sn (t) of
a plurality of piezoelectric elements 1-1 to 1-n. A correlation calculation circuit unit 3 for
selecting a second vibration signal kr including a noise vibration component and a target voice
component, and a first vibration signal kb and a second vibration signal kr selected by the
correlation calculation circuit unit 3 And an audio signal acquisition unit 4 for acquiring an audio
vibration signal S (t). [Selected figure] Figure 1
Audio signal sampling device and program
[0001]
The present invention relates to an audio signal extraction apparatus and program, and can be
applied to, for example, an audio signal extraction apparatus capable of extracting an audio
signal that is a target audio under high noise.
[0002]
It is desirable to extract an audio signal even under high noise such as large noise or noise.
When collecting an audio signal as a target sound using an acoustic microphone under high
04-05-2019
1
noise, the audio signal is buried in a high-power noise signal, which makes it difficult to extract
an audio signal, and the cost is high for noise removal. It takes time.
[0003]
Patent Document 1 refers to a bone conduction microphone (hereinafter referred to as a bone
conduction microphone). A technique for acquiring an audio signal using the In the technology
described in Patent Document 1, an ear pad is attached to the open end side of an ear cup of a
headset, and a bone conduction microphone is provided at a position pressed around the user's
ear. By using the bone conduction microphone, it is possible to collect only the sound signal as
the target sound as the vibration without collecting a large noise signal.
[0004]
JP, 2010-157974, A
[0005]
However, in the conventional technique for collecting an audio signal using a bone conduction
microphone, it is necessary for the user to wear a headset or the like as intended by the bone
conduction microphone mounted headset described in Patent Document 1 There is.
It is necessary for the user to intentionally wear a headset, etc. which the user does not normally
wear, which is troublesome for the user.
[0006]
Therefore, according to the present invention, the user is provided with a plurality of
piezoelectric means (piezoelectric elements) constituting the bone conduction microphone on an
object which the user is not aware of or which the user contacts for another purpose. It is an
object of the present invention to provide an audio signal extraction apparatus and program
capable of extracting an audio signal as a target sound.
[0007]
In order to solve the problems as described above, the audio signal collecting apparatus
04-05-2019
2
according to the first aspect of the present invention extracts a voice vibration signal which is a
target voice from vibration signals from a plurality of piezoelectric means for converting
vibration into an electric signal. A sound signal sampling apparatus comprising: (1) a first
vibration signal including a noise vibration component and a second vibration signal including a
noise vibration component and a target sound component from vibration signals of a plurality of
piezoelectric means And (2) voice vibration signal collecting means for collecting a voice
vibration signal based on the first vibration signal and the second vibration signal selected by the
signal selecting means. Do.
[0008]
An audio signal extraction program according to a second aspect of the present invention is an
audio signal extraction program for extracting an audio vibration signal as a target audio from
vibration signals from a plurality of piezoelectric means for converting vibration into an electric
signal, and (1) Signal selection means for selecting a first vibration signal including a noise
vibration component and a second vibration signal including a noise vibration component and a
target voice component from vibration signals of a plurality of piezoelectric devices ( 2) The
apparatus is characterized in that it functions as voice vibration signal collecting means for
collecting a voice vibration signal based on the first vibration signal and the second vibration
signal selected by the signal selection means.
[0009]
According to the present invention, the user is provided with a plurality of piezoelectric means
(piezoelectric elements) constituting the bone conduction microphone on an object which the
user is not aware of or which the user contacts for another purpose. An audio signal to be a
target sound can be sampled.
[0010]
It is a block diagram showing functional composition of an audio signal extraction device
concerning a 1st embodiment.
It is an explanatory view explaining installation of a piezoelectric element concerning a 1st
embodiment.
It is an explanatory view explaining installation of a piezoelectric element concerning a 2nd
04-05-2019
3
embodiment.
It is a block diagram which shows the functional structure of the audio ¦ voice signal extraction
apparatus which concerns on 2nd Embodiment.
[0011]
(A) First Embodiment Hereinafter, a first embodiment of an audio signal sampling device and
program according to the present invention will be described in detail with reference to the
drawings.
[0012]
(A-1) Configuration of First Embodiment FIG. 1 is a block diagram showing a functional
configuration of an audio signal collecting apparatus according to the first embodiment.
[0013]
The audio signal collecting apparatus 10 according to the first embodiment includes a plurality
of piezoelectric elements constituting a bone conduction microphone, a dedicated IC chip, and
the like as a hardware configuration.
The process for realizing the function of the voice signal extraction device 10 may be realized by,
for example, a hardware configuration, or may be configured as software as a CPU and a
program executed by the CPU, but functional Can be represented in FIG.
[0014]
As shown in FIG. 1, the audio signal sampling apparatus 10 according to the first embodiment
includes piezoelectric elements 1-1 to 1-n (n is an integer), an analog / digital (A / D) converter 2,
correlation calculation. A circuit unit 3 and an audio signal sampling unit 4 are provided.
[0015]
The piezoelectric elements 1-1 to 1-n are piezoelectric elements constituting a bone conduction
04-05-2019
4
microphone, and an object or a target device with which a user contacts (hereinafter simply
referred to as an "object").
Provided).
It is desirable that the objects provided with the piezoelectric elements 1-1 to 1-n be tangible
objects such as objects or devices with which the user unintentionally contacts when the user
emits voice.
All the piezoelectric elements 1-1 to 1-n do not have to be provided on the part where the user
touching the object necessarily comes in contact, but are provided on the part where the user
seems to be involuntarily touching. The piezoelectric elements 1-1 to 1-n are provided in contact
with the user's body. The piezoelectric elements 1-1 to 1-n are referred to as a user's body
vibration signal ( voice vibration signal ) generated by emitting a voice. ) Containing the
vibration and convert it into an electrical signal. The piezoelectric elements 1-1 to 1-n also
capture the vibration of the object provided with each of the piezoelectric elements 1-1 to 1-n,
and convert the vibration conducted from the object to an electrical signal.
[0016]
FIG. 2 is an explanatory view for explaining the installation of the piezoelectric elements 1-1 to 1n according to the first embodiment. For example, in the first embodiment, a steering wheel of a
car (hereinafter referred to as a steering wheel). 2 illustrates the case where a plurality of (13 in
FIG. 2) piezoelectric elements 1-1 to 1-13 are provided.
[0017]
As shown in FIG. 2, piezoelectric elements 1-1 to 1-13 are provided at a portion of a circular
steering wheel 5 of an automobile where a user who steers the automobile grips the steering
wheel 5. As the parts held by the user may be various parts depending on the aspect of the user,
it is desirable that the piezoelectric elements 1-1 to 1-n be arranged in many parts in
consideration of them.
04-05-2019
5
[0018]
In addition, although the case where the piezoelectric elements 1-1 to 1-13 are provided on the
front side (driver side) of the steering wheel 5 of FIG. 2 is illustrated, a portion where the driver's
hand may come into contact with the steering wheel 5 In this case, it may be provided on the
back side of the steering wheel 5 (the side not facing the driver), or may be provided on both the
front side and the rear side of the steering wheel 5. The number of piezoelectric elements 1-1 to
1-13 is not particularly limited.
[0019]
In FIG. 2, in order to capture the vibration of a traveling automobile, the piezoelectric element 11 is provided in a portion where the driver's body to be steered is unlikely to contact, for
example, a T-shaped portion of the steering wheel 5 in FIG. It is like that. In this case, the
piezoelectric element 1-1 functions as a noise vibration capturing piezoelectric means for
capturing noise vibration transmitted from the handle 5.
[0020]
The A / D conversion unit 2 converts analog signals of the plurality of piezoelectric elements 1-1
to 1-n into digital signals S1 (t) to Sn (t). t is a parameter representing time. An amplifier for
amplifying signals from the piezoelectric elements 1-1 to 1-n, or a filter for removing noise
components if noise components can be predicted in advance may be provided. .
[0021]
The correlation calculation circuit unit 3 calculates the correlation between the signals S1 (t) to
Sn (t) based on the frequency characteristics of the respective signals of the piezoelectric
elements 1-1 to 1-n. That is, the correlation calculation circuit unit 3 derives a signal including a
noise vibration component conducted from the steering wheel 5 of a traveling automobile from
the vibration signals from the piezoelectric elements 1-1 to 1-n, and the noise vibration
component and the voice. To derive a signal that includes the user's voice vibration signal.
04-05-2019
6
[0022]
As shown in FIG. 1, the correlation calculation circuit unit 3 includes an FFT unit 31 and a signal
selection unit 32.
[0023]
The FFT unit 31 uses the piezoelectric elements 1-1 to 1-n to analyze frequency characteristics
(frequency information) of the signals S1 (t) to Sn (t) sampled by the piezoelectric elements 1-1
to 1-n. The signals S1 (t) to Sn (t) thus acquired are converted into frequency information
(frequency characteristics) S1 (k) to Sn (k) in frame units.
Here, k represents a frame. In addition, the conversion method to the frequency information of a
signal is not limited to Fourier transform (FFT).
[0024]
The signal selection unit 32 uses the piezoelectric element 1-1 provided in advance at a portion
where the user's body is unlikely to come in contact, in order to select a signal including noise
and vibration signals that may occur with a traveling vehicle. The output frequency information
S1 (k) is output as a signal Sb (k).
[0025]
Further, the signal selection unit 32 selects the vibration signal on which the noise vibration
signal and the voice vibration signal of the user emitting the voice are superimposed, the signal
S1 (t) of the piezoelectric element 1-1, and the piezoelectric element 1-. The correlation value
with each of the signals S2 (t) to Sn (t) sampled by 2-1-n is calculated.
At this time, the signal selection unit 32 selects the signal S1 (t) that is the most different from
the signal S1 (k). That is, the signal selection unit 32 uses the frame unit frequency information
S1 (k) to Sn (k) of each signal converted by the FFT unit 31 and has the lowest correlation value
with the signal S1 (k). Select one and output it as a signal Sr (k).
[0026]
04-05-2019
7
The signal selection unit 32 regards and outputs the signal S1 (k) captured by the piezoelectric
element 1-1 as a noise vibration signal that captures the vibration generated by the traveling of
the vehicle. The signal selection unit 32 adds the signal with the lowest correlation value to the
noise vibration signal of the automobile based on the correlation results of the signal S1 (k) and
the signals S2 (k) to Sn (k). A signal Sr (k) is output considering that it is a vibration signal on
which the voice vibration signal of the user who emits the voice is superimposed.
[0027]
This means that the piezoelectric element 1-1 which is not in contact with the user's body will
capture the vibration signal of the vibration of the traveling car being conducted through the
handle 5. On the other hand, among the piezoelectric elements 1-2 to 1-n, those which may come
in contact with the user's body emitting the voice are superimposed on the user's voice vibration
signal in addition to the vibration of the traveling automobile. To capture the vibration signal.
Therefore, the signal selection unit 32 selects one of the correlation values of the signal S1 (k)
and the signals S2 (k) to Sn (k) that has the lowest correlation value, the noise vibration signal
and the voice vibration signal of the automobile. Output as including.
[0028]
The audio signal sampling unit 4 calculates the difference between the signal Sb (k) output from
the correlation calculation circuit unit 3 and the signal Sr (k), and outputs a signal S (t). The audio
signal sampling unit 4 may calculate the difference between the signal Sb (k) and the signal Sr (k)
after inverse conversion to the time domain and then calculate in the frequency domain.
[0029]
(A-2) Operation of the First Embodiment Next, the operation of the audio signal sampling
apparatus 10 according to the first embodiment will be described with reference to the abovedescribed drawings.
[0030]
04-05-2019
8
Each of the piezoelectric elements 1-1 to 1-13 constituting the bone conduction microphone
according to the first embodiment is provided on a steering wheel 5 of a car.
When the user who steers the vehicle grips the steering wheel 5, any one of the piezoelectric
elements 1-1 to 1-13 can contact via the body of the user.
[0031]
Among the piezoelectric elements 1-1 to 1-13, the piezoelectric elements 1-1 to 1-13 in contact
with the user's body when the user utters a voice are vibration signals of the traveling
automobile and the user's It captures audio vibration signals conducted through the body and
converts them into electrical signals.
[0032]
On the other hand, among the piezoelectric elements 1-1 to 1-13, the piezoelectric elements 1-1
to 1-13 which are not in contact with the user's body capture vibration signals obtained from the
car body of a traveling automobile, and the electric signals Convert to
[0033]
The signals output from the respective piezoelectric elements 1-1 to 1-13 are converted into
digital signals by the A / D conversion unit 2 and supplied to the correlation calculation circuit
unit 3.
[0034]
In the correlation calculation circuit unit 3, the signals obtained by the piezoelectric elements 1-1
to 1-13 are Fourier-transformed by the FFT unit 31 to obtain frequency information S1 (k) to S13
(k).
The frequency information S1 (k) to S13 (k) is provided to the signal selection unit 32.
[0035]
In the signal selection unit 32, among the frequency information S1 (k) to S13 (n), the signal S1
(k) having a low possibility of contact with the user's body is used to select a signal including a
04-05-2019
9
vibration signal of the automobile. The selected signal Sb (k) is output to the audio signal
sampling unit 4.
[0036]
Further, in the signal selection unit 32, the correlation value between the signal S1 (k) and each
of the other signals S2 (k) to S13 (k) is calculated.
Then, from the correlation value between the signal S1 (k) and each of the signals S2 (k) to S13
(k), in order to select the signal including the vibration signal of the automobile and the voice
vibration signal of the user who emits voice. The signal with the lowest correlation value is
selected, and the signal Sr (k) is output to the audio signal sampling unit 4.
[0037]
The signal Sb (k) including the vibration signal of the vehicle and the signal Sr (k) including the
vibration signal of the user who issued the vibration signal of the vehicle and the voice output
from the correlation calculation circuit unit 3 are voice signals The sampling unit 4 calculates the
difference.
As a result, the vibration signal of the automobile is removed, and a signal S (t) including the
voice vibration signal of the user is obtained and output.
[0038]
(A-3) Effects of the First Embodiment As described above, according to the first embodiment, a
plurality of bone conduction microphones are configured on an object or a target device with
which a user who emits a sound comes into contact unintentionally Voice vibration from the
difference between a signal including an audio vibration signal collected by the piezoelectric
device in contact with the user's body and a signal including the vibration signal collected by the
non-contact piezoelectric device. A signal can be collected.
04-05-2019
10
As a result, the user can eliminate the influence of the noise signal under high noise without the
user's intention of wearing the bone conduction microphone, and the speech signal can be
collected by the bone conduction microphone.
[0039]
(B) Second Embodiment Next, a second embodiment of the audio signal sampling device and
program according to the present invention will be described in detail with reference to the
drawings.
[0040]
The second embodiment is an embodiment in which the object on which the piezoelectric
elements 1-1 to 1-n constituting the bone conduction microphone are installed is provided on the
backrest 5A of the chair of the car instead of the handle 5 of the car. It is.
[0041]
FIG. 3 is an explanatory view for explaining the installation of the piezoelectric elements 1-1 to 1n according to the second embodiment.
For example, in the second embodiment, a case where a plurality of (nine in FIG. 3) piezoelectric
elements 1-1 to 1-9 are provided on the backrest 5A of a chair in a car is illustrated.
As shown in FIG. 3, piezoelectric elements 1-1 to 1-9 are provided at the backrest 5A of the chair
where the back of the driver who steers the vehicle contacts.
[0042]
The piezoelectric elements 1-1 to 1-9 are a seat or an armrest of a chair, and the piezoelectric
element 1--1 is a portion which can be in contact with the user sitting on the seat or the user's
body placed on the armrest. It may be provided in 1 to 1-9. The number of piezoelectric elements
1-n is not particularly limited.
04-05-2019
11
[0043]
FIG. 4 is a block diagram showing a functional configuration of the audio signal sampling device
according to the second embodiment.
[0044]
An audio signal extraction apparatus 10A according to the second embodiment includes
piezoelectric elements 1-1 to 1-n (n is an integer), an analog / digital (A / D) converter 2, a
correlation calculation circuit 3A, an audio signal output It has a part 4.
[0045]
In the second embodiment, the processing of the correlation calculation circuit unit 3A is
different from that of the first embodiment, and the other piezoelectric elements 1-1 to 1-n (n is
an integer), analog / digital (A / D) The processing of the conversion unit 2 and the audio signal
collection unit 4 are the same as or correspond to those of the first embodiment.
Therefore, in the second embodiment, the processing of the correlation calculation circuit unit 3A
will be mainly described in detail.
[0046]
The correlation calculation circuit unit 3A includes an FFT unit 31 and a signal selection unit
32A.
[0047]
As in the first embodiment, the FFT unit 31 performs frequency information (frequency
characteristics) S1 (k) on a frame basis of the signals S1 (t) to Sn (t) sampled by the piezoelectric
elements 1-1 to 1-n. ) To Sn (k).
[0048]
The signal selection unit 32A outputs a signal that most approximates the noise vibration
characteristic calculated from the signals sampled by the piezoelectric elements 1-1 to 1-9 to the
audio signal sampling unit 4 as a signal Sb (k).
04-05-2019
12
[0049]
Further, the signal selection unit 32A selects the vibration signal of the traveling automobile and
the vibration signal of the user emitting the voice, and selects the vibration signal on which the
signal S1 is sampled by the piezoelectric elements 1-1 to 1-9. The correlation between (t) and S9
(t) is taken, the signal Sr (t) is selected based on each correlation result, and the signal Sr (t) is
output to the audio signal sampling unit 4.
[0050]
Here, in the second embodiment, the method of selecting the signal Sb (t) and the signal Sr (k) by
the signal selection unit 32A is different from that in the first embodiment.
[0051]
In the first embodiment, a signal from a piezoelectric element provided at a position where
contact with the user's body is unlikely is output as Sb (k), and the correlation between this signal
Sb (k) and other signals is output. The case where the one with the lowest value is selected is
illustrated.
[0052]
On the other hand, in the second embodiment, the signal selection unit 32A obtains the signal S1
(taken by the piezoelectric elements 1-1 to 1-9 provided at the position where the signal
selection unit 32A may come in contact with the user's body). k) Noise vibration characteristics
are calculated from S9 (k), and a signal most approximate to the noise vibration characteristics is
set as a signal Sb (k), and a signal farthest from the noise vibration characteristics is output as Sr
(k).
[0053]
For example, in order to calculate all the correlation values between two signals, for example, the
signal selection unit 32A uses piezoelectric elements 1-1 to 1-9 provided at positions that may
come in contact with the user's body. Of the sampled signals S1 (k) to S9 (k), a combination of
two signals is selected, and a correlation value between these two signals is calculated.
At this time, the signal selection unit 32 calculates a correlation value between two signals for all
combinations.
04-05-2019
13
[0054]
The signal selection unit 32 calculates, for each of all the signals S1 (k) to S9 (k), a correlation
average value which is an average value of correlation values of combinations involving itself, in
order to calculate the noise and vibration characteristics. .
Furthermore, the signal selection unit 32 obtains an overall correlation average value using each
correlation average value of the signals S1 (k) to S9 (k).
The overall correlation average value is an average value of all the correlation average values in
which each of the signals S1 (k) to S9 (k) is involved, and strongly includes the influence of noise
and vibration components.
[0055]
Therefore, the signal selection unit 32 selects the correlation average value closest to the overall
correlation average value (the correlation average value having the smallest difference with the
overall correlation average value) among the correlation average values of the signals S1 (k) to
S9 (k). Signal Sb (k), and the signal farthest from the overall correlation average value (having the
largest correlation average value with the overall correlation average value) is output as Sr (k).
[0056]
Thus, even when the piezoelectric element 1-1 is not provided at a position where the
contactability of the user's body is low, the signal Sb (k) including the vibration signal and the
signal Sr including the vibration signal and the voice vibration signal (K) can be output.
Note that the method of selecting the signal Sb (k) and the signal Sr (k) by the signal selection
unit 32 is not limited to the above modification.
[0057]
As described above, according to the second embodiment, in addition to the effects described in
the first embodiment, even when the piezoelectric element is not provided in a portion where the
04-05-2019
14
user is unlikely to contact, the user can not recognize the bone It is possible to eliminate the
influence of the noise signal under high noise without intending to attach the conductive
microphone, and to obtain the audio signal by the bone conductive microphone.
[0058]
(C) Other Embodiments Although the first and second embodiments described above also refer to
various modified embodiments of the present invention, the present invention can also be
applied to the following modified embodiments.
[0059]
(C-1) In the first and second embodiments described above, the case where the object provided
with the plurality of piezoelectric elements is the handle of a car or the backrest of a chair is
exemplified, but the object provided with the piezoelectric elements is the handle And not limited
to chairs.
[0060]
Generally, the inside of a traveling automobile is an environment where loud noise and noise may
occur.
Therefore, in the first and second embodiments, an example in a car as an example under high
noise is illustrated, and as another example in which a user who emits speech unknowingly
touches, another purpose when speaking He cited the handlebars and backrests of chairs that
come in contact with the car (steering the car).
[0061]
For example, there are public facilities such as a construction site, an airport, a station yard, and
the like in an environment where large noise or noise may occur.
Therefore, the piezoelectric element may be provided at a site where a user who speaks
unintentionally contacts at a construction site, a public facility or the like.
04-05-2019
15
In the embodiment described above, the piezoelectric element is provided at a position in contact
with the hand, back, etc. from the viewpoint of contact with the user's body. An element may be
provided.
[0062]
(C-2) In the first and second embodiments described above, the signal selection unit correlates
the signal including the noise vibration component with the signal including the noise vibration
component and the voice vibration component by two signals. The case of selecting by
The signal selection unit may apply a band pass filter for extracting a frequency band including
an audio component according to the frequency characteristic of the audio component captured
by the piezoelectric element constituting the bone conduction microphone.
In addition, the signal component of the frequency band that is considered to be a voice
component may be multiplied by a gain.
[0063]
(C-3) In the first and second embodiments described above, the audio signal sampling unit 4
multiplies the difference signal between the signal Sb (k) and the signal Sr (k) by the gain and
outputs the product, or a band pass filter To extract the audio component.
[0064]
10 and 10A: voice signal collecting apparatus, 1-1 to 1-n: piezoelectric element, 2 .. A / D
conversion unit, 3: correlation calculation circuit unit, 31: FFT unit, 32: signal selection unit, 4:
voice signal Collection department.
04-05-2019
16

Download Report