close

Вход

Забыли?

вход по аккаунту

?

JPH0435300

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 JPH0435300
[0001]
The present invention relates to a sound processing apparatus for processing a signal received
from a microphone array in a sound field where unnecessary sound sources such as noise and
echo are present with a filter to extract a desired sound signal. “Conventional technology J: In
the case of receiving a desired acoustic signal in the presence of an unnecessary acoustic signal,
a directional microphone is conventionally used. However, since the directivity characteristic is
fixed, depending on the positional relationship between the desired acoustic signal source and
the unnecessary acoustic signal source, the blind spot of the directional microphone can not
always be directed to the incoming direction of the noise. Noise (... There is a fundamental
problem that it is difficult to suppress noise if the positive amount is not sufficient and noise
comes from multiple directions. On the other hand, there is a method called adaptive microphone
array as a method using signal processing. This is a method of using a plurality of microphones
and digital filters to form a dead spot of directivity in the direction of arrival of unnecessary
acoustic signals while maintaining sensitivity in a target direction, and receiving only a desired
acoustic signal. is there. In the method using this adaptive microphone array, it is necessary to
pick up the unnecessary acoustic signals such as noise and echo while they are coming, and learn
the direction of arrival of the unnecessary acoustic signals from the picked up data. There is.
Therefore, we are learning in advance. However, not all noise or echo occurs during learning.
Rather, it is rare that all noises or echoes occur at the same time, often with a time lag. For this
reason, after forming a directivity pattern having a dead angle in the direction of arrival of noise
at the time of learning, in order to suppress newly generated noise, it is necessary to re-create
the directivity pattern. Therefore, only the noise and the echo have to be picked up again,
meanwhile the desired acoustic signal has to be interrupted, which is inefficient, and in the worst
case, the desired acoustic signal is suppressed as noise when it can not be interrupted. There was
a problem of saying that In addition, it is difficult to capture the state of only an unnecessary
08-05-2019
1
acoustic signal, and when learning is performed in a state in which a desired acoustic signal is
simultaneously received, there is a problem that the arrival direction of the desired acoustic
signal also becomes a blind spot. In particular, with regard to printer sounds in the office,
telephones, private languages in seats, etc., the sound source position is clear, but it is less likely
to occur simultaneously, which is a big problem in this case. An object of the present invention is
to extract a desired acoustic signal without learning on the way if the change in the state of the
sound field is understood under the sound field where unnecessary acoustic signals are present,
and the change is not generated frequently. An object of the present invention is to provide a
sound reception processing device capable of receiving sound.
[Means for Solving the Problems] According to the present invention, a plurality of sound field
states of unnecessary sound sources predicted in the use environment are stored in advance in
the memory unit, and one or more sound field states are stored from the memory unit. Is selected
and the target sound field state is generated by the sound field state generator, and the filter
coefficient is calculated by the filter coefficient calculator based on the generated sound field
state, and the filter coefficient is calculated by the microphone array The desired sound signal is
extracted by setting it as a filter for processing the received sound output. A conference
apparatus can be considered as an example of actually using the sound reception processing
apparatus of the present invention. In fact, in environments such as offices where the use of
conferencing equipment is expected, the location is often constant even though there are
multiple noise sources. That is, the installation position and the seat of the telephone are almost
fixed, and the speaker for communication is also fixed. Therefore, even if there are a plurality of
noise and echo arrival directions, the directions are constant to a certain extent and can be
predicted. Moreover, when a person is considered to be a noise source depending on whether or
not a person is sitting for each seat, the condition of the unnecessary sound field is different, and
it is assumed that a certain speaker is used. The generation state of the echo is different when
not present, and a plurality of sound field states due to these unnecessary sound sources are
predicted, and this is stored in advance in the memory unit. -Since the sound field can be
considered as a linear system in JQ, the superposition theory holds. Therefore, the sum of the
data obtained by converting the sound collected when the echo is coming into digital signal
information and the data collected when the noise is coming is collected when the echo and the
noise come simultaneously It is equivalent to the Therefore, according to the attendance
condition of the meeting of the day, that is, the seating condition of the seat, the speaker used,
etc. in the conference apparatus, one or a plurality of landing locations are selected from the
memory unit, Create a field condition to remove unwanted acoustic signals. Embodiment FIG. 1
shows an embodiment of the present invention. Each microphone output from the microphone
array 10 consisting of a small number (or at least two microphones l) for right-handing is
supplied to the filter 11 for processing and the desired acoustic signal is taken out and output to
the output terminal 12. According to the present invention, a plurality of sound field states due to
unnecessary sound sources are stored in the memory unit I3 and one or more sound field states
08-05-2019
2
are selected from the memory unit 13 according to the setting for the sound field setting unit 14
The target sound field state is generated by the sound field state generation unit 15, and the
target sound field state is supplied to the filter coefficient calculation unit 16, the optimum filter
coefficient is calculated, and the filter coefficient is the filter 11 Set to
For example, unnecessary sound sources such as echoes and noises in the use environment are
separately output, the unnecessary acoustic signals are received by the microphone array 10,
and the outputs of the respective microphones l are manually input to the memory unit 13 to
obtain digital signals. The sound field state of each unnecessary sound source is stored in the
memory unit 13 in this manner. By operating the sound field setting unit 14 in response to the
unnecessary sound source predicted according to the use condition, the sound condition of the
sound field state in the memory unit 13 may be all or one or more arbitrary sound field
conditions. By combining in the field state generation unit 15, a target sound field state
equivalent to the sound field state actually collected from the sound field is generated. From the
sound field state obtained from the sound field state generation unit 15, the filter coefficient
calculation unit 16 calculates an optimum filter coefficient, and the optimum filter coefficient is
centered on the filter 11, and the microphone array 10 makes the target sound For example, in a
general room with a volume of 80 rri and a reverberation time of 0.4 seconds, communication
speakers as shown in Fig. 2 that form a directional dead angle in the direction in which the
arrival of noise or echo in a field condition is expected. When stereo reproduction is performed
by SPI SF3 and it is desired to receive voices uttered by the speaker S who listens to the
reproduction by the microphone array 10, the non-stationary noise source SP3 is assumed to be
at the positions shown in the figure. In order to suppress howling and echo, the communication
speaker SP1. A dead angle must be formed in the directivity of the microphone array 10 in the
direction of SP2. In order to simulate the echo coming from the communication speaker SPI, SP2,
a signal (white noise or pseudo sound) is generated in advance from the communication speaker
SPI, SF3 and collected by the microphone array 10, The sound field state is stored in the memory
unit as digital signal information. Let that data be sound field state l. At the same time, when nonstationary noise is coming from the noise source SP3, the noise is collected by the microphone
array 10, and the sound field state is stored in another area of the memory unit. The data is
referred to as sound field state 2. As mentioned earlier,-in fishing, the sound field can be
considered as a linear system, so the theory of superposition Z holds. Therefore, the state in
which the echoes come from the communication speakers SPI and SP2 and the noise comes from
the noise 11sP3 is equivalent to the sum of the sound field state 1 and the sound field shape B2.
The directional characteristics of the filter coefficient obtained from the sound field shape B1 are
shown in FIG. 3, and the directional characteristics of the filter coefficient obtained from the
sound field state 2 are shown in FIG.
It can be seen from FIG. 3 that although the dead angle of directivity is formed in the direction of
08-05-2019
3
the communication speakers SPI and SF3, the dead angle of directivity is not formed in the
direction of the noise tASP3. In FIG. 4, it can be seen that although the dead angle of directivity is
formed in the direction of the noise source SP3, the dead angle is not formed in the direction of
the communication speakers SPI, SP2. Now, both digital signals of sound field state 1 and sound
field state 2 are added, and the sum is made sound field state 3. The directivity characteristic of
the filter coefficient obtained from this sound field shape B3 is shown in FIG. It can be seen from
FIG. 5 that with the filter coefficient obtained from the sound field state 3, a dead angle is formed
in all directions of the communication speaker 5PISP2 and the noise source SP3. FIG. 6 shows
the frequency response in each direction of the speaker S of the filter 11, the communication
speaker SPI and SP2, and the noise source SP3 obtained from the sound field state 3. It can be
seen that the response in the incoming direction of noise or echo is suppressed over the entire
band with respect to the response in the 11th signal direction, that is, the speaker S direction.
The average of all bands is about 18.3 dll for the communication speaker SPI, about 18 dB for
the communication speaker SP2, and about 19 dB for the lf'lt sound source SP3. Therefore, when
the noise 1 sP3 does not exist, the base station state 1 is selected to be supplied to the filter
coefficient calculator 16 by IRL, and when the echo from the communication speakers SPI and
SP2 does not exist, the sound field state 2 is selected and the filter is selected. The filter
coefficient calculator 16 is supplied to the coefficient calculator 16 and when the echo of the
communication speaker SPI, the echo of the SP2, and the noise of the noise source SP5 are both
present, the sound field B3 obtained by adding the sound field state 1 and the sound field state 2
If supplied to the system, it is possible to suppress noise (non-stationary noise) and echo which
are not necessarily present simultaneously. In this example, when the three states are known in
advance, the sound field setting unit 14 is manually selected to select the most appropriate
sound field state and supplied to the filter coefficient calculation unit 16. The arrival direction of
noise is not limited to one. When there is a plurality of noises that are expected to arrive, each
sound field state by them may be stored in advance as data, and they may be selectively
combined according to the situation. In that case, the amount of suppression in each direction
can also be controlled by weighting and adding a plurality of necessary sound field conditions.
For example, as shown in FIG. 7, it is assumed that six noises 6spx to SP6 exist. First, driving
noise from the speaker at the position of each of the noise sources SPI to SP6 is performed for
each of the noise sources in advance to pick up each sound field state.
Then the noise fisPl. When suppressing noise of only SF3, a sound field state (target sound field
state) of the sum of the sound field states of the noise sources SP1 and SF3 is created. The
directivity characteristic by the filter coefficient obtained from the sound field state data of this
sum is shown in FIG. As shown, directional dead angles are formed in the directions of the noise
sources SPI and SP5, respectively, and noise can be suppressed. Likewise, noise! SP2.
SP4. The directional characteristics obtained from the sound field state data of the sum of
the respective sound field states of SP6 added at the ratio of weight 1; 1: 1 are shown in FIG.
From the figure, noise source SP2. SP4. It can be seen that directional blind spots are formed
08-05-2019
4
in each direction of SP6. Noise source S)) 2 □ 3P4. FIG. 10 shows an example of directivity
characteristics obtained from sound field state data in which each sound field state of SF3 is
added at a weight ratio of 1: 5. By making the weight larger than those in FIGS. 9 and 10, the
amount of suppression in the direction of noise #sp, + can be made sufficiently large. The amount
of suppression in each direction can be controlled by weighting the sound field conditions only
when the noise that is in the direction that follows it, or when some noises are large or small. The
directivity characteristics obtained from the sound field state data of the sum obtained by adding
the sound field states of noises aspx to SP6 at the same rate are shown in FIG. (Effect J of the
Invention As described above, according to the present invention, each separate sound field state
for each unnecessary sound source is stored in advance (by, for example, the unnecessary
predicted by the day, the time, etc. By creating from the sound field state where the sound field
state by the sound source is stored, it is possible to suppress noise and echo existing in the use
environment and extract and receive one desired sound signal, regardless of the timing of
learning. .
[0002]
Brief description of the drawings
[0003]
1 is a block diagram showing an embodiment of the present invention, FIG. 2 is an example of a
positional relationship between a J11 trusted speaker and a microphone array 941 sound source
and a speaker, and FIG. 3 is a communication speaker in FIG. The directivity characteristic
diagram determined by receiving only the echo from SPI and SF3; FIG. 4 is the directivity
characteristic diagram determined by receiving only the noise from the noise source SP3 in FIG.
2; Directivity characteristics figure determined from the sum of echo and noise separately
received, Fig. 6 is a frequency response characteristic diagram in each direction, Fig. 7 is a
microphone array, and other examples of positional relationship between noise source and
speaker FIG. 8 shows directivity characteristics of noise sources SP1 and SF3. FIG. 9 shows noise
sources SP2.
SP4. FIG. 10 is a noise source SP2. SP4. FIG. 1] is a directivity characteristic diagram in
the case of noise sources SPI to SP6 when weighting is performed on the sound field state of SPG.
Patent Applicant Hiki Telegraph & Telephone Corporation
08-05-2019
5
Документ
Категория
Без категории
Просмотров
0
Размер файла
15 Кб
Теги
jph0435300
1/--страниц
Пожаловаться на содержимое документа