close

Вход

Забыли?

вход по аккаунту

?

JP2007096389

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 JP2007096389
An object of the present invention is to provide a regressive sound removing device that stably
operates an adaptive filter in a short time even when the acoustic environment changes rapidly
and nonlinearly. A control unit 7 supplies sound environment instruction data to a sound
collection directivity control unit 5 and an adaptive filter 11, and the sound collection directivity
control unit 5 accordingly collects a sound collection signal having a predetermined sound
collection directivity. Generate The adaptive filter 11 detects sound collection directivity from the
sound environment instruction data, and reads out from the memory 13 a filter parameter
corresponding to the sound collection directivity. The adaptive filter 11 sets delay coefficients
and filter coefficients of the FIR filter, and generates a pseudo echo signal by an impulse
response to the received speech signal. The adaptive filter 11 sets a more optimal filter
parameter based on the error signal obtained by subtracting the pseudo echo signal from the
sound collection signal in the adder 12 and generates the next pseudo echo signal. [Selected
figure] Figure 1
Regression sound removal device
[0001]
The present invention relates to a regression sound removing device for preventing acoustic echo
and howling caused by sound emitted from a speaker coming around a microphone and being
collected, and more particularly to a regression sound removing device using an adaptive filter. is
there.
[0002]
Conventionally, various devices using adaptive filters to prevent acoustic echo and howling have
04-05-2019
1
been disclosed.
[0003]
The echo canceler of Patent Document 1 includes a plurality of microphones, and the transfer
function of the transmission path from each microphone is updated and set by the error signal
after echo removal, and the filter function of the FIR filter (adaptive filter) is performed by this
transfer function. Set
[0004]
The echo canceller device of Patent Document 2 includes a plurality of microphones, and
calculates pseudo echo path characteristics of each transmission path (echo path) from the
plurality of total pseudo echo path characteristics estimated in the past and the transfer function
at this time, A new comprehensive pseudo echo path characteristic is set from this pseudo echo
path characteristic and the transfer function of this time.
Japanese Patent Application Laid-Open No. 62-120734 Patent No. 2938076
[0005]
However, in Patent Document 1, since the filter coefficient of the adaptive filter is set using the
error signal, it is necessary to keep updating the adaptive filter until the error signal converges,
and it takes time to set the filter coefficient.
Further, in Patent Document 2, the current integrated pseudo echo path characteristic is
calculated by performing matrix operation using the previous integrated pseudo echo path
characteristic, the transfer function, and the current transfer function. And complicated
arithmetic processing is required.
[0006]
In particular, in recent years, in a sound system using a speaker array in which a plurality of
04-05-2019
2
speakers are arrayed and a microphone array in which a plurality of microphones are arrayed,
the directivity of the speaker array and the microphone array is controlled to control sound.
Environments are often changed rapidly and nonlinearly.
In such a situation, as in the above-mentioned patent documents, in the method of setting the
current filter coefficient based on the previous error signal or the filter setting, the setting of the
filter coefficient may follow the change of the acoustic environment. It can not be done, and it
takes a long time for the adaptive filter to operate stably.
[0007]
Therefore, an object of the present invention is to provide a regressive sound removing device
that operates the adaptive filter stably in a short time to effectively remove recurrent sound even
when the acoustic environment changes rapidly and nonlinearly. is there.
[0008]
A regression sound removal apparatus according to the present invention includes control means
for instructing an acoustic environment to both of an acoustic environment formation unit that
includes at least a speaker system and a microphone system and that realizes one of a plurality of
acoustic environments and a regression sound removal unit; Recurrent sound removing means
for generating a pseudo regressive sound signal based on an audio signal to be input to the
system and subtracting the pseudo regressive sound signal from the sound collection signal
output from the microphone system.
Then, the regression sound removal means is a storage means for storing a plurality of adaptive
filter parameters set according to each of the plurality of acoustic environments, and when the
acoustic environment instruction is performed from the control means, based on the acoustic
environment instruction. The corresponding parameter is read out from the storage means, the
pseudo-regression sound signal is generated using the read-out parameter, and the parameter is
updated based on the result of subtracting the pseudo-regression sound signal at that time from
the previous sound collection signal. And an adaptive filter that generates a pseudo-regression
sound signal.
[0009]
04-05-2019
3
In this configuration, when the acoustic environment is instructed by the control means, the
acoustic environment formation means controls the directivity of the speaker system or the
microphone system to form a predetermined acoustic environment. The adaptive filter of the
regression sound removal means reads from the storage means a parameter corresponding to
the contents of the acoustic environment instruction, and sets the parameter. The adaptive filter
then filters the audio signal with the set parameters to generate the pseudo-regression sound
signal. The regression sound removal means obtains an output signal by subtracting the pseudo
regression sound signal from the sound collection signal. As described above, when the acoustic
environment is changed, the adaptive filter generates the pseudo-regression sound signal based
on the parameter stored in advance in the storage unit according to the newly set acoustic
environment. Then, after the initial processing after the change of the acoustic environment, the
operation of the normal adaptive filter, that is, the operation of generating the pseudo-regression
sound signal while sequentially updating the parameters to the optimum condition based on the
previous error signal is repeated. .
[0010]
Thus, even if the acoustic environment changes rapidly and non-linearly, initial parameters
suitable for the new acoustic environment can be set immediately, and optimum parameters can
be obtained in a short time.
[0011]
Further, the adaptive filter of the regression sound removing device according to the present
invention updates and stores the currently used parameter in the storage means when a new
acoustic environment instruction is received, and reads out the parameter based on the new
acoustic environment instruction. It is characterized by
[0012]
In this configuration, the parameter optimized by the adaptive filter is fed back to the storage
means and stored.
As a result, when the same acoustic environment instruction is given next, the initial parameter
setting contents become closer to the optimum condition for the instructed acoustic
environment, and the optimal parameter can be obtained in a short time. Can.
04-05-2019
4
[0013]
Further, in the regression sound removal apparatus according to the present invention, the
speaker system is a speaker array, the acoustic environment is set by the directivity of the
speaker, and the directivity of the speaker array is changed according to the acoustic
environment instruction and the parameter of the adaptive filter is switched. It is characterized
by
[0014]
In this configuration, parameters of the adaptive filter are stored corresponding to the directivity
of the speaker array, and the parameters are read out and set in the adaptive filter based on the
directivity of the instructed speaker array.
[0015]
Further, in the regression sound removal apparatus according to the present invention, the
microphone system is a microphone array, the acoustic environment is set by the directivity of
the microphone, and the directivity of the microphone array is changed according to the acoustic
environment instruction and the parameter of the adaptive filter is switched. It is characterized
by
[0016]
In this configuration, the parameters of the adaptive filter are stored corresponding to the
directivity of the microphone array, and the parameters are read out and set in the adaptive filter
based on the directivity of the instructed microphone array.
[0017]
Further, in the regression sound removal apparatus of the present invention, the speaker system
is a speaker array, and the microphone system is a microphone array, and the acoustic
environment is set by the directivity of the speaker and the directivity of the microphone. It is
characterized in that the directivity of the array and the directivity of the microphone array are
changed and the parameters of the adaptive filter are switched.
[0018]
In this configuration, the parameters of the adaptive filter are stored corresponding to the
directivity of the speaker array and the microphone array, and the parameters are read out based
on the directivity of the instructed speaker array and the microphone array, and adaptive Set to
04-05-2019
5
filter.
[0019]
According to the present invention, since the parameter suitable for the designated acoustic
environment is set to the adaptive filter at the initial stage of change, the adaptive filter can be
changed in a short time even if control is performed to change the acoustic environment rapidly
and nonlinearly. Stable operation.
[0020]
A regression sound removing device according to a first embodiment of the present invention
will be described with reference to FIGS. 1 to 3.
In the present embodiment, an echo canceller will be described as an example of the regression
sound removing device.
FIG. 1 is a block diagram showing the main part of the echo canceller of this embodiment.
FIG. 2 is a conceptual view of filter parameters stored in the memory 13 shown in FIG.
FIG. 3 is a flow chart showing an echo cancellation processing flow of the echo canceller of this
embodiment.
[0021]
The echo canceller of this embodiment includes an echo cancellation unit 1, a speaker unit 3, a
microphone unit 4, a sound collection directivity control unit 5, a control unit 7, and an operation
input unit 8.
[0022]
The control unit 7 controls the entire echo canceller, and based on the sound environment
04-05-2019
6
setting received by the operation input unit 8, the sound environment instruction data is
collected by the directivity control unit 5 and the adaptive type of the echo cancellation unit 1.
Give to the filter 11.
The operation input unit 8 includes operators such as a plurality of buttons, receives various
setting inputs from the user, and gives the input to the control unit 7.
[0023]
The speaker unit 3 is composed of a single speaker, converts the received sound signal and emits
the sound.
The microphone unit 4 is composed of a microphone array in which a plurality of microphones
are arranged and formed, and each microphone collects an external sound including the talk
sound of the talker and outputs it to the sound collection directivity control unit 5.
[0024]
The sound collection directivity control unit 5 delays and adds the output signals from the
microphones of the microphone array based on the sound environment instruction data supplied
from the control unit 7 to obtain a sound collection signal having a sound collection directivity in
a predetermined direction. Generate
[0025]
The echo cancellation unit 1 includes an adaptive filter 11, an adder (subtractor) 12, and a
memory 13.
The adaptive filter 11 includes an FIR filter, and setting the delay coefficient and the filter
coefficient of the FIR filter to predetermined values enables pseudo echo (regression using
regression of the received voice signal input from the voice signal input terminal 2). Sound)
Generate a signal.
04-05-2019
7
The adder 12 subtracts the pseudo echo signal from the sound collection signal input from the
sound collection directivity control unit 5 and outputs the result.
This output signal becomes an error signal and a transmission voice signal, and the transmission
voice signal is transmitted to the other side via the voice signal output terminal 6, and the error
signal is fed back to the adaptive filter 11.
As shown in FIG. 2A, the memory 13 stores filter parameters in advance for each sound
collection directivity. Specifically, the filter parameter is set for each sound collection directivity
set by the microphone unit 4 and the sound collection directivity control unit 5, and is configured
by the delay coefficient and the filter coefficient of the FIR filter of the adaptive filter 11. Be done.
For example, as shown in FIG. 2, if there are only sound collecting directivity A and B to M, filter
parameters corresponding to sound collecting directivity A and B to M exist only a0 and b0 to
m0 Do. Further, detailed delay coefficients and filter coefficients are set for each of the filter
parameters a0 and b0 to m0.
[0026]
Next, the operation of the adaptive filter 11 will be specifically described along the flowchart of
FIG.
[0027]
When the user operates the operation input unit 8 to set the acoustic environment, the control
unit 7 generates acoustic environment instruction data and applies the data to the adaptive filter
11.
When the acoustic environment instruction data is input from the control unit 7 (S101), the
adaptive filter 11 receives the acoustic environment instruction data and identifies the
designated sound collection directivity (S102).
[0028]
The adaptive filter 11 reads out each delay coefficient and filter coefficient currently set in the
04-05-2019
8
FIR filter, and writes them in the memory 13 as filter parameters corresponding to the
corresponding sound collection directivity (S103). At this time, although the filter parameters of
the previous (initial state or state of the previous update) are stored in the memory 13, the
adaptive filter 11 sets new filter parameters to the filter parameters already stored. Overwrite.
For example, in the initial state as shown in FIG. 2A, the filter parameter stored for the sound
collection directivity B is b0, but in the adaptive filter 11, the filter parameter b1 for the sound
collection directivity B is If present, the adaptive filter 11 overwrites the filter parameter b1 with
the filter parameter b0.
[0029]
When the adaptive filter 11 writes the filter parameter set to itself in the memory 13, the
adaptive filter 11 reads the filter parameter corresponding to the identified sound collection
directivity (S104). Then, the adaptive filter 11 sets the delay coefficient and the filter coefficient
of the FIR filter based on the read filter parameter (S105).
[0030]
The adaptive filter 11 performs a convolution operation or multiplication process on the input
reception voice signal with a delay coefficient and a filter coefficient (impulse response) set based
on the acoustic environment instruction data, and generates a pseudo echo signal. (S106). Then,
the adder 12 subtracts the pseudo echo signal from the sound collection signal as described
above and outputs it.
[0031]
As described above, by reading and using the filter parameters according to the new acoustic
environment stored in the memory 13 simultaneously with the change of the acoustic
environment, the filter suitable for the acoustic environment from the initial state after the
acoustic environment change Since the parameters are obtained, the filter parameters of the
adaptive filter 11 can be optimized in a short time. Thereby, stable echo cancellation can be
realized in a short time.
[0032]
04-05-2019
9
The adaptive filter 11 inputs an error signal generated by the subtraction process of the adder
12 (S107), and calculates and sets an optimal filter parameter at that time using a known
learning identification method or the like (S108). . Then, if there is no input of the sound
environment instruction data, the adaptive filter 11 generates a pseudo echo signal by using the
optimized filter parameter (S108 → S101 → S106).
[0033]
The generation of the pseudo echo signal, the input of the error signal, and the calculation and
setting of the optimum filter parameters (S106 → S107 → S108) are the operations of the
normal adaptive filter 11, and continuously unless the acoustic environment instruction data is
input. To be executed. Thereby, the filter parameters are updated as needed, and asymptotically
approach more truly optimal filter parameters.
[0034]
Then, when the acoustic environment instruction data is input, the adaptive filter 11 overwrites
and stores, in the memory 13, filter parameters more optimized for the current acoustic
environment. By performing such processing, when the same acoustic environment is set next
time, the currently optimized filter parameter can be used, so the adaptive filter 11 can be
optimized in a shorter time next time. Can be done. As a result, stable echo cancellation can be
realized in a shorter time.
[0035]
Next, a regression sound removing device according to a second embodiment will be described
with reference to FIG. 4 and FIG. Also in the present embodiment, an echo canceller will be
described as an example of the regression sound removing device. FIG. 4 is a block diagram
showing the main part of the echo canceller of this embodiment. FIG. 5 is a conceptual view of
filter parameters stored in the memory 13 shown in FIG. The echo canceller shown in FIG. 4 is, in
contrast to FIG. 1, formed of a speaker array in which the speaker unit 3 is formed by arranging
a plurality of speakers, and the sound emission directivity control unit 9 is disposed between the
echo cancellation unit 1 and the speaker unit 3. Is inserted. Further, the echo canceller shown in
04-05-2019
10
FIG. 4 gives the sound environment instruction data to the sound emission directivity control unit
9 as well as the sound collection directivity control unit 5 from the control unit 7.
[0036]
In such an echo canceller, when the sound environment setting is input, the control unit 7
supplies sound environment instruction data to the sound emission directivity control unit 9 and
the sound collection directivity control unit 5. The sound emission directivity control unit 9
performs delay control of the audio signal output to each speaker of the speaker array based on
the sound environment instruction data, and controls the directivity of the sound emitted from
the speaker unit 3. The sound collection directivity control unit 5 delay-controls the output signal
from each microphone of the microphone array, and generates a sound collection signal having
sound collection directivity in a predetermined direction. As described above, the speaker unit 3
is configured by the speaker array, the microphone unit 4 is configured by the microphone array,
and the sound emission directivity control unit 9 and the sound collection directivity control unit
5 are provided to further enhance the acoustic environment. It can be realized.
[0037]
As shown in FIG. 5, the memory 13 stores filter parameters for each combination of sound
collection directivity and sound emission directivity. For example, when there are only sound
collecting directivity of A and B to M and sound emission directivity of only α and β to 、,
corresponding to each combination, Aα0 to A0∼0, Bα0 to B ・ ・ ・ 0,. , Mα0 to Mρ0 are set
and stored.
[0038]
When the acoustic environment instruction data is input from the control unit 7, the adaptive
filter 11 analyzes the acoustic environment instruction data to detect a combination of sound
collection directivity and sound emission directivity. Then, the adaptive filter 11 reads the
corresponding filter parameter, and sets the delay coefficient and the filter coefficient of the FIR
filter. The other operation processing of the adaptive filter 11 is the same as that of the first
embodiment, so the description will be omitted.
04-05-2019
11
[0039]
As described above, even in the sound environment in which both the sound emission directivity
and the sound collection directivity can be set, that is, the sound environment that can be set in
more various ways than in the first embodiment, it is stored in the memory. Since it is sufficient
to read out and set the filter parameters, stable adaptive echo cancellation can be realized by
optimizing the adaptive filter in a short time according to the set acoustic environment.
[0040]
In particular, when the acoustic environment is diverse as in the present embodiment, stable
echo cancellation can be effectively realized in a shorter time than in the prior art by using the
configuration of the present invention.
[0041]
Next, a regression sound removing device according to a third embodiment will be described
with reference to FIG.
The present embodiment is different from the second embodiment in the filter parameter storage
and setting method of the apparatus, and the other configuration is the same.
Therefore, the description of the same parts of the configuration will be omitted. FIG. 6 is a
conceptual view of filter parameters stored in the memory of the echo canceller of this
embodiment. In the echo canceller of this embodiment, filter parameters are set in advance and
stored in the memory 13 only for the combination of the sound collection directivity and the
sound emission directivity which are previously determined to be used (see FIG. 6 (A)).
[0042]
Then, the echo canceller reads out the filter parameter corresponding to the combination when
the combination (sound environment setting) of the sound collection directivity and the sound
emission directivity is stored, and sends it to the adaptive filter 11. Set
[0043]
04-05-2019
12
By using such a configuration, it is possible to reduce the number of combinations of sound
collection directivity and sound emission directivity and filter parameters stored in the memory
13 as much as possible and save memory resources. be able to.
Note that even with such a method for storing and setting filter parameters, the above-described
updated storage of filter parameters can be performed.
[0044]
By the way, when setting such filter parameters, there are cases where the user instructs a
combination of sound collection directivity and sound emission directivity which are not set and
stored in advance. In this case, the echo canceller may set the filter parameters of the adaptive
filter 11 by any of the following methods.
[0045]
(1) Store general-purpose filter parameters that are not related to the combination of sound
collection directivity and sound emission directivity. (2) Continue to use the filter parameters
before the user sets the sound environment. (3) A similar combination is detected from the
combination of sound collection directivity and sound emission directivity already stored for the
designated sound collection directivity and sound emission directivity, and this similar collection
is collected. Use filter parameters corresponding to the combination of sound directivity and
sound emission directivity. For example, each sound collection directivity and each sound
emission directivity are made ID based on the characteristics of each directivity, and a similar ID
is selected from the characteristics of each directivity newly set and detected by the user. To
realize.
[0046]
Furthermore, the echo canceller of the present embodiment may be provided with a learning
function of filter parameters shown below. The echo canceller secures, in the memory 13, an
area for storing filter parameters for the combination of the sound collecting directivity and the
sound emitting directivity, when a combination of the sound collecting directivity and the sound
emitting directivity which is not stored is instructed. (Refer FIG. 6 (B).).
04-05-2019
13
[0047]
After that, the adaptive filter 11 operates as in the above embodiment to update the filter
coefficients. Then, when a different acoustic environment is set by the user, the adaptive filter 11
stores the latest filter parameter set in itself in the corresponding area (the newly secured area
described above) of the memory 13 (FIG. 6). (C).).
[0048]
With such a configuration, the combination of the added sound collecting directivity and the
sound emitting directivity and the filter parameter is stored, and the added combination of the
sound collecting directivity and the sound emitting directivity is instructed again. In this case, an
optimal filter coefficient can be obtained in a short time.
[0049]
In addition, as a new storage method of filter parameters, when securing an area corresponding
to the new filter parameters in the memory 13, for example, sound collection directivity and
noise emission direction with the least number of usage or the shortest usage time There is also a
method of eliminating the combination of the gender and the filter parameter.
In this case, in the memory 13, the number of times of use and the use time are accumulated and
stored together with the combination of sound collection directivity and sound emission
directivity and the filter parameter. The adaptive filter 11 reads out the number of times of use
and the use time, and arranges combinations of sound collection directivity and sound emission
directivity and filter parameters to erase the lowermost set. Then, in the area formed by this
processing, a new combination of sound collection directivity and sound emission directivity and
a set of filter parameters are stored.
[0050]
In such a configuration, as well as saving memory resources and storing easy-to-use filter
parameters, it is possible to realize a user-friendly echo canceller with a limited memory.
04-05-2019
14
[0051]
In each of the embodiments described above, the case of one transmission line of the received
voice signal is shown, but as shown in FIG. The above-described configuration can be applied to
achieve the above-described effects.
[0052]
FIG. 7 is a block diagram showing the main part of an echo canceller of another configuration.
The echo canceller shown in FIG. 7 has three transmission lines of the received voice signal, and
is constituted by, for example, a speaker array by performing delay control and amplitude control
of each received voice signal by the sound emission directivity control unit 9. The speaker system
3 realizes a plurality of virtual point sound sources.
Further, in the echo canceller shown in FIG. 7, the microphone unit 4 includes only a single
microphone, and the sound collection directivity control unit 5 is omitted.
[0053]
In such a configuration, the adaptive filter 11 is provided with three functional units respectively
corresponding to each channel, and each functional unit generates a pseudo echo signal for each
received audio signal of each channel. At this time, in the memory 13, filter parameters are
stored and set for each sound emission directivity corresponding to each received voice signal.
[0054]
The microphone unit 4 may be configured by a microphone array, and a sound collection
directivity control unit may be provided. In this case, a combination of sound emission directivity
and sound collection directivity for each received audio signal is possible. The filter parameters
are stored and set every time.
[0055]
04-05-2019
15
Further, although the example shown in FIG. 7 shows the case where a plurality of virtual point
sound sources are realized, the configuration of the present invention can be applied even when
a plurality of speakers are actually installed and sound is emitted.
Furthermore, if not only the speaker unit and the microphone unit but also the acoustic space
(the size and shape of the room, etc.) is variable, the above configuration is applied by setting the
filter parameters including these as well. Can.
[0056]
In the above description, the coefficients of the adaptive filter are switched according to the
sound emission directivity of the speaker array and the sound collection directivity of the
microphone array, but in each embodiment of the present invention, directivity control by the
array is performed. It is not limited to For example, even if it is one speaker unit and a
microphone unit, this invention is applicable if it can control and detect the installation direction.
[0057]
Also, in the above description, the echo canceller has been described, but in the case of a device
in which the sound emitted from the speaker gets into the microphone and is collected
(regressed), the configuration of the present invention is applied. The aforementioned effects can
be achieved. An example of this is a howling canceler.
[0058]
In the above description, the filter parameter optimized by the adaptive filter 11 is overwritten in
the memory 13. However, this process is not performed, and the filter preset in the memory 13 is
received each time the acoustic environment instruction data is received. Parameters may be
used each time.
[0059]
It is a block diagram which shows the principal part of the echo canceller of 1st Embodiment.
04-05-2019
16
It is a conceptual diagram of the filter parameter memorize | stored in the memory 13 shown in
FIG. It is a flowchart which shows the echo cancellation processing flow of the echo canceller of
1st Embodiment. It is a block diagram which shows the principal part of the echo canceller of
2nd Embodiment. FIG. 5 is a conceptual view of filter parameters stored in a memory 13 shown
in FIG. 4; It is a conceptual diagram of the filter parameter memorize | stored in the memory of
the echo canceller of 3rd Embodiment. FIG. 10 is a block diagram showing the main part of
another configuration of the echo canceller.
Explanation of sign
[0060]
1-echo canceller, 11-adaptive filter, 12-adder, 13-memory, 2-voice signal input terminal, 3speaker unit, 4-microphone unit, 5-sound collection directivity control unit, 6-voice signal Output
terminal, 7-control unit, 8-operation input unit, 9-noise emission directivity control unit
04-05-2019
17
Документ
Категория
Без категории
Просмотров
0
Размер файла
27 Кб
Теги
jp2007096389
1/--страниц
Пожаловаться на содержимое документа