close

Вход

Забыли?

вход по аккаунту

?

DESCRIPTION JP2008288718

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 JP2008288718
An acoustic echo canceller is provided that can set optimum parameters based on the
reverberation characteristics of an acoustic space in which an acoustic echo canceller is installed,
and can perform echo cancellation processing simply and optimally. SOLUTION: A process of
updating a filter coefficient in an acoustic space in which an acoustic echo canceller 27 is
installed with a sound such as a power on sound is performed for a plurality of tap numbers
before a user makes a speech, and the filter coefficient is previously determined. Determine the
shortest number of taps that converges below a defined threshold. Then, the echo suppressor 12
performs acoustic processing using the parameter associated with the determined number of
taps. [Selected figure] Figure 2
Acoustic echo canceller
[0001]
The present invention relates to an acoustic echo canceller.
[0002]
In recent years, the use of a call form called a speech call has been spreading.
In the conventional telephone, while the handset (handset) is held to make a call, in the case of a
loud-phone call, the speaker and the microphone installed on the desktop make the call. Acoustic
echo is the most important problem in realizing a speech communication. Acoustic echo is a
15-04-2019
1
phenomenon where one's voice is emitted from the other party's speaker, enters the other party's
microphone, returns to one's side and is emitted from the speaker, and as a result, one's own
voice is perceived as being delayed in speech It is. Depending on the degree of acoustic echo, it
may be difficult to keep talking because of discomfort, or it may be impossible to talk by causing
howling.
[0003]
Although there are various methods for controlling the above-mentioned acoustic echo, among
them, the acoustic echo canceller using an adaptive filter is the most effective method. In the
acoustic echo canceller, the transmission path of the echo from the speaker to the microphone is
identified, a pseudo echo is generated by convolving the filter and the reference signal, and the
acoustic echo is controlled by subtracting the pseudo echo from the signal generated by the
microphone. Do. The above filter needs to be used with an optimum setting specific to each
acoustic space in which the acoustic echo canceller is installed. Therefore, it is necessary to set
the filter based on the acoustic characteristics of the space where the terminal is installed.
[0004]
Patent Document 1 discloses, as an acoustic echo canceller using a FIR (Finite Impulse Response)
filter, a technology related to an acoustic echo canceller with a long tap length and a fast
convergence time of filter coefficients. Unexamined-Japanese-Patent No. 2006-157498
[0005]
However, in the technique disclosed in Patent Document 1, only the tap length of the filter is
selected according to the reverberation characteristic of the room, and in that method, the echo
cancellation processing is optimized, including the double talk characteristic. It is not possible.
Further, since the setting of the filter coefficient is performed by the voice of the conversation,
the filter coefficient does not converge immediately after the start of the conversation, and there
is a problem that the operation becomes unstable.
[0006]
The present invention has been made in view of the above problems, and can set optimum
15-04-2019
2
parameters based on the reverberation characteristics of the acoustic space in which the acoustic
echo canceller is installed, and can perform echo cancellation processing simply and optimally.
The purpose is to provide technology to
[0007]
The configuration of the acoustic echo canceller according to the present invention comprises:
receiving means for receiving a first audio signal from a communication network; storage means
for storing electronic sound data; and the first audio signal received by the receiving means
Means for reproducing electronic sound data stored in the means; input means for picking up the
sound and generating a second sound signal corresponding to the sound; and the first sound
signal received by the receiving means A filter for filtering using a filter coefficient to generate a
third audio signal from the first audio signal, and a second audio signal generated by the input
means and a third audio signal generated by the filter Filter coefficient setting means for setting
the filter coefficient so as to minimize the difference; and subtracting the third audio signal from
the second audio signal generated by the input means to generate a fourth audio signal
Subtraction processing means, sound processing means for performing acoustic processing on
the fourth sound signal generated by the subtraction means based on the parameters to generate
a fifth sound signal, and fifth generation means for generating the fifth sound signal Output
means for outputting the voice signal of the signal to the communication network, measurement
means for measuring the acoustic characteristics of the room, and the number of taps of the filter
and the parameters of the acoustic processing means in correspondence with the acoustic
characteristics of the room Setting the number of taps corresponding to the indoor acoustic
characteristic measured by the measuring means in the second memory means for storing the
second table and the table stored in the second memory means in the filter; And setting means
for setting, in the sound processing means, parameters corresponding to the indoor acoustic
characteristics measured by the measuring means.
In the above configuration, the measurement unit may measure the reverberation time in the
room.
[0008]
Further, in any one of the above configurations, the parameter is a gain spectrum in which a gain
is associated with a frequency, and the acoustic processing unit converts the fourth audio signal
into a signal in a frequency domain. At the same time, each frequency component of the
15-04-2019
3
converted frequency domain signal may be multiplied by the gain associated with the
corresponding frequency in the gain spectrum. In any of the above configurations, the
reproduction means may reproduce the electronic sound data when the reception means does
not receive the first audio signal.
[0009]
Further, another configuration of the acoustic echo canceller according to the present invention
is a receiving unit for receiving a first audio signal from a communication network, a storage unit
for storing electronic sound data, and a first audio received by the receiving unit. A reproduction
means for reproducing the signal and the electronic sound data stored in the storage means; an
input means for picking up a sound and generating a second audio signal corresponding to the
sound; a first means received by the reception means A filter that generates a third audio signal
from the first audio signal, the second audio signal generated by the input unit, and a third filter
generated by the filter. Filter coefficient setting means for setting the filter coefficient so as to
minimize the difference from the audio signal; and subtracting the third audio signal from the
second audio signal generated by the input means; sound A subtractor for generating a signal, a
sound processor for performing acoustic processing on the fourth audio signal generated by the
subtractor based on a parameter to generate a fifth audio signal, and the audio processor for
generating the fifth audio signal Output means for outputting the fifth voice signal to the
communication network; second storage means for storing a table in which the number of taps of
the filter and the parameter of the sound processing means are associated with each other;
Selection means for selecting the number of taps of the filter, and the number of taps selected by
the selection means are set in the filter, and a parameter corresponding to the number of taps
selected by the selection means in the table is the acoustic processing means And setting means
for setting to.
[0010]
According to the present invention, it is possible to set an optimal parameter based on the
reverberation characteristic of the acoustic space in which the echo canceller is installed, and to
perform an echo cancellation process optimally.
[0011]
Before describing the present invention, acoustic echo will be briefly described.
Echoes in communication can be broadly divided into two types: line echoes and acoustic echoes.
15-04-2019
4
A line echo is an echo produced by an electrical circuit, and the impulse response may be
considered to be short once the line connection is established. On the other hand, the acoustic
echo is generated by acoustic feedback in which the sound emitted from the speaker gets into
the microphone. In echo propagation from the speaker to the microphone, in addition to the
direct propagation path, there is a path reflected from the room wall or the speaker. Acoustic
echo is characterized in that its characteristic changes rapidly due to changes in the acoustic
space, such as the movement of a speaker, a speaker or a microphone, which has a long duration.
[0012]
(A: Configuration) Hereinafter, a conference system configured using a conference terminal
equipped with an acoustic echo canceller according to an embodiment of the present invention
will be described. FIG. 1 is a diagram showing the configuration of a conference system for
performing a teleconference. The conference terminals 100A and 100B having the same
configuration are installed in the acoustic spaces 200A and 200B, and both are connected by a
communication network 300 such as the Internet. The conference terminal 100A has an acoustic
echo canceller 27A, a speaker 15A and a microphone 16A connected to the acoustic echo
canceller 27A, and the conference terminal 100B has an acoustic echo canceller 27B, a speaker
15B and a microphone 16B connected to the acoustic echo canceller 27B. . The speakers 15A,
15B convert the received audio signal into an analog signal and emit the audio. The microphones
16A and 16B generate an audio signal representing the collected voice, convert the audio signal
into a digital signal, and output the digital signal.
[0013]
Hereinafter, when there is no need to distinguish between the conference terminals 100A and
100B, they are referred to as the conference terminal 100. The same applies to the acoustic
spaces 200A and 200B. Further, when it is not necessary to indicate to which of the conference
terminals 100A or 100B the components of the conference terminal 100 also belong, they are
represented without adding an alphabet to the code, such as the speaker 15, the microphone 16
or the like.
15-04-2019
5
[0014]
Next, the configuration of the acoustic echo canceller 27 will be described with reference to FIG.
The acoustic echo canceller 27 includes a control unit 20 and a storage unit 30. The control unit
20 controls the operation of each unit of the acoustic echo canceller 27. The storage unit 30
stores a control program for executing the process of the acoustic echo canceller 27, a parameter
table, and electronic sound data.
[0015]
Here, the parameter table stored in the storage unit 30 will be described. As shown in FIG. 3,
parameters for optimizing the cancellation function of acoustic echoes are written in the
parameter table in association with the number of taps. In the present embodiment, gain
spectrum data is associated as a parameter for optimizing the acoustic echo cancellation
function. The gain spectrum data is a parameter used when the echo suppressor 12 removes an
acoustic echo component, the details of which will be described later.
[0016]
Hereinafter, the flow of the audio signal in the acoustic echo canceller 27 will be described with
reference to FIG. A voice signal (hereinafter referred to as reference signal X) transmitted from
the conference terminal 100 on the opposite side via the communication network 300 is
received via the reference signal input terminal 1. The received reference signal X is sent to the
speaker 15 via the reference signal output terminal 2. The reference signal X is also sent to the
reference signal storage unit 5 via the reference signal input terminal 1. The reference signal
storage unit 5 receives and stores the reference signal X. Since the reference signal storage unit
5 deletes new data when new data is received and writes new data, the reference signal storage
unit 5 always stores the reference signal X for the latest predetermined time.
[0017]
The filter coefficient storage unit 6 stores filter coefficients used by the filter processing unit 7 to
generate the pseudo echo Z from the reference signal X. The filter processing unit 7 generates a
pseudo echo Z by filtering the reference signal X stored in the reference signal storage unit 5
15-04-2019
6
with the filter coefficient stored in the filter coefficient storage unit 6, and generates the
generated pseudo echo Z The error calculation processing unit 8 and the second error
calculation processing unit 10 are output. The method of generating the pseudo echo Z by the
filter processing unit 7 will be described later.
[0018]
The voice signal input from the microphone 16 through the transmission signal input terminal 3
is sent to the first error calculation processing unit 8. The voice signal is also sent to the
transmission input storage unit 9. The transmission input storage unit 9 stores the received
audio signal. In addition, when data is newly received, the transmission input storage unit 9
always stores an audio signal for the latest predetermined time in order to delete old data and
write new data. The second error calculation processing unit 10 receives an audio signal from the
transmission input storage unit 9, subtracts the pseudo echo Z received from the filter processing
unit 7 to generate a difference signal, and outputs the difference signal to the filter coefficient
updating unit 11. The filter coefficient update unit 11 generates a new filter coefficient based on
the difference signal received from the second error calculation processing unit 10 and updates
the filter coefficient written in the filter coefficient storage unit 6.
[0019]
The first error calculation processing unit 8 subtracts the pseudo echo Z received from the filter
processing unit 7 from the audio signal received via the transmission signal input terminal 3 and
outputs the result to the echo suppressor 12. The echo suppressor 12 subjects the audio signal
received from the first error calculation processing unit 8 to acoustic processing in accordance
with the set parameters, and outputs the processed audio signal to the transmission signal output
terminal 4. The voice signal is output to the communication network 300 via the transmission
signal output terminal 4.
[0020]
Next, a method in which the filter processing unit 7 generates the pseudo echo Z will be
described. As shown in FIG. 4, the filter processing unit 7 is configured by an adaptive FIR filter.
The adaptive FIR filter is formed by combining a plurality of constituent units (taps) surrounded
by a broken line (L is the number of taps), and each tap corresponds to a unit delay element D
15-04-2019
7
and a multiplier Wn (n = 1, 2, ..., L). The unit delay element D delays the reference signal X by
one unit time. Each multiplier Wn indicates that the time-varying filter coefficient Wn (k) (the
value in parentheses is the time k). The same applies to the following. The input reference signal
X is subjected to product-sum operation by an adaptive FIR filter, and the operation result is
generated as a pseudo echo Z. That is, the pseudo echo Z (k) is generated by product-sum
operation using the filter coefficients of the time k and the reference signal input before the time
k.
[0021]
(B: Operation) Next, the operation of the acoustic echo canceller 27 will be described. (B-1: Initial
Setting Process) When starting the conference, the acoustic echo canceller 27 according to the
present embodiment evaluates the reverberation characteristics of the installed acoustic space
200 and performs an initial setting process of various parameters.
[0022]
FIG. 5 is a flowchart showing a flow of initial setting processing performed by the acoustic echo
canceller 27. When the acoustic echo canceller 27 is powered on by the user, the control unit 20
sets the number of taps of the adaptive FIR filter of the filter processing unit 7 to a
predetermined initial value (for example, 1024) (step SA100). Then, control unit 20 causes filter
coefficient storage unit 6 to set the filter coefficient at each tap to a predetermined initial value
(step SA200).
[0023]
Next, the control unit 20 causes the speaker 15 to reproduce the electronic sound data stored in
the storage unit 30 as a “power start-up sound” for a predetermined time (step SA300). A high
frequency (for example, 3 kHz) is preferable as the sound to be reproduced as the power start-up
sound, but it is sufficient if it has an acoustic characteristic suitable for measuring a
reverberation characteristic described later. The power up sound causes an echo in the acoustic
space 200. The microphone 16 generates an audio signal representing the echo (hereinafter, an
echo Y). The control unit 20 uses the echo Y to update the filter coefficient set in step SA200
(step SA400). Briefly, the filter processing unit 7 generates the pseudo echo Z from the reference
signal X (in this case, an audio signal representing the power-on sound) and the filter coefficient
15-04-2019
8
set in step SA200. The second error calculation processing unit 10 generates a difference signal
between the generated pseudo echo Z and the echo Y. The filter coefficient update unit 11
updates the filter coefficient stored in the filter coefficient storage unit 6 according to a
predetermined algorithm based on the difference signal.
[0024]
The LMS (Least Mean Square) method is used for the above algorithm. Briefly, the pseudo echo Z
generated by the filter processing of the filter processing unit 7 is most approximate to the echo
Y so that the difference signal generated by the second error calculation processing unit 10
approaches a silent audio signal. As described above, the filter coefficient update unit 11 updates
the filter coefficient. By updating the filter coefficients in this manner, filter coefficients
optimized to generate a pseudo echo Z imitating an echo Y from the reference signal X are
determined. The algorithm applied to the filter coefficient updating unit 11 may be other than
the LMS method.
[0025]
Now, as shown in FIG. 6, the filter coefficient shows a large value at the tap (the tap located on
the left side in FIG. 4) which processes the reference signal X having a small delay time, and the
filter coefficient at the tap having a large delay time is smaller. Show a tendency to
asymptotically to 0. The control unit 20 determines whether or not the absolute value of the
filter coefficient falls below a predetermined value (broken line) at the tap position with the
largest delay (1024 in the case of FIG. 6) (step SA500). It writes in storage part 30 corresponding
to each tap number.
[0026]
The storage unit 30 stores a plurality of values of the number of taps in addition to the value set
as the initial value (1024 in the present embodiment). After finishing the processing of steps
SA100 to SA500, the control unit 20 determines whether or not the filter coefficient update
processing has been executed for all of the plurality of stored tap numbers (step SA600). At this
point in time, it is at a stage where processing has been performed for the number of taps set as
the initial value, so the determination result in step SA600 is "No", and step SA700 is performed.
In step SA700, the number of taps which have not been tried is selected, and the processing of
15-04-2019
9
steps SA300 to 500 is performed using the number of taps.
[0027]
If data regarding whether or not the absolute value of the filter coefficient at the tap position
with the largest delay falls below a predetermined value is obtained for the filter with all taps, the
determination result in step SA600 becomes “Yes”, and the process proceeds to step SA800 .
In step SA800, control unit 20 compares the convergence status of the filter coefficients at each
number of taps, and the smallest number of taps among the filters for which the absolute value
of the filter coefficient at the tap position with the largest delay falls below a predetermined
value. Choose Also, the control unit 20 stores the filter coefficient optimized for the selected
number of taps in the filter coefficient storage unit 6. In addition, instead of the absolute value of
the filter coefficient at the tap position with the largest delay, the average value of the absolute
values of the filter coefficients at a plurality of (for example, 10) tap positions is calculated from
the tap position with the largest delay The average value may be compared with the
predetermined value.
[0028]
Next, control unit 20 refers to the parameter table stored in storage unit 30 to select gain
spectrum data corresponding to the number of taps selected in step SA800. Then, control unit 20
sets the parameter represented by the selected gain spectrum data in echo suppressor 12 (step
SA900).
[0029]
By performing the initial setting process in the mode described above, the following effects can
be obtained. That is, by setting the minimum number of taps for canceling the acoustic echo
generated in the acoustic space 200 in the filter processing unit 7, it is possible to realize an
acoustic echo canceller with a fast convergence time of the filter coefficient. Further, by
measuring the above-mentioned initial setting process before the start of the conference, the
echo cancellation and the double talk characteristic become optimum.
[0030]
15-04-2019
10
(B-2: Echo Canceling Process) Hereinafter, an echo canceling process performed by the acoustic
echo canceller 27 whose various parameters are optimized by the above-described initial setting
process will be described. FIG. 7 is a flowchart showing the flow of the echo canceling process.
First, when an audio signal (reference signal X) is transmitted from the conference terminal 100
on the opposite side, the audio represented by the reference signal X is emitted from the speaker
15 toward the acoustic space 200. The reference signal X is stored in the reference signal
storage unit 5 (step SB100).
[0031]
The filter processing unit 7 reads the filter coefficient from the filter coefficient storage unit 6
and generates a pseudo echo Z from the reference signal X stored in the reference signal storage
unit 5 using the filter coefficient (step SB200), The error calculation processing unit 8 and the
second error calculation processing unit 10 are output. The first error calculation processing unit
8 subtracts the pseudo echo Z generated by the filter processing unit 7 from the sound signal
received from the microphone 16 (step SB300), and outputs the generated sound signal to the
echo suppressor 12.
[0032]
At this time, in the audio signal received by the first error calculation processing unit 8 from the
microphone 16, the speech content of the user who uses the conference terminal 100 on which
the acoustic echo canceler 27 is mounted, and the reference signal X Contains an echo Y
generated when the sound is emitted. Therefore, the first error calculation processing unit 8
subtracts the pseudo echo Z generated by the filter processing unit 7 from the sound signal
received from the microphone 16 to generate a sound signal from which the user's speech
content is extracted. However, since it is difficult to generate the pseudo echo Z which completely
matches the echo Y by the filtering by the adaptive FIR filter, the voice signal output from the
first error calculation processing unit 8 is added to the content of the user's speech. , And an
audio signal corresponding to the difference between the echo Y and the pseudo echo Z.
[0033]
15-04-2019
11
The echo suppressor 12 performs echo suppressor processing on the audio signal received from
the first error calculation processing unit 8 (step SB400). Specifically, the echo suppressor 12
converts the received speech signal into a spectrum in the frequency domain by FFT (Fast Fourier
Transform). The frequency spectrum is filtered for each frequency to remove only the component
derived from the echo. The gain spectrum set in step SA900 is used as the filtering coefficient
used at that time.
[0034]
Specifically, the gain spectrum is a spectrum of gains defined in a specific frequency domain as
shown in FIG. In the gain spectrum, for the input signal, a high value is for the frequency
component derived from the signal to be passed (speaker's remark), and a low value for the
signal to be removed (signal from the echo). It is set. The echo suppressor 12 multiplies the FFTprocessed speech signal (spectrum) and the set gain spectrum to change the frequency
characteristic.
[0035]
The characteristics of the gain spectrum are appropriately set by the user through theory or
experiment. For example, when it is possible to predict in advance that the audio signal
transmitted and received by the conference terminal 100 is limited to a specific frequency width,
the gain may be set to 0 for bands other than the frequency band. When the conference terminal
100 is installed in the acoustic space 200 having specific acoustic characteristics, it is possible to
predict in advance the frequency that is likely to be included in the echo and set the gain in the
frequency band low. . Also, instead of or in addition to the multiplication of the spectra in the
frequency domain, processing may be performed to cut the frequency component for the
frequency component whose sound pressure level is below a predetermined threshold.
[0036]
Note that the acoustic processing performed by the echo suppressor 12 is not limited to the
above-described processing, and it is an acoustic processing that removes components that can
not be removed by the echo canceling processing using the above-described adaptive FIR filter.
good. When other acoustic processing is performed, parameters related to the processing may be
associated with the parameter table stored in the storage unit 30 instead of or in combination
15-04-2019
12
with the gain spectrum.
[0037]
The frequency domain signal whose frequency characteristic has been transformed is again
transformed into a time domain signal by inverse FFT, and is output to the transmission signal
output terminal 4. In step SB500, control unit 20 determines whether or not all processing of
reference signal X accumulated in reference signal accumulation unit 5 is completed. If all the
processes have been completed, the determination result in step SB500 is "Yes", and the process
ends. On the other hand, when the conference terminal 100 further receives the reference signal
X, the new reference signal X is written in the reference signal storage unit 5, and the
determination result in step SB500 is "No", and the reference signal X is determined. The
processes in and after step SB100 are executed.
[0038]
(C: Modifications) Although the embodiments of the present invention have been described
above, the present invention can be implemented in various aspects as follows.
[0039]
(1) In the above embodiment, the case of using the power-up sound by the electronic sound data
has been described when performing the process related to the updating of the filter coefficient.
However, the voice may not be the power-up sound. For example, it may be a tone or ringing
tone generated in communication with the opposite conference terminal, or a sound other than
the electronic tone may be emitted to update the filter coefficient.
[0040]
(2) Although the case where two conference terminals 100 are connected to the communication
network 300 has been described in the above embodiment, three or more conference terminals
100 may be connected to the communication network 300.
[0041]
15-04-2019
13
(3) Although the case where the present invention is applied to a general acoustic echo canceler
has been described in the above embodiment, the present invention may be applied to an
acoustic echo canceller of other aspects.
An example is shown below. When speaking toward a microphone placed in a very echo-prone
acoustic space, echoes generated in the acoustic space are also blown into the microphone, so
that an audible audio signal containing echoes is generated. As described above, it is possible to
cancel the echo by the voice which is not the voice emitted from the speaker or the like by using
the configuration of the present acoustic echo canceller. In that case, the electronic sound data
stored in advance in the storage unit 30 before voice generation is reproduced, and the collected
electronic sound is used as the reference signal X, and the same initial setting process as in the
above embodiment can be performed. Just do it. Therefore, the present invention can be applied
not only to a communication device that communicates voice mutually as in the conference
terminal 100 but also to a broadcast station, a server, a voice signal distribution device or the like
that unilaterally sends voice signals. In that case, the receiving means (reference signal input
terminal 1) for receiving an audio signal from the communication terminal is unnecessary.
Further, the present invention is similarly applicable to an audio signal generation apparatus
such as a recorder which does not transmit and receive an audio signal, but in such a case,
receiving means for receiving an audio signal from a communication network (reference signal
input terminal 1) In addition to the above, the output means (transmission signal output terminal
4) for outputting the voice signal to the communication network is also unnecessary.
[0042]
(4) In the above embodiment, the case where the characteristic function of the acoustic echo
canceller according to the present invention is realized by the software module has been
described, but it may be realized by the hardware module. In the embodiment described above,
the case where the program for causing the control unit 20 to execute the processing
characteristic of the communication terminal according to the present invention is stored in
advance in the storage unit 30 has been described. The program may be written and distributed
in a computer-readable recording medium such as a ROM (Compact Disk-Read Only Memory), or
may be distributed via a telecommunication line such as the Internet.
[0043]
15-04-2019
14
(5) In the above embodiment, the case where the conference terminal 100 is connected to the
network by the wired communication network 300 has been described, but the respective
devices may be connected wirelessly. The communication network 300 may be a LAN (Local
Area Network) or the like other than the Internet.
[0044]
(6) In the above embodiment, when determining the number of taps set in the filter processing
unit 7 in the initial setting process, it is determined whether or not the absolute value of the filter
coefficient at the tap position with the largest delay time is less than a predetermined value.
Then, an appropriate number of taps was selected from the number of taps for which the
judgment result was positive. However, the number of taps may be selected by the following
method. The number of taps is preferably set based on the acoustic characteristics of the acoustic
space 200 in which the conference terminal 100 is installed, in particular, the reverberation time.
Specifically, when the reverberation time is long, a larger number of taps may be set. Therefore,
reverberation time measurement means is provided for emitting a predetermined sound (for
example, electronic sound data in the above embodiment) in the acoustic space 200 and
measuring the reverberation time of the generated acoustic echo, and measurement is performed
by the reverberation time measurement means The number of taps corresponding to the
reverberation time may be set in the filter processing unit 7. In that case, a table in which a large
number of taps is associated with a long reverberation time may be stored in the storage unit 30
in advance. In the measurement of the reverberation time, the time required for the reverberation
curve indicating the passage of time of the sound pressure level to attenuate by a predetermined
ratio (for example, 60 dB) may be measured. Further, in the measurement of the reverberation
time, EDT (Early Decay Time) may be measured. EDT is a parameter that places particular
emphasis on the initial attenuation, and is the time taken to attenuate by 60 dB in the regression
line of the first 10 dB attenuation region of the reverberation curve. Also, the time required for
the sound pressure level of a predetermined impulse sound or the like to fall below a
predetermined threshold may be used.
[0045]
(7) In the above embodiment, when determining the number of taps set in the filter processing
unit 7 in the initial setting process, whether the absolute value of the filter coefficient at the tap
position where the delay time is the largest is smaller than a predetermined value It was judged.
Now, when the filter coefficients at each tap position are graphically displayed, as shown in FIG.
15-04-2019
15
6, in addition to the tendency that the filter coefficient becomes smaller as the tap position
becomes larger, small variations are observed between adjacent tap positions. It is common to As
a result, even if the filter coefficient at the tap position where the delay is large approaches 0, the
small variation causes the absolute value of the filter coefficient at the tap position where the
delay is largest (for example, 1024) to increase or the delay Even if the filter coefficient at a large
tap position does not asymptotically approach zero, the filter coefficient at the tap position with
the largest delay may accidentally show a value close to zero. In such a case, if the above
determination is performed based only on the filter coefficient at the tap position with the largest
delay, there is a possibility that an inappropriate determination may be made regarding the
convergence state of the filter coefficient. Therefore, the following method may be used. For
example, a low pass filter is applied to the graph of filter coefficients at each tap position (for
example, FIG. 6), and the above determination is performed for the filter coefficient at the tap
position with the largest delay in the graph generated as a result of filtering. Also good. In the
graph of filter coefficient values at each tap position, an envelope may be calculated, and the
above determination may be made for the filter coefficient at the tap position with the largest
delay in the envelope. By performing the processing as described above, it is possible to suppress
an inappropriate determination due to a small change in the filter coefficient.
[0046]
(8) In the modification (6), the reverberation characteristic of the acoustic space 200 is
measured, and the number of taps and the parameter in the echo suppressor 12 are set based on
the measured reverberation characteristic. However, the various parameters may be set based on
various acoustic characteristics in addition to the reverberation characteristics. For example, a C
value and a D value may be used which is a balance between the initial reflection sound and the
subsequent reverberation sound. These values are commonly used to evaluate the acoustic
performance of a hall or a classroom because they are closely related to clarity and
reverberation. The C value (Clarity [dB]) represents the energy ratio of the early reflection and
the late reverberation, and is defined by the following equation (1). Here, p (t) represents an
impulse response waveform. (Expression 1) <img class = "EMIRef" id = "202974240-00003" />
In Expression (1), t is a time indicating the boundary between the initial reflection sound and the
subsequent reverberation sound. The C value when t = 80 ms is called C80 etc. and is generally
used, but the value of t may be set appropriately and used. The D value (Definition or Deutlichkeit
[%]) represents an initial energy ratio to the total energy, and is defined by the following equation
(2). (Expression 2) <img class = "EMIRef" id = "202974240-00004" /> In the expression (2), t is a
time indicating the boundary between the initial reflection sound and the subsequent
reverberation sound. The D value when t = 50 ms is called D50 etc. and is generally used, but the
value of t may be set appropriately and used. Besides the method of actually generating sound in
the acoustic space to measure the acoustic characteristics as described above, there is a method
15-04-2019
16
of calculating the average sound absorption coefficient of the acoustic space to estimate the
acoustic characteristics. The average sound absorption coefficient is the product of the sound
absorption coefficient of the interior material of the acoustic space and the area where the
interior material is used for all areas of the inner surface (wall, ceiling, floor, etc.) of the acoustic
space. It is a value obtained by dividing the combined value by the inner surface area of the
acoustic space. The average sound absorption coefficient has a strong correlation with the
reverberation time in the acoustic space, and is used as a parameter for estimating the degree of
reverberation in the acoustic space. A table that associates the number of taps and the
parameters of the echo suppressor 12 with the various acoustic characteristics described above
is stored in the storage unit 30, and the number of taps and parameters corresponding to the
measured acoustic characteristics are set in the table. It may be done.
[0047]
(9) In the above embodiment, the case where various parameters are automatically set based on
the reverberation characteristic of the acoustic space 200 has been described. However, the user
of the acoustic echo canceller 27 may set (select) the various parameters based on the acoustic
characteristics of the acoustic space 200. The processing method in that case is described below.
The acoustic echo canceller 27 stores in the storage unit 30 a table in which the number of taps
of the adaptive FIR filter and the parameters of the echo suppressor 12 are associated with
various acoustic characteristics and written. As the above various acoustic characteristics, for
example, the type of acoustic space such as "hole", "anechoic room", "conference room" may be
used, and the degree of reverberation is "high", "medium", "low", etc. It may be a description of
acoustic characteristics, or may be the value of the number of taps set in the above adaptive FIR
filter. The acoustic echo canceller 27 further includes display means for displaying options
regarding the various acoustic characteristics, and selection means for the user to select the
options displayed on the display means, and the user displays on the display means From the
options selected, the option that is considered to be the best reflecting of the acoustic
characteristics of the acoustic space 200 is selected. Then, the acoustic echo canceller 27 may
set, in the adaptive FIR filter and the echo suppressor 12, the number of taps and parameters
associated with the choice of acoustic characteristics selected by the user in the above table. In
the above table, the number of taps of the adaptive FIR filter and the parameter of the echo
suppressor 12 may be associated and written. In that case, when the user selects the number of
taps, the selected number of taps is set in the adaptive FIR filter, and a parameter value
corresponding to the selected number of taps is set in the echo suppressor 12 in the table. You
may
[0048]
15-04-2019
17
It is a figure showing composition of a meeting system. FIG. 6 is a block diagram showing the
configuration of an acoustic echo canceller 27. It is a figure showing an example of a parameter
table. It is a figure showing the generation method of a false echo signal. 5 is a flowchart showing
a flow of initial setting processing. It is a graph showing the convergence condition of a filter
coefficient. It is a flowchart which shows the flow of an echo canceling process. It is a figure
which shows an example of a gain spectrum.
Explanation of sign
[0049]
DESCRIPTION OF SYMBOLS 1 ... reference signal input terminal, 2 ... reference signal output
terminal, 3 ... transmission signal input terminal, 4 ... transmission signal output terminal, 5 ...
reference signal storage part, 6 ... filter coefficient storage part, 7 ... filter processing part, 8: first
error calculation processing unit, 9: transmission input storage unit, 10: second error calculation
processing unit, 11: filter coefficient update unit, 12: echo suppressor, 15: speaker, 16:
microphone, 20: 20 Control unit 27 Acoustic echo canceller 30 Storage unit 100 Conference
terminal 200 Acoustic space 300 Communication network
15-04-2019
18
Документ
Категория
Без категории
Просмотров
0
Размер файла
33 Кб
Теги
description, jp2008288718
1/--страниц
Пожаловаться на содержимое документа