close

Вход

Забыли?

вход по аккаунту

?

DESCRIPTION JP2004297791

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 JP2004297791
The present invention provides an echo cancellation apparatus which can immediately
communicate even when howling occurs and which exhibits good echo cancellation performance
from the start of a call. A first subtraction means 206 for estimating a transfer function
corresponding to reverberation in a room or the like and subtracting an output signal of a first
filter means 204 operating using the transfer function from a signal from the microphone side.
When the estimation accuracy of the transfer function is high, the second filter means 205 that
operates using the transfer function copied from the first filter means 204, and the output signal
of the second filter means 205 from the signal from the microphone side , And the howling
detection means 201, the notch filter means 202, and the switch means 208. [Selected figure]
Figure 2
Echo cancellation device, echo cancellation method, program and recording medium
[0001]
The present invention relates to an echo cancellation technology that prevents howling and echo
caused by a voice from a far-end speaker output from a speaker being input to a microphone
together with a voice of a near-end speaker.
[0002]
In an audio hands-free device such as a speakerphone type telephone, there is an echo
cancellation technology to prevent howling and echo.
15-04-2019
1
According to this echo cancellation technology, by subtracting the synthesized pseudo echo so
that the sound output from the speaker becomes the transfer function of the space from the
sound (echo) input to the microphone through the space such as a room, It can be as if there
were no echo.
[0003]
Below, the conventional echo cancellation technique is demonstrated. FIG. 6 is a functional block
diagram showing a conventional echo cancellation apparatus.
[0004]
In FIG. 6, a speaker 601 reproduces a received voice (voice from a far-end speaker) in a
speakerphone type telephone or the like, a microphone 602 picks up a transmitted voice (voice
from a near-end speaker), and 603 directly propagates A first echo cancellation unit that cancels
the echo that has passed through the path, a double talk detection unit that detects a double talk
state using the output signal of the first echo cancellation unit 603, and a 605 that cancels the
echo that has passed through the indirect propagation path. Second echo cancellation unit.
Unexamined-Japanese-Patent No. 5-48547
[0005]
However, in the above-mentioned conventional echo cancellation device, the environmental noise
may not be able to fully utilize the performance of the echo canceller and may lack stability. As a
result, setting of the learning timing of the first echo cancellation unit is difficult, unstable
behavior is exhibited at the start of a call, etc., it is difficult to completely avoid occurrence of
howling, and automatic recovery is not possible. I had the problem of stopping.
[0006]
In order to solve the above problems, an echo cancellation apparatus according to the present
invention includes transfer function estimation means for estimating a transfer function
corresponding to reverberation in a room or the like, which is added to voice before being output
from a speaker and input to a microphone; When the estimation accuracy of the transfer function
15-04-2019
2
estimation means is high, the first filter means that operates using the transfer function, the first
subtraction means that subtracts the output signal of the first filter means from the signal from
the microphone side, , Second filter means operating with the transfer function copied from the
first filter means, second subtraction means subtracting the output signal of the second filter
means from the signal from the microphone side, and howling detected Howling detection
means, notch filter means for notching a certain frequency band component from the signal from
the far end talker side, and To mainly characterized in that a switch means for selecting whether
processed signal in the signal or notch filter means.
[0007]
According to the above configuration, the echo cancellation apparatus of the present invention
can immediately communicate even if howling occurs, and exhibits good echo cancellation
performance from the start of the call.
[0008]
An embodiment of the present invention will be described with reference to FIGS.
[0009]
First Embodiment FIG. 1 is a block diagram showing a basic configuration of an echo cancellation
apparatus according to a first embodiment of the present invention, and FIG. 2 is a central
processing unit of the echo cancellation apparatus according to the first embodiment of the
present invention. FIG. 2 is a functional block diagram showing an echo cancellation method in a
speakerphone type telephone.
FIG. 3 is a flowchart showing the operation of the central processing unit of FIG.
Also, this function shows an outline of the program stored in the ROM.
[0010]
In FIG. 1, 101 is a telephone circuit device having an interface with a telephone line, 102 is an A
/ D converter for converting a received voice electric signal which is an analog electric signal into
a digital electric signal, and 103 is an analog electric signal. D / A converter for converting into a
speaker, 104 is a speaker for converting an analog electrical signal from the D / A converter 103
15-04-2019
3
into voice, 105 is a microphone for converting voice into an analog electrical signal, 106 is an
analog electrical signal from the microphone 105 A / D converter for converting digital electric
signals into digital electric signals, D / A converter 107 for converting digital electric signals into
analog electric signals (sending voice electric signals), 108 with A / D converter 102 and A / D
Digital signal processing is performed on the digital electrical signal obtained from the
conversion device 106, and Processing unit that outputs the D / A converter 103 and the D / A
converter 107, 109 is a ROM (Read Only Memory) in which a program for operating the central
processing unit 108 is stored, and 110 is a ROM 109 It is a RAM (Random Access Memory) used
when the central processing unit 108 operates according to the program being executed.
[0011]
In FIG. 2, reference numeral 201 denotes a howling detection means for detecting the occurrence
of howling.
When the howling detection means 201 detects a frequency band having a projecting portion in
the frequency spectrum of the signal from the far end talker side (hereinafter referred to as
received speech), it is in the frequency band where the protruding peak is generated. It is
determined that howling has occurred.
Reference numeral 202 denotes a notch filter means which is a bandstop type filter means for
notching a specific frequency band component, and 203 denotes a speaker 104 and a
microphone 105 according to the steepest descent method represented by a normalized LMS
(Least Mean Square) method or the like. Transfer function estimation means for estimating an
impulse response (impulse responce) in the space between the first and second filter means for
performing convolution operation of the estimated impulse response and the received speech;
First and second subtraction means for subtracting the output signals of the first and second
filter means 204 and 205 from the signal from the end talker side (hereinafter referred to as
transmission voice); It is a switch means for selecting whether or not the received voice passes
through the notch filter means 202 according to the detection result.
[0012]
The operation of the central processing unit 108 configured as described above will be described
with reference to FIG.
15-04-2019
4
[0013]
In FIG. 3, the transfer function estimation means 203 estimates an impulse response and outputs
it to the first filter means 204.
The first filter unit 204 performs a convolution operation of the impulse response input from the
transfer function estimation unit 203 and the received speech, and outputs the result of the
operation to the first subtraction unit 206. The first subtraction means 206 subtracts the
calculation result inputted from the first filter means 204 from the transmission voice inputted
from the microphone 105, and outputs the subtraction result to the transfer function estimation
means 203 (301). The transfer function estimation means 203 monitors the subtraction result
input from the first subtraction means 206 (302).
[0014]
If the estimation accuracy of the transfer function estimation means 203 is low and the
subtraction result input from the subtraction means 206 is not stable, the operation returns to
step 301.
[0015]
On the other hand, when the estimation accuracy of the transfer function estimation means 203
is high and the subtraction result input from the subtraction means 206 is stable, the second
filter means 205 is a filter representing the impulse response used in the first filter means 204.
Copy and store coefficients (303).
[0016]
The howling detection means 201 performs the howling detection operation (304), and if the
howling is not detected, the operation returns to step 301.
At the same time, the second filter means 205 performs a convolution operation of the impulse
response and the received voice using the filter coefficients stored in step 303, and outputs the
result of the convolution operation to the second subtraction means 207.
15-04-2019
5
The second subtraction means 207 subtracts the calculation result input from the second filter
means 205 from the transmission voice input from the microphone 105, and performs D / A
conversion on the far end speaker side of the transmission voice subjected to the echo
cancellation processing. Output to the device 107 side.
[0017]
On the other hand, when the howling detection means 201 detects howling, the switch means
208 switches to the notch filter means 202 side, and the received speech passes through the
notch filter means 202 and is sent to the D / A converter 103 on the near end speaker side. It is
output (305). The copying of the filter coefficients from the first filter means 204 to the second
filter means 205 is stopped (306). The second filter means 205 continues the echo cancellation
process using the filter coefficients stored before the howling detection means 201 detects
howling. The first filter means 204 initializes the filter coefficients (307). When using a
normalized LMS to estimate the impulse response, the initialization is to make the filter
coefficient 0. The transfer function estimation means 203 performs relearning from the state
where the filter coefficient of the first filter means 204 becomes 0, and approximates the impulse
response according to the subtraction result of the first subtraction means 206 (308). . When
learning is complete, the operation returns to step 301 (309).
[0018]
It is also possible to control the notch filter means 202 to be of a variable frequency type so that
the frequency band to be notched matches the frequency band in which the howling obtained by
the howling detection means 201 occurs.
[0019]
Further, in the present embodiment, the steepest descent method is used to estimate the transfer
function, but it is also possible to estimate the transfer function using other methods.
[0020]
As described above, according to the present embodiment, the transfer function estimating
means 203 for estimating the transfer function of the space between the speaker 104 and the
microphone 105 and the operation using the transfer function estimated by the transfer function
estimating means 203 If the estimation accuracy of the transfer function estimation unit 203 is
15-04-2019
6
high, the first filter unit 204 that performs the first filter, the first subtraction unit 206 that
subtracts the output signal of the first filter unit 204 from the signal from the microphone 105
side, A second filter means 205 operating using the transfer function copied from the first filter
means 204, and a second subtraction means 207 for subtracting the output signal of the second
filter means 205 from the signal from the microphone 105 side; Howling detection means 201
for detecting howling, and a notch for notching a certain frequency band component from a
signal from the far end talker side Filter means 202 and switch means 208 for selecting either
the signal from the far end talker side processed by notch filter means 202 or the signal from the
far end talker side not processed by notch filter means 202 Therefore, by performing echo
cancellation processing by removing the howling frequency by the notch filter means 202 at the
time of howling detection and using the transfer function stored before the howling detection as
it is, it is possible to talk immediately even if howling occurs. Become.
Also, at the time of howling detection, the transfer function of the first filter means 204 is
initialized, and the transfer function is learned using the signal from the far end talker side from
which the howling frequency component has been removed by the notch filter means 202. By
copying the transfer function from the first filter means 204 to the second filter means 205 after
learning of the function is completed, good echo cancellation performance can be exhibited from
the beginning of the call.
[0021]
Further, by using a computer by the program for executing the steps of the echo cancellation
method of FIG. 3, it becomes possible to execute the echo cancellation method of the present
embodiment at any time and at any time.
In addition, by reading a recording medium recording the above program by a computer, it
becomes possible to execute the above program at an arbitrary place at an arbitrary time.
[0022]
Second Embodiment FIG. 4 is a functional block diagram showing a central processing unit of an
echo cancellation apparatus according to a second embodiment of the present invention, and FIG.
5 is a flowchart showing an operation of the central processing unit of FIG. It is. The basic
configuration of the echo cancellation apparatus according to the present embodiment is the
15-04-2019
7
same as that shown in FIG. Also, this function shows an outline of the program stored in the
ROM.
[0023]
In FIG. 4, in a speakerphone type telephone or the like, a far-end speaker's speech, a near-end
speaker's speech and double talk (far-end speaker and near-end speaker Speaker detection
means for detecting simultaneous utterances), transfer function estimation means for estimating
the transfer function of the space between the speaker 104 and the microphone 105 by the
steepest descent method represented by the normalized LMS method etc., 403 is directly Direct
echo filter means for performing a convolution operation of a transfer function corresponding to
an echo component and received speech, 404 is an indirect echo filter means for performing a
convolution operation of a transfer function corresponding to an indirect echo component and
received speech, and 405 is a subtraction means is there. Here, the direct echo component refers
to the sound emitted from the speaker 104 and input directly to the microphone 105, and the
indirect echo component is emitted from the speaker 104 and reflected by an object such as a
wall, floor, or ceiling in space. It is a sound to be input to the microphone 105 later.
[0024]
The schematic operation of the echo cancellation apparatus configured as described above will
be described with reference to FIG.
[0025]
In FIG. 5, when echo cancellation processing is started (501), the speaker detection means 401
determines whether the speech state is far-end speaker utterance, near-end speaker utterance, or
double talk (502). If the speech state is the far-end speaker's speech, the transfer function
estimation means 402 performs direct echo component transfer function estimation (503) and
indirect echo component transfer function estimation (504) using an algorithm such as NLMS,
and direct echo filter means 403 Performs a convolution operation of the estimation result of the
direct echo component transfer function estimation (503) and the received speech (505), and the
indirect echo filter means 404 convolves the estimated result of the indirect echo component
transfer function estimation (504) with the received speech Calculation is performed (506), and
the transmitted voice from the microphone 105 and the result of the convolution calculation are
directly Removing code component and the indirect echo component (507).
[0026]
15-04-2019
8
This enables echo cancellation processing that realizes both speeding up and high precision of
transfer function estimation.
[0027]
As described above, according to the present embodiment, the direct echo filter unit 403
performs the convolution operation of the estimation result of the direct echo component
transfer function estimation (503) and the received speech, and the indirect echo filter unit 404
performs indirect echo component transfer. The convolution result of the estimation result of the
function estimation (504) and the reception voice is performed, and the direct transmission
component and the indirect echo component are removed by subtracting the transmission voice
from the microphone 105 and the convolution calculation result using the subtraction means
405 Thus, even when the volume from the speaker is increased, the determination accuracy of
double talk can be increased, and the double talk detection accuracy can be increased even when
the audio power ratio between the reception voice and the transmission voice is the same. .
[0028]
The echo cancellation apparatus, the echo cancellation method, the program and the recording
medium according to the present invention can immediately communicate even if howling
occurs, and the speakerphone type telephone etc. which needs the good echo cancellation
performance from the start of the call It can also be applied to applications.
[0029]
A block diagram showing a basic configuration of an echo cancellation apparatus according to a
first embodiment of the present invention A functional block diagram showing a central
processing unit of the echo cancellation apparatus according to the first embodiment of the
present invention Flowchart: A functional block diagram showing a central processing unit of an
echo cancellation apparatus according to a second embodiment of the present invention. A
flowchart showing an operation of the central processing unit of FIG.
Explanation of sign
[0030]
DESCRIPTION OF SYMBOLS 101 Telephone circuit apparatus 102, 106 A / D conversion
apparatus 103, 107 D / A conversion apparatus 104 Speaker 105 Microphone 108 Central
processing unit 109 ROM 110 RAM 201 Howling detection means 202 Notch filter means 203
Transfer function estimation means 204 1st Filter means 205 Second filter means 206 First
15-04-2019
9
subtraction means 207 Second subtraction means 208 Switch means
15-04-2019
10
Документ
Категория
Без категории
Просмотров
0
Размер файла
19 Кб
Теги
jp2004297791, description
1/--страниц
Пожаловаться на содержимое документа