Multiplex communications – Crosstalk suppression
Reexamination Certificate
1998-03-02
2001-08-07
Vu, Huy D. (Department: 2663)
Multiplex communications
Crosstalk suppression
C370S286000, C379S413020
Reexamination Certificate
active
06272106
ABSTRACT:
TECHNICAL FIELD
This invention relates to a double talk detecting method which is carried out in a mobile communications network and in a long-distance telephone line network when echo is canceled, a double talk detecting apparatus and an echo canceler, which are suited to be used for carrying out the afore-mentioned method.
BACKGROUND ART
In a long-distance telephone line via a submarine cable or via a communication satellite, the subscriber's line, in general, connected to both ends of the line is of a two-wire circuit and its long-distance transmission portion is of a four-wire circuit employed for amplification of signal and some other purposes. Similarly, in the mobile communications network using a mobile telephone (or cellular phone), the subscriber's line of a terrestrial analog telephone is of a two-wire circuit and its portion from a terminal of the mobile telephone to a switch, etc. is of a four-wire circuit. In this case, the connection region between the two-wire and the four-wire is provided with a hybrid circuit for performing a four-wire/two-wire conversion. This hybrid circuit is designed to match with the impedance of the two-wire circuit. However, since it is difficult to obtain always a good matching condition, a received signal reaching an input side of the four-wire of the hybrid circuit tends to leak toward an output side of the four-wire, thereby generating a so-called echo. Since such an echo reaches the talker at a lower sound level than the talker's voice and after a delay of a predetermined time period, a speech hindrance is created. Such a speech hindrance caused by echo becomes significant as the signal propagation time becomes longer. Particularly, in the case of a mobile communication with the aid of a mobile telephone, since various processing procedures are carried out in the radio communication section leading to the switch, etc., the delay of signal is increased, thus resulting, particularly, in the problem of speech hindrance caused by echo.
As an apparatus for preventing a generation of echo, there are an echo suppressor and an echo canceler.
FIG. 1
shows a schematic construction of an echo canceler which can be used in a mobile communications network. The echo canceler
1
illustrated here is located on a front stage of a hybrid circuit
2
. In this illustration, the subscriber of an analog telephone is referred to as the “near-end talker” and the subscriber of a mobile telephone as the “far-end talker”. A far-end speech signal input into the echo canceler
1
is represented by Rin; a far-end speech signal output from the echo canceler
1
, by Rout; a near-end speech signal input into the echo canceler
1
, by Sin; and a near-end speech signal output from the echo canceler
1
; by Sout, respectively.
The echo canceler
1
shown in
FIG. 1
comprises an echo path estimation circuit/echo replica generator
3
, a control unit
4
, an adder
5
, and a non-linear processor
6
. Here, the echo path estimation circuit/echo replica generator
3
detects a response characteristic of the hybrid circuit
2
based on both the far-end speech input Rin and near-end speech input Sin and estimates an echo path (namely, echo propagating line). Then, an anticipated echo (namely, echo replica) from the hybrid circuit
2
is generated through a convolutional operation as a result of estimation and the far-end speech input Rin. In the adder
5
, this echo replica is subtracted from the near-end speech input Sin, thereby canceling the echo. As the above-mentioned echo path estimation algorithm, a learning identification algorithm is used. Among many adaptive algorithms, this learning identification algorithm is comparatively small in computational complexity and good in convergence characteristic.
As shown in
FIG. 1
, the echo path estimation circuit/echo replica generator
3
includes an echo path estimation circuit
3
a,
an H-register
3
b,
and an echo replica generator
3
c.
In this case, the echo path estimation circuit
3
a
estimates an echo path using the learning identification algorithm which is, among many other adaptive algorithms, generally comparatively small in computational complexity and good in convergence characteristic, and writes a tap coefficient (as later described) corresponding to the estimated echo path in the H-register
3
b.
The echo replica generator
3
c
comprises an FIR adaptive digital filter. The generator
3
c
generates an echo replica using the tap coefficient in the H-register
3
b
and through a convolutional operation with the far-end speech input Rin. The learning identification algorithm is a known estimation algorithm as disclosed, for example, in Institute of Electronics and Communication Engineers of Japan (IECE) Journal '77/11 Vol. J60-A NO.11, article under the heading of “Regarding Echo Canceling Characteristic of Echo Canceler Using Learning Identification Algorithm” (written by: Itakura and Nishikawa). The outline of the learning identification algorithm discussed in this article will be briefly described.
Firstly, if the impulse response h(t) and input signal x(t) are used presuming that the signal propagation characteristic of the echo path is linear, the echo y
k
at the time kT (where T is a sampling interval) can be expressed as follows.
y
k
=h
t
x
k
(1)
where:
h
=(
h
1
, h
2
, . . . , h
n
),
h
j
=h
(
j
T
)
x
k
=(
x
k−1
, x
k−2
, . . . , x
k−n
)
t
, x
j
=x
(
j
T
) (2)
(where
t
is transposition of vector)
On the other hand, if the estimation value of h at the time kT is represented by H
k
(hereinafter referred to as the “tap coefficient”), an estimation value Y
k
of y
k
can be given by the following expression.
Y
k
=H
k
t
x
k
(3)
Then, a successive correction of H
k
according to the learning identification algorithm is made by
H
k
+
1
=
{
H
k
+
α
(
x
k
e
k
)
/
(
x
k
x
h
)
(
x
t
x
k
≥
ne
2
)
H
k
(
x
t
x
k
<
ne
2
)
(
4
)
where:
e
k
=y
k
−Y
k
(5)
Namely, e
k
is a residual echo. This residual echo appears on the output side of the adder
5
. As apparent from the above expression (5), the next tap coefficient H
k+1
is calculated so that the residual echo will be reduced. Through calculation in the digital circuit, the above-mentioned algorithm can be specifically expressed as listed below. Firstly, the far-end speech input Rin, which is taken into the echo path estimation circuit
3
a,
is handled as a digital signal Xt (where t is a sampling time) having N pieces of sample values.
X
t
=(
x
(
t
),
x
(
t−
1), . . . , . . .
x
(
t−
(
N−
1)) (6)
If the tap coefficient H
t
at the time t written in the H-register
3
b
can be expressed as follows,
H
t
=(
h
t
(0),
h
t
(
t
), . . . ,
h
t
(
N−
1)) (7)
the convolutional operation in the echo replica generator
3
c
(FIR filter) can be expressed as follows.
Y
(
t
)
=
∑
i
=
0
N
-
1
(
t
-
1
)
×
h
t
(
i
)
(
8
)
If the inner product of the vector is represented by “*” here, the above expression (8) can be rewritten as follows.
Y
(
t
)=
x
t
*H
t
(9)
Now, if the residual echo obtained on the output side of the adder
5
is represented by er(t), the following expression can be obtained.
er
(
t
)=
e
(
t
)−
Y
(
t
) (10)
From the expressions so far listed, a fluctuation &Dgr;H
t
of H
t
can be expressed as follows.
&Dgr;
H
t
=g×er
(
t
)×
x
t
/(
x
t
*X
t
) (11)
H
t+1
can be expressed as follows.
H
t+1
=H
t
+&Dgr;H
t
(12)
Therefore, the echo path estimation circuit
3
a
reads the tap coefficient H in the H-register
3
b.
By adding &Dgr;H
t
, which is calculated in the expression (11), to the tap coefficient H thus read, the echo path estimation circuit
3
a,
in turn, calculates the next tap coefficient H
t+1
and writes it in the H-register
3
b.
In this way, the
Kawahara Toshiro
Naka Nobuhiko
Ohya Tomoyuki
Saigusa Masahito
NIT Mobile Communications Network Inc.
Vu Huy D.
LandOfFree
Method and device for detecting double-talk, and echo canceler does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and device for detecting double-talk, and echo canceler, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and device for detecting double-talk, and echo canceler will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2527514