Pulse or digital communications – Equalizers – Automatic
Reexamination Certificate
1999-03-05
2002-05-28
Pham, Chi (Department: 2631)
Pulse or digital communications
Equalizers
Automatic
C370S290000, C708S322000
Reexamination Certificate
active
06396872
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates to a method and its device for identifying the characteristics of an unknown system such as a transmission line, an acoustic coupling path and the like.
As applications of an unknown system identification performed by an adaptive filter, an echo canceler, a noise canceler, a howling canceler, an adaptive equalizer and the like are known. Here, taking an acoustic echo canceler which removes an acoustic echo leaking from a speaker into a microphone in the acoustic coupling path as an example, the prior art will be described below.
An echo canceler operates so as to suppress an acoustic echo leaking from a speaker into a microphone in the acoustic coupling path by generating an echo replica corresponding to a sending signal using an adaptive filter having a large number of tap coefficients more than that of an impulse response of the echo path. At this time, each tap coefficient of the adaptive filter is modified by correlating the error signal obtained by subtracting the echo replica from the mixed signal which is a compination of echo and the near-end signal, and far end signals.
As typical coefficient adaptation algorithms of these adaptive filters, ‘LMS algorithm’ in “PROCEEDINGS OF IEEE”, Vol. 63, No. 12, pp. 1692-1716, December, 1975 (reference 1) and ‘Learning Identification Method, LIM’ in “IEEE TRANSACTIONS ON AUTOMATIC CONTROL”, Vol. 12, No. 3, pp. 282-287, March, 1967(reference 2) are known.
The impulse response length of an acoustic coupling where an acoustic echo canceler is actually used depends upon a physical size of the acoustic space, a reflection factor of the wall or the like. For example, assuming a room used for videoconference and the like, the impulse response length may reach 1,000 taps, or occasionally several thousands taps. Therefore, it often difficulty in realizing it from the viewpoint of calculation amount or hardware size.
Thus, mainly for solving the problem of a calculation amount increase, subband adaptive filters in which input signals are divided into subbands and each subband uses an adaptive filter have been proposed in “IEEE SIGNAL PROCESSING MAGAZINE”, PP. 14-37, January, 1992 (reference 3).
As for the adaptive filter structure in each subband, various forms can be taken, but most common one is called an FIR (Finite Impulse Response)-type, and there is a reference made in “IEEE TRANSACTIONS ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING”, Vol. 27, No. 6, pp. 768-781, June, 1979 (reference 4).
The number N of taps of an adaptive FIR filter corresponding to each subband must be equal to the corresponding impulse response length, or longer than that. Commonly, in the case where an acoustic echo is divided into subbands, an impulse response in lower subbands is longer than that in higher subbands. This is mainly because an impulse response length of an acoustic coupling is determined by reflection, the reflection factor in a higher subband is smaller and higher subband components are more likely to be attenuated.
In
FIG. 16
, a typical impulse response of each subband in the case where an acoustic echo is divided into four (4) subbands is shown. Assuming that an impulse response length corresponding to the subband
1
through the subband
4
is M
1
,M
2
,M
3
, and M
4
, respectively, the impulse response lengths M
1
,M
2
,M
3
, and M
4
are given by an inequality M
1
≧M
2
≧M
3
≧M
4
. A subband adaptive filter which can reduce the calculation amount and shorten the convergence time by considering these characteristics and by adaptively assigning the number of taps of an adaptive filter of each subband has been proposed in “IEEE PROCEEDINGS OF INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING, Vol. V, pp. 3051-3053, April, 1995” (reference 5).
In
FIG. 17
, a block diagram of a subband adaptive filter which has been described in ‘reference 5’ is shown. The number of taps of an adaptive filter is adaptively assigned by evaluating a coefficient values and information of the subband input signal obtained from each adaptive filter.
According to the method of ‘reference 5’ shown in
FIG. 17
, at first, input signals are divided into a plurality of subbands in the analysis filter bank
3
and subband input signals are generated. Each subband input signal is decimated by a factor of 1/L
i
in the decimating circuit
50
i
(i=1,2, . . . , K), and is supplied to an independent adaptive filter
60
i
(i=1,2, . . . , K), respectively, and usually, where L
i
is K. On the other hand, the output of an unknown system
2
which is to be identified, that is to say, the echo to be cancelled the echo canceler is decimated by w /a factor of 1/L
i
in the decimating circuit
51
i
(i=1,2, . . . , K) after becoming a subband echo divided into a plurality of subbands in another analysis filter bank
4
having the entirely the same characteristics with the analysis filter bank
3
.
In an adaptive filter
60
i
, the subband error signal that is the difference between the subband replica which is the output of the adaptive filter and the subband echo which is decimated is generated. An adaptive filter
60
i
performs a coefficient adaptation using this subband error signal. This subband error signal is supplied to the synthesis filter bank
8
after it is interpolated w/a factor of L
i
in an interpolation circuit
70
i
(i=1,2, . . . , K), and transported to the output terminal
9
after it is synthesized into the full band.
Specifically, the signal obtained at the output terminal
9
be comes an echo-cancelled signal in the full band if each subband error signal is sufficiently small, that is to say, if the echo is sufficiently cancelled in each subband.
The tap assignment control circuit
10
receives a coefficient value
100
from adaptive filters
60
i
(i=1,2, . . . , K) of each subband, the number of taps
101
at a time (m−1)UT and a subband input signal
102
, calculates a signal
103
representing the number of taps of each subband, and transports it to an adaptive filter
60
i
of each subband. Here, T is a sampling period.
In
FIG. 18
, an example of the tap assignment control circuit
10
is shown. A coefficient value
100
is supplied from adaptive filters
60
i
(i=1,2, . . . , K) of each subband to the coefficient square value evaluation circuit
110
. Assuming the coefficient vector C
i,k
of the adaptive filter
60
i
at a time kT, each of coefficients C
ij,k
(j=1,2, . . . , N) is determined by
c
i,k
=[c
i,1,k
c
i,2,k
. . . c
i,N,k
]
T
(1)
where [.]
T
denotes vector transpos of [.].
In the coefficient square value evaluation circuit
110
, the following value,
{overscore (c)}
i,k
=[c
i,N
i
−P+1,k
c
i,N
i
−P+2,k
. . . c
i,N
i
,k]
T
(2)
is first calculated corresponding to each subband. Here, P is a positive integer, N
i,k
is the number of taps of an adaptive filter of the i-th subband at time kT. That is to say, coefficients from the tail P tap portion is utilized for the evaluation in each subband. Moreover, the coefficient square value evaluation circuit
110
calculates as follows,
c
~
k
=
[
c
_
1
,
k
T
c
_
1
,
k
c
_
2
,
k
T
c
_
2
,
k
…
c
_
K
,
k
T
c
_
K
,
k
]
(
3
)
where the result is transported to the calculation circuit of number of taps
130
as a tsubband tail coefficient vector
111
. The input signal
102
of each adaptive filter in addition to an output
111
of the coefficient square value evaluation circuit
110
is supplied to the.
Now, the input signal power vector V
k
at time kT are determined by
V
k
=
(
[
χ
1
,
k
2
χ
2
,
k
2
…
χ
K
,
k
2
]
)
T
(
4
)
using an input signal &khgr;
i,k
(i=1,2, . . . , K) for the i-th. The calculation circuit of number of taps
130
calculates the number of taps of each adaptive filt
Dickstein , Shapiro, Morin & Oshinsky, LLP
Pham Chi
Phu Phuong
LandOfFree
Unknown system identification method by subband adaptive... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Unknown system identification method by subband adaptive..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Unknown system identification method by subband adaptive... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2822656