Data processing: measuring – calibrating – or testing – Measurement system – Measured signal processing
Reexamination Certificate
1999-09-07
2002-07-02
Bui, Bryan (Department: 2857)
Data processing: measuring, calibrating, or testing
Measurement system
Measured signal processing
C702S190000, C702S197000, C708S322000, C379S406080, C370S290000
Reexamination Certificate
active
06415247
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates to an on-line system identification method used in adaptive echo canceller, adaptive noise canceller or the like, and more particularly to high speed processing and sequential updating processing of blind system.
BACKGROUND OF THE INVENTION
The meanings of symbols used in the invention are shown in Table 1.
TABLE 1
Symbols used in formulas are meant as follows.
Symbol
Meaning
x(n)
Input signal at time n
y
1
(n)
Output from unknown system 1 at time n
y
2
(n)
Output from unknown system 2 at time n
h
1
Impulse response matrix of unknown system 1
h
2
Impulse response matrix of unknown system 2
f
1
Impulse response matrix identifying unknown system 1
f
2
Impulse response matrix identifying unknown system 2
Convolution
L
1
Tap length of unknown system 1
L
2
Tap length of unknown system 2
r
y1y2
(&tgr;)
Cross correlation function
r
y1y1
(&tgr;)
Auto correlation function
e
1
(n)
Error function 1 at time n
e
2
(n)
Error function 2 at time n
J
Evaluation function
Ĵ
Evaluation function defined newly by excluding expected value
of evaluation function J
&lgr;
Forgetfulness coefficient
&mgr;
Step coefficient
I
Unit matrix
c
Arbitrary integer
t
Transposition
E[ ]
Operation of expected value
A conventional system identification method is described while referring to FIG.
8
.
FIG. 8
is a block diagram showing a conventional system identification device. In
FIG. 8
, from one input line
501
, a digital signal x (n) is transmitted through two unknown systems
502
,
504
, and digital signals y
1
(n) and y
2
(n) are output to output lines
506
,
509
. A first system identification device
503
is connected parallel to the first unknown system
502
, and a second system identification device
505
, to the second unknown system
504
. To these system identification devices, a digital signal x (n) is entered same as in the unknown systems, and corresponding signals (formula 24) and (formula 25) are output to the output lines
508
,
511
.
ŷ
1
(
n
) [Formula 24]
ŷ
2
(
n
) [Formula 25]
The digital signal y
1
(n) output from the first unknown system
502
and the digital signal (formula 24) output from the first identification device
503
are fed into an adder
507
. The digital signal y
2
(n) output from the second unknown system
504
and the digital signal (formula 25) output from the second identification device
505
are fed into an adder
510
. At this time, it is supposed that the system identification device
503
and system identification device
505
are identified by a representative adaptive algorithm, such as least mean square (LMS) or recursive least square (RLS). The system identification by adaptive algorithm is realized by following calculation. The adder
507
calculates error e
1
(n) between the output digital signal y
1
(n) of the first unknown system
502
and the output digital signal (formula 24) of the system identification device
503
, feeding back the error, and changing the filter coefficient of the first system identification device
503
so that the error e
1
(n) may be closer to zero. This is the same in the second system identification device
505
. Thus, in the prior art, when identifying the system, the system identification device required the same input digital signal as in the unknown system.
Now suppose to identify an unknown system by employing a method proposed in a paper disclosed in “A Least-Squares Approach to Blind Channel Identification” (IEEE Transactions on Signal Processing Vol. 43, No. 12, 1995, pp. 2982-2993) (hereinafter called reference 1). This is explained by reference to FIG.
1
.
FIG. 1
is a block diagram showing a general system identification method shown in reference 1. In
FIG. 1
, first unknown systems
102
~m-th unknown systems
104
show m different unknown systems, or m different unknown systems spuriously decomposing one unknown system. In
FIG. 1
, m unknown systems are shown, but the generality is not lost if the number of unknown systems is limited to two, and therefore, in the following explanation, the number of unknown systems is limited to two for the sake of simplicity.
In
FIG. 1
, from one input line
101
, a digital signal x (n) is transmitted through two unknown systems
102
and
103
, and they output digital signals y
1
(n) and y
2
(n) to output lines
105
and
106
. According to reference 1, in a system identification device
108
, using only these two input digital signals y
1
(n), y
2
(n), it is possible to identify the first unknown system
102
and second unknown system
103
. It means that only the outputs of the unknown systems are used as the input to the system identification device
108
to be identified, and the input digital signal x (n) to the unknown systems is not necessary. Besides, “Equalization Based on Blind System Identification Using Second Order Statistics (a paper included in pre-print A-4-22 of General Assembly of Japan Society of Electronic Information and Communication; hereinafter called reference 2) can be also used for building up a system identification device same as reference 1. The both are known as the blind system.
In reference 1 and reference 2, however, the method of identifying the unknown system by the digital signal output from the unknown system, that is, only the off-line processing is mentioned, and the system identification method in reference 1 or reference 2 cannot be used in on-line processing for updating sequentially while operating the object unknown system, and hence it cannot be used in the adaptive echo canceller requiring real-time processing. Another problem is that the formula development mentioned in reference 1 or reference 2 is not a formula development in which sequential updating process can be introduced.
In the case of an off-line processing, if the characteristic of the unknown system is changed due to some reason (for example, a change in time), a different value from the intended unknown system is identified. To identify according to the method of reference 1 or reference 2, it is necessary to calculate the inverse matrix or eigenvalue, and the quantity of calculation becomes tremendous. In the system identification device and system identification method for use in adaptive echo canceller or the like, and in the recording medium in which the execution program of identification method is recorded, by formula development different from reference 1 or reference 2, that is, by formula development for updating sequentially and not performing operation of inverse matrix, it is required to perform sequential updating process and prevent increase to tremendous quantity of calculation.
SUMMARY OF THE INVENTION
It is hence an object of the invention to present a system identification method capable of performing sequential updating process and preventing increase to tremendous quantity of calculation.
To achieve the object, the system identification method of the invention is a system identification method for identifying an unknown system by feeding output digital signal y
1
(n) and output digital signal y
2
(n) of the unknown system, comprising a first delay step for delaying the input digital signal y
2
(n) by one unit time, a formula 2 matrix generating step for generating an input matrix (formula 2) shown in formula 1 from the input digital signal y
1
(n) and the digital signal y
2
(n−1) output at the first delay step, a formula 13 matrix calculating step for calculating the state matrix (formula 13) at the present time shown in formula 12 from the matrix output at the formula 2 matrix generating step and the state matrix (formula 11) calculated one unit time before, a formula 15 matrix calculating step for calculating the state matrix (formula 15) at the present time shown in formula 14 from the matrix output at the formula 2 matrix generating step, the matrix output at the formula 13 matrix calculating step, and the state matrix (formula 11) calculated one unit time before, a second delay ste
Kimura Takashi
Ochi Hiroshi
Sasaki Hideaki
Bui Bryan
Matsushita Electric - Industrial Co., Ltd.
Ratner & Prestia
LandOfFree
System identification method and its recording device does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System identification method and its recording device, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System identification method and its recording device will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2895854