 |
Claims  |
|
|
What is claimed is:
1. A method of identifying a system with an adaptive filter, comprising the
steps of:
calculating an error signal by subtracting an output signal produced by
processing a reference input signal with an adaptive filter from an
observed signal composed of a mixture of an output signal from an unknown
system and noise;
estimating characteristics of the unknown system by correcting coefficients
of said adaptive filter in order to minimize the error signal using at
least said error signal, said reference input signal, and a step size; and
producing a value as said step size by estimating the power of said
reference input signal and processing the estimated power of said
reference input signal according to a function which has a maximum value
when the power of said reference input signal is equal to a first
threshold, monotonically increases when the power of said reference input
signal is smaller than said first threshold, and monotonically decreases
when the power of said reference input signal is greater than said first
threshold,
further comprising the steps of
estimating the level of noise mixed in the observed signal using said error
signal and at least one of an output signal from said adaptive filter and
said reference input signal, and
controlling said first threshold depending on the level of noise, thereby
determining said first threshold.
2. A method of identifying a system with an adaptive filter, comprising the
steps of:
calculating an error signal by subtracting an output signal produced by
processing a reference input signal with an adaptive filter from an
observed signal composed of a mixture of an output signal from an unknown
system and noise;
estimating characteristics of the unknown system by correcting coefficients
of said adaptive filter in order to minimize the error signal using at
least said error signal, said reference input signal, and a step size; and
producing a value as said step size by estimating the power of said
reference input signal and processing the estimated power of said
reference input signal according to a function which has a maximum value
when the power of said reference input signal is equal to a first
threshold, monotonically increases when the power of said reference input
signal is smaller than said first threshold, and monotonically decreases
when the power of said reference input signal is greater than said first
threshold,
further comprising the steps of
estimating the level of noise mixed in the observed signal using said error
signal and at least one of an output signal from said adaptive filter and
said reference input signal,
estimating the gain of said unknown system using the coefficients of said
adaptive filter, and
controlling said first threshold depending on the level of noise and the
gain of said unknown system, thereby determining said first threshold.
3. A method of identifying a system with an adaptive filter, comprising the
steps of:
calculating an error signal by subtracting an output signal produced by
processing a reference input signal with an adaptive filter from an
observed signal composed of a mixture of an output signal from an unknown
system and noise;
estimating characteristics of the unknown system by correcting coefficients
of said adaptive filter in order to minimize the error signal using at
least said error signal, said reference input signal, and a step size; and
producing a value as said step size by estimating the power of said
reference input signal and processing the estimated power of said
reference input signal according to a function which has a maximum value
when the power of said reference input signal is equal to a first
threshold, monotonically increases when the power of said reference input
signal is smaller than said first threshold, and monotonically decreases
when the power of said reference input signal is greater than said first
threshold,
further comprising the steps of
estimating the level of noise mixed in the observed signal using said error
signal and at least one of an output signal from said adaptive filter and
said reference input signal, and
controlling said first threshold and a maximum value of said step size
depending on the level of noise, thereby determining said first threshold
and the maximum value of said step size.
4. A method of identifying a system with an adaptive filter, comprising the
steps of:
calculating an error signal by subtracting an output signal produced by
processing a reference input signal with an adaptive filter from an
observed signal composed of a mixture of an output signal from an unknown
system and noise;
estimating characteristics of the unknown system by correcting coefficients
of said adaptive filter in order to minimize the error signal using at
least said error signal, said reference input signal, and a step size; and
producing a value as said step size by estimating the power of said
reference input signal and processing the estimated power of said
reference input signal according to a function which has a maximum value
when the power of said reference input signal is equal to a first
threshold, monotonically increases when the power of said reference input
signal is smaller than said first threshold, and monotonically decreases
when the power of said reference input signal is greater than said first
threshold,
further comprising the steps of
estimating the level of noise mixed in the observed signal using said error
signal and at least one of an output signal from said adaptive filter and
said reference input signal,
estimating the gain of said unknown system using the coefficients of said
adaptive filter, and
controlling said first threshold and a maximum value of said step size
depending on the level of noise and the gain of said unknown system,
thereby determining said first threshold and the maximum value of said
step size.
5. A method according to claim 1, further comprising the steps of updating
an estimated value of the level of noise only when the output signal from
said adaptive filter has a level smaller than a second threshold, and
otherwise holding a preceding estimated value of the level of noise,
thereby estimating said level of noise.
6. A method according to claim 2, further comprising the steps of updating
an estimated value of the level of noise only when the output signal from
said adaptive filter has a level smaller than a second threshold, and
otherwise holding a preceding estimated value of the level of noise,
thereby estimating said level of noise.
7. A method according to claim 3, further comprising the steps of updating
an estimated value of the level of noise only when the output signal from
said adaptive filter has a level smaller than a second threshold, and
otherwise holding a preceding estimated value of the level of noise,
thereby estimating said level of noise.
8. A method according to claim 4, further comprising the steps of updating
an estimated value of the level of noise only when the output signal from
said adaptive filter has a level smaller than a second threshold, and
otherwise holding a preceding estimated value of the level of noise,
thereby estimating said level of noise.
9. A method according to claim 5, further comprising the step of
controlling said second threshold depending on the level of said error
signal, thereby estimating said level of noise.
10. A method according to claim 6, further comprising the step of
controlling said second threshold depending on the level of said error
signal, thereby estimating said level of noise.
11. A method according to claim 7, further comprising the step of
controlling said second threshold depending on the level of said error
signal, thereby estimating said level of noise.
12. A method according to claim 8, further comprising the step of
controlling said second threshold depending on the level of said error
signal, thereby estimating said level of noise.
13. A method according to claim 1, further comprising the steps of updating
an estimated value of the level of noise only when said reference input
signal has a level smaller than a third threshold, and otherwise holding a
preceding estimated value of the level of noise, thereby estimating said
level of noise.
14. A method according to claim 2, further comprising the steps of updating
an estimated value of the level of noise only when said reference input
signal has a level smaller than a third threshold, and otherwise holding a
preceding estimated value of the level of noise, thereby estimating said
level of noise.
15. A method according to claim 3, further comprising the steps of updating
an estimated value of the level of noise only when said reference input
signal has a level smaller than a third threshold, and otherwise holding a
preceding estimated value of the level of noise, thereby estimating said
level of noise.
16. A method according to claim 4, further comprising the steps of updating
an estimated value of the level of noise only when said reference input
signal has a level smaller than a third threshold, and otherwise holding a
preceding estimated value of the level of noise, thereby estimating said
level of noise.
17. A method according to claim 13, further comprising the step of
controlling said third threshold depending on the level of said error
signal, thereby estimating said level of noise.
18. A method according to claim 14, further comprising the step of
controlling said third threshold depending on the level of said error
signal, thereby estimating said level of noise.
19. A method according to claim 15, further comprising the step of
controlling said third threshold depending on the level of said error
signal, thereby estimating said level of noise.
20. A method according to claim 16, further comprising the step of
controlling said third threshold depending on the level of said error
signal, thereby estimating said level of noise.
21. A method according to claim 1, further comprising the steps of adding
the product of an estimated value of the level of noise and a first
coefficient and the product of the level of the error signal and a second
coefficient thereby to produce a new estimated value of the level of
noise, and controlling said first coefficient and said second coefficient
depending on the level of said error signal and the level of at least one
of the output signal from said adaptive filter or said reference input
signal, thereby estimating the level of noise.
22. A method according to claim 2, further comprising the steps of adding
the product of an estimated value of the level of noise and a first
coefficient and the product of the level of the error signal and a second
coefficient thereby to produce a new estimated value of the level of
noise, and controlling said first coefficient and said second coefficient
depending on the level of said error signal and the level of at least one
of the output signal from said adaptive filter or said reference input
signal, thereby estimating the level of noise.
23. A method according to claim 3, further comprising the steps of adding
the product of an estimated value of the level of noise and a first
coefficient and the product of the level of the error signal and a second
coefficient thereby to produce a new estimated value of the level of
noise, and controlling said first coefficient and said second coefficient
depending on the level of said error signal and the level of at least one
of the output signal from said adaptive filter or said reference input
signal, thereby estimating the level of noise.
24. A method according to claim 4, further comprising the steps of adding
the product of an estimated value of the level of noise and a first
coefficient and the product of the level of the error signal and a second
coefficient thereby to produce a new estimated value of the level of
noise, and controlling said first coefficient and said second coefficient
depending on the level of said error signal and the level of at least one
of the output signal from said adaptive filter or said reference input
signal, thereby estimating the level of noise.
25. An apparatus for identifying a system by calculating an error signal by
subtracting an output signal produced by processing a reference input
signal with an adaptive filter from an observed signal composed of a
mixture of an output signal from the unknown system and noise, and
estimating characteristics of the unknown system by correcting
coefficients of said adaptive filter based on said error signal,
comprising:
an adaptive filter for producing an output signal using at least the
reference input signal and filter coefficients thereof;
a subtractor for calculating an error signal by subtracting the output
signal of the adaptive filter from an observed signal composed of a
mixture of an output signal from the unknown system and noise, and
supplying the error signal to said adaptive filter;
a power estimating circuit for estimating the power of the reference input
signal;
a noise level estimating circuit for estimating the level of noise mixed in
the observed signal using said error signal and at least one of the output
signal from said adaptive filter or said reference input signal;
a threshold setting circuit for calculating a first threshold based on the
estimated level of noise from said noise level estimating circuit; and
a step size determining circuit for calculating a step size based on the
estimated power from said power estimating circuit and said first
threshold from said threshold setting circuit, and supplying the
calculated step size to said adaptive filter;
the arrangement being such that said step size determining circuit
generates a step size according to a function of the power of the
reference input signal which monotonously increases if the power of the
reference input signal is smaller than the first threshold and
monotonously decreases if the power of the reference input signal is
greater than the first threshold, and said adaptive filter corrects the
filter coefficients thereof in order to minimize said error signal using
at least said error signal, said reference input signal, and said step
size, for thereby estimating the characteristics of the unknown system.
26. An apparatus for identifying a system by calculating an error signal by
subtracting an output signal produced by processing a reference input
signal with an adaptive filter from an observed signal composed of a
mixture of an output signal from the unknown system and noise, and
estimating characteristics of the unknown system by correcting
coefficients of said adaptive filter based on said error signal,
comprising:
an adaptive filter for producing an output signal using at least the
reference input signal and filter coefficients thereof;
a subtractor for calculating an error signal by subtracting the output
signal of the adaptive filter from an observed signal composed of a
mixture of an output signal from the unknown system and noise, and
supplying the error signal to said adaptive filter;
a power estimating circuit for estimating the power of the reference input
signal;
a noise level estimating circuit for estimating the level of noise mixed in
the observed signal using said error signal and at least one of the output
signals from said adaptive filter and said reference input signal;
a gain calculating circuit for estimating the gain of the unknown system
using the filter coefficient of said adaptive filter;
a threshold setting circuit for calculating a first threshold based on the
estimated level of noise from said noise level estimating circuit and the
estimated gain from said gain calculating circuit; and
a step size determining circuit for calculating a step size based on the
estimated power from said power estimating circuit and said first
threshold from said threshold setting circuit, and supplying the
calculated step size to said adaptive filter;
the arrangement being such that said step size determining circuit
generates a step size according to a function of the power of the
reference input signal which monotonously increases if the power of the
reference input signal is smaller than the first threshold and
monotonously decreases if the power of the reference input signal is
greater than the first threshold, and said adaptive filter corrects the
filter coefficients thereof in order to minimize said error signal using
at least said error signal, said reference input signal, and said step
size, for thereby estimating the characteristics of the unknown system.
27. An apparatus for identifying a system by calculating an error signal by
subtracting an output signal produced by processing a reference input
signal with an adaptive filter from an observed signal composed of a
mixture of an output signal from the unknown system and noise, and
estimating characteristics of the unknown system by correcting
coefficients of said adaptive filter based on said error signal,
comprising:
an adaptive filter for producing an output signal using at least the
reference input signal and filter coefficients thereof;
a subtractor for calculating an error signal by subtracting the output
signal of the adaptive filter from an observed signal composed of a
mixture of an output signal from the unknown system and noise, and
supplying the error signal to said adaptive filter;
a power estimating circuit for estimating the power of the reference input
signal;
a noise level estimating circuit for estimating the level of noise mixed in
the observed signal using said error signal and at least one of the output
signal from said adaptive filter or said reference input signal; and
a step size determining circuit for calculating a step size based on the
estimated power from said power estimating circuit and the estimated level
of noise from said noise level estimating circuit, and supplying the
calculated step size to said adaptive filter;
the arrangement being such that said step size determining circuit
establishes a first threshold and a maximum value of the step size based
on the estimated level of noise, generates a step size according to a
function of the power of the reference input signal which monotonously
increases if the power of the reference input signal is smaller than said
first threshold and monotonously decreases if the power of the reference
input signal is greater than the first threshold, and said adaptive filter
corrects the filter coefficients thereof in order to minimize said error
signal using at least said error signal, said reference input signal, and
said step size, for thereby estimating the characteristics of the unknown
system.
28. An apparatus for identifying a system by calculating an error signal by
subtracting an output signal produced by processing a reference input
signal with an adaptive filter from an observed signal composed of a
mixture of an output signal from the unknown system and noise, and
estimating characteristics of the unknown system by correcting
coefficients of said adaptive filter based on said error signal,
comprising:
an adaptive filter for producing an output signal using at least the
reference input signal and a filter coefficient thereof;
a subtractor for calculating an error signal by subtracting the output
signal of the adaptive filter from an observed signal composed of a
mixture of an output signal from the unknown system and noise, and
supplying the error signal to said adaptive filter;
a power estimating circuit for estimating the power of the reference input
signal;
a noise level estimating circuit for estimating the level of noise mixed in
the observed signal using said error signal and at least one of the output
signal from said adaptive filter or said reference input signal;
a gain calculating circuit for estimating the gain of the unknown system
using the filter coefficient of said adaptive filter; and
a step size determining circuit for calculating a step size based on the
estimated power from said power estimating circuit, the estimated level of
noise from said noise level estimating circuit, and the estimated gain
from said gain estimating circuit, and supplying the calculated step size
to said adaptive filter;
the arrangement being such that said step size determining circuit
establishes a first threshold and a maximum value of the step size based
on the estimated level of noise and the estimated gain of the unknown
system, generates a step size according to a function of the power of the
reference input signal which monotonously increases if the power of the
reference input signal is smaller than said first threshold and
monotonously decreases if the power of the reference input signal is
greater than the first threshold, and said adaptive filter corrects the
filter coefficients thereof in order to minimize said error signal using
at least said error signal, said reference input signal, and said step
size, for thereby estimating the characteristics of the unknown system.
29. An apparatus according to claim 25, wherein said noise level estimating
circuit comprises:
an output level estimating circuit for estimating the level of the output
signal from said adaptive filter;
a register for storing a second threshold;
a comparator for comparing the estimated level of the output signal from
said adaptive filter with said second threshold; and
a noise level calculating circuit for updating the estimated level of noise
using said error signal only if said comparator determines that the
estimated level of the output signal from said adaptive filter is smaller
than said second threshold.
30. An apparatus according to claim 26, wherein said noise level estimating
circuit comprises:
an output level estimating circuit for estimating the level of the output
signal from said adaptive filter;
a register for storing a second threshold;
a comparator for comparing the estimated level of the output signal from
said adaptive filter with said second threshold; and
a noise level calculating circuit for updating the estimated level of noise
using said error signal only if said comparator determines that the
estimated level of the output signal from said adaptive filter is smaller
than said second threshold.
31. An apparatus according to claim 27, wherein said noise level estimating
circuit comprises:
an output level estimating circuit for estimating the level of the output
signal from said adaptive filter;
a register for storing a second threshold;
a comparator for comparing the estimated level of the output signal from
said adaptive filter with said second threshold; and
a noise level calculating circuit for updating the estimated level of noise
using said error signal only if said comparator determines that the
estimated level of the output signal from said adaptive filter is smaller
than said second threshold.
32. An apparatus according to claim 28, wherein said noise level estimating
circuit comprises:
an output level estimating circuit for estimating the level of the output
signal from said adaptive filter;
a register for storing a second threshold;
a comparator for comparing the estimated level of the output signal from
said adaptive filter with said second threshold; and
a noise level calculating circuit for updating the estimated level of noise
using said error signal only if said comparator determines that the
estimated level of the output signal from said adaptive filter is smaller
than said second threshold.
33. An apparatus according to claim 29, wherein said register is replaced
with a threshold generator for generating a threshold depending on said
error signal which is supplied as an input signal thereto.
34. An apparatus according to claim 30, wherein said register is replaced
with a threshold generator for generating a threshold depending on said
error signal which is supplied as an input signal thereto.
35. An apparatus according to claim 31, wherein said register is replaced
with a threshold generator for generating a threshold depending on said
error signal which is supplied as an input signal thereto.
36. An apparatus according to claim 32, wherein said register is replaced
with a threshold generator for generating a threshold depending on said
error signal which is supplied as an input signal thereto.
37. An apparatus according to claim 29, wherein said output level
estimating circuit is replaced with a reference input signal level
estimating circuit for estimating the level of the reference input signal,
and said register stores the third threshold.
38. An apparatus according to claim 30, wherein said output level
estimating circuit is replaced with a reference input signal level
estimating circuit for estimating the level of the reference input signal,
and said register stores the third threshold.
39. An apparatus according to claim 31, wherein said output level
estimating circuit is replaced with a reference input signal level
estimating circuit for estimating the level of the reference input signal,
and said register stores the third threshold.
40. An apparatus according to claim 32, wherein said output level
estimating circuit is replaced with a reference input signal level
estimating circuit for estimating the level of the reference input signal,
and said register stores the third threshold.
41. An apparatus according to claim 37, wherein said register is replaced
with a threshold generator for generating a threshold depending on said
error signal which is supplied as an input signal thereto.
42. An apparatus according to claim 38, wherein said register is replaced
with a threshold generator for generating a threshold depending on said
error signal which is supplied as an input signal thereto.
43. An apparatus according to claim 39, wherein said register is replaced
with a threshold generator for generating a threshold depending on said
error signal which is supplied as an input signal thereto.
44. An apparatus according to claim 40, wherein said register is replaced
with a threshold generator for generating a threshold depending on said
error signal which is supplied as an input signal thereto.
45. An apparatus according to claim 25, wherein said noise level estimating
circuit comprises:
a register for storing the estimated level of noise;
a first multiplier for multiplying the estimated level of noise stored in
said register by a first coefficient;
a nonlinear converter for nonlinearly converting said error signal into a
converted signal;
a second multiplier for multiplying the converted signal by a second
coefficient;
an adder for adding a product signal from said first multiplier and a
product signal from said second multiplier, and storing a resultant sum
signal in said register as the estimated level of noise; and
a coefficient generator for establishing the first and second coefficients
based on said error signal and at least one of the output signal from the
adaptive filter and the reference signal.
46. An apparatus according to claim 26, wherein said noise level estimating
circuit comprises:
a register for storing the estimated level of noise;
a first multiplier for multiplying the estimated level of noise stored in
said register by a first coefficient;
a nonlinear converter for nonlinearly converting said error signal into a
converted signal;
a second multiplier for multiplying the converted signal by a second
coefficient;
an adder for adding a product signal from said first multiplier and a
product signal from said second multiplier, and storing a resultant sum
signal in said register as the estimated level of noise; and
a coefficient generator for establishing the first and second coefficients
based on said error signal and at least one of the results of the output
signal from the adaptive filter and the reference signal.
47. An apparatus according to claim 27, wherein said noise level estimating
circuit comprises:
a register for storing the estimated level of noise;
a first multiplier for multiplying the estimated level of noise stored in
said register by a first coefficient;
a nonlinear converter for nonlinearly converting said error signal into a
converted signal;
a second multiplier for multiplying the converted signal by a second
coefficient;
an adder for adding a product signal from said first multiplier and a
product signal from said second multiplier, and storing a resultant sum
signal in said register as the estimated level of noise; and
a coefficient generator for establishing the first and second coefficients
based on said error signal and at least one of the output signal from the
adaptive filter and the reference signal.
48. An apparatus according to claim 28, wherein said noise level estimating
circuit comprises:
a register for storing the estimated level of noise;
a first multiplier for multiplying the estimated level of noise stored in
said register by a first coefficient;
a nonlinear converter for nonlinearly converting said error signal into a
converted signal;
a second multiplier for multiplying the converted signal by a second
coefficient;
an adder for adding a product signal from said first multiplier and a
product signal from said second multiplier, and storing a resultant sum
signal in said register as the estimated level of noise; and
a coefficient generator for establishing the first and second coefficients
based on said error signal and at least one of the output signal from the
adaptive filter and the reference signal.
49. A method of identifying a system with an adaptive filter, comprising
the steps of:
calculating an error signal by subtracting an output signal produced by
processing a reference input signal with an adaptive filter from an
observed signal composed of a mixture of an output signal from an unknown
system and noise;
estimating characteristics of the unknown system by correcting coefficients
of said adaptive filter in order to minimize the error signal using at
least said error signal, said reference input signal, and a step size; and
producing a value as said step size by estimating the power of said
reference input signal and processing the estimated power of said
reference input signal according to a function which has a maximum value
when the power of said reference input signal is equal to a first
threshold, has a direct proportional relationship with the power of said
reference input signal when the power of said reference input signal is
smaller than said first threshold, and has an inverse proportional
relationship with the power of said reference input signal when the power
of said reference input signal is greater than said first threshold.
50. An apparatus for identifying a system by calculating an error signal by
subtracting an output signal produced by processing a reference input
signal with an adaptive filter from an observed signal composed of a
mixture of an output signal from the unknown system and noise, and
estimating characteristics of the unknown system by correcting
coefficients of said adaptive filter based on said error signal,
comprising:
an adaptive filter for producing an output signal using at least the
reference input signal and filter coefficients thereof;
a subtractor for calculating an error signal by subtracting the output
signal of the adaptive filter from an observed signal composed of a
mixture of an output signal from the unknown system and noise, and
supplying the error signal to said adaptive filter;
a power estimating circuit for estimating the power of the reference input
signal; and
a step size determining circuit for calculating a step size based on the
estimated power from said power estimating circuit and supplying the
calculated step size to said adaptive filter;
the arrangement being such that said step size determining circuit
generates a step size according to a function of the power of the
reference input signal, wherein the step size has a direct proportional
relationship to the power of the reference input signal if the power of
the reference input signal is smaller than a first threshold and has an
inverse proportional relationship to the power of the reference input
signal if the power of the reference input signal is greater than the
first threshold, and said adaptive filter corrects the filter coefficients
thereof in order to minimize said error signal using at least said error
signal, said reference input signal, and said step size, for thereby
estimating the characteristics of the unknown system. |
|
 |
|
 |
Claims |
|
|
|
Description |
|
|
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a method of and an apparatus for
estimating characteristics of an unknown system using an adaptive filter
in an echo canceller active noise control, equalizer, line enhancer,
adaptive array, adaptive loudspeaker or a noise canceller.
2. Description of the Related Art
Transversal adaptive filters based on the learning identification method
described in IEEE transactions on automatic control, Vol. AC-12, No. 3,
pp. 282-287, 1967, USA (hereinafter referred to as Literature 1) are
widely used in methods of and apparatus for estimating characteristics of
an unknown system. The principles of operation of an acoustic echo
canceller incorporating a transversal adaptive filter based on the
learning identification method will be described below.
FIG. 1 of the accompanying drawings is a block diagram of an acoustic echo
canceller based on the learning identification method. A system
identification device is used as an echo canceller 100. A reference input
signal 1 is converted by a loudspeaker 10 into an acoustic signal which is
propagated through an acoustic path 11 as an unknown system and reaches,
as an acoustic echo, a microphone 12. The microphone 12 converts the
acoustic echo, with noise 2 added thereto, into an electric signal as an
observed signal 3. An adaptive filter 101 effects a convolutional
calculation on the reference input signal 1 and filter coefficients, and
supplies the result as an output signal 5 to a subtractor 102. The
subtractor 102 subtracts the output signal 5 from the observed signal 3,
and produces a resultant error signal 4 as an output signal from the echo
canceller 100, which is supplied to the adaptive filter 101. A power
estimating circuit 103 estimates the power of the reference input signal
1, and supplies the estimated power to a divider 115. The divider 115
divides a positive constant .mu..sub.0 stored in a register 114 by the
estimated power, and outputs the quotient as a step size 105. The adaptive
filter 101 updates the filter coefficients in order to minimize the error
signal 4, using the step size 105 supplied from the divider 115, the
reference input signal 1, and the error signal 4.
The above process is expressed by equations as follows: It is assumed that
the unknown system has an impulse response h.sub.i (i=0, . . . , N-1), the
reference input signal 1 at a time "t" is represented by x(t), the noise 2
at the time "t" by n(t), and the observed signal 3 at the time "t" by
d(t). The relationship between the reference input signal x(t), the noise
n(t), and the observed signal d(t) is given by:
##EQU1##
If the adaptive filter 101 has a tap number N and the filter coefficient
is represented by w.sub.i (t) (i=0. . . , N-1), then the adaptive filter
101 produces an output signal y(t) expressed by:
##EQU2##
From the equations (1) and (2), the error signal e(t) is indicated by:
##EQU3##
Using the step size .mu.(t), the filter coefficient w.sub.i (t) is updated
as follows:
w.sub.i (t+1)=w.sub.i (t)+.mu.(t)e(t)x(t-i) (4)
The step size .mu.(t) is given by:
##EQU4##
where p.sub.x (t) is the power of the reference input signal 1 which is
determined by the equation:
##EQU5##
and .mu..sub.0 is a constant in the range of:
0<.mu..sub.0 <2 (7)
The learning identification method updates the filter coefficient using the
error signal e(t).
It can be seen from the equation (3) that the error signal e(t) contains
the noise n(t) in addition to the system identification error h.sub.i
-w.sub.i (t). When the noise n(t) is sufficiently smaller than the output
signal of the unknown system, the filter coefficient can be updated
properly and the characteristics of the unknown system can be identified
according to the learning identification method. However, when the noise
n(t) is larger, the filter coefficient cannot be corrected properly.
Furthermore, if the reference input signal x(t) is a non-stationary signal
such as a speech signal, then the filter coefficient may not be updated
properly even when the noise n(t) is relatively small. The reasons for
this are considered to be as follows: Since the step size .mu.(t) is
inversely proportional to the power P.sub.x (t) of the reference input
signal x(t), the step size .mu.(t) is very large if the reference input
signal x(t) is very small. The output signal from the unknown system 11 is
very small, and the error signal e(t) contains larger noise n(t).
Therefore, the filter coefficient w.sub.i (t) is updated greatly using the
noise n(t) rather than the identification error h.sub.i -w.sub.i (t) with
respect to the unknown system. As a result, the filter coefficients cannot
be updated corrected properly.
As described above with reference to FIG. 1, the adaptive filter based on
the learning identification method cannot update the filter coefficient
properly when the noise is large.
SUMMARY OF THE INVENTION
It is therefore an object of the present invention to provide a method of
and an apparatus for identifying an unknown system using an adaptive
filter which can update the filter coefficient properly even in an
environment in which noise is large.
To achieve the above object of the present invention, there is provided a
first method of identifying a system with an adaptive filter, comprising
the steps of calculating an error signal by subtracting an output signal
produced by processing a reference input signal with an adaptive filter
from an observed signal composed of a mixture of an output signal from an
unknown system and noise, estimating characteristics of the unknown system
by correcting coefficients of the adaptive filter in order to minimize the
error signal using at least the error signal, the reference input signal,
and a step size, and producing a value as the step size by estimating the
power of the reference input signal and processing the estimated power of
the reference input signal according to a function which has a maximum
value when the power of the reference input signal is equal to a first
threshold, monotonously increases when the power of the reference input
signal is smaller than the first threshold, and monotonously decreases
when the power of the reference input signal is greater than the first
threshold.
As a second means of achieving the object of the present invention, the
first method further comprises the steps of estimating the level of noise
mixed in the observed signal using the error signal and at least one of an
output signal from the adaptive filter and the reference input signal, and
controlling the first threshold depending on the level of noise, thereby
determining the first threshold.
As a third means of achieving the object of the present invention, the
first method further comprises the steps of estimating the level of noise
mixed in the observed signal using the error signal and at least one of an
output signal from the adaptive filter and the reference input signal,
estimating the gain of the unknown system using the coefficient of the
adaptive filter, and controlling the first threshold depending on the
level of noise and the gain of the unknown system, thereby determining the
first threshold.
As a fourth means of achieving the object of the present invention, the
first method further comprises the steps of estimating the level of noise
mixed in the observed signal using the error signal and at least one of an
output signal from the adaptive filter and the reference input signal, and
controlling the first threshold and a maximum value of the step size
depending on the level of noise, thereby determining the first threshold
and the maximum value of the step size.
As a fifth means of achieving the object of the present invention, the
first method further comprises the steps of estimating the level of noise
mixed in the observed signal using the error signal and at least one of an
output signal from the adaptive filter and the reference input signal,
estimating the gain of the unknown system using the coefficient of the
adaptive filter, and controlling the first threshold and a maximum value
of the step size depending on the level of noise and the gain of the
unknown system, thereby determining the first threshold and the maximum
value of the step size.
As a sixth means of achieving the object of the present invention, the
methods of each of the second, third, fourth, and fifth means further
comprise the steps of updating an estimated value of the level of noise
only when the output signal from the adaptive filter has a level smaller
than a second threshold, and otherwise holding a preceding estimated value
of the level of noise, thereby estimating the level of noise.
As a seventh means of achieving the object of the present invention, the
sixth method further comprises the step of controlling the second
threshold depending on the level of the error signal, thereby estimating
the level of noise.
As an eighth means of achieving the object of the present invention, the
method of each of the second, third, fourth, and fifth inventions further
comprise the steps of updating an estimated value of the level of noise
only when the reference input signal has a level smaller than a third
threshold, and otherwise holding a preceding estimated value of the level
of noise, thereby estimating the level of noise.
As a ninth means of achieving the object of the present invention, the
eighth method further comprises the step of controlling the third
threshold depending on the level of the error signal, thereby estimating
the level of noise.
As a tenth means of achieving the object of the present invention, the
method of each of the second, third, fourth, and fifth inventions further
comprise the steps of adding the product of an estimated value of the
level of noise and a first coefficient and the product of the level of the
error signal and a second coefficient thereby to produce a new estimated
value of the level of noise, and controlling the first coefficient and the
second coefficient depending on the level of the error signal and the
level of at least one of the output signal from the adaptive filter and
the reference input signal, thereby estimating the level of noise.
As an eleventh means of achieving the object of the present invention,
there is provided an apparatus for identifying a system by calculating an
error signal by subtracting an output signal produced by processing a
reference input signal with an adaptive filter from an observed signal
composed of a mixture of an output signal from the unknown system and
noise, and estimating characteristics of the unknown system by correcting
coefficients of the adaptive filter based on the error signal, comprising
an adaptive filter for producing an output signal using at least the
reference input signal and a filter coefficient thereof, a subtractor for
calculating an error signal by subtracting the output signal of the
| | |
