Modified affine projection algorithm for non-stationary signal
View Patent ↗An affine projection or like algorithm is used to process a non-stationary signal. The affine projection algorithm creates an inverse matrix and includes a factor γ to avoid numerical instability in the inverse matrix. The factor γ is adaptively adjusted according to the characteristics of the signal.
1. A method of processing a non-stationary signal comprising:
applying an adaptive echo cancellation algorithm that creates an inverse matrix to said non-stationary signal; and
adaptively adjusting the condition of said matrix according to the characteristics of said signal;
wherein said adaptive echo cancellation algorithm is an affine projection algorithm including a factor γ provided to avoid numerical instability in said inverse matrix;
wherein said factor γ is adaptively adjusted in accordance with the characteristics of said signal; and
wherein said factor γ is set to have a value that is a predetermined amount below the average power of the last N set of signal components or the estimated power, where N is the order of projection of the algorithm.
2. A method as claimed in claim 1 , wherein said algorithm is a proportionate normalized least-mean-squares (LMS) algorithm.
3. A method as claimed in claim 1 , wherein said fixed value is 30 dB.
4. A method as claimed in claim 3 , wherein said factor γ is defined by the relationship:
Υ
=
max
(
1
1000
N
∑
diag
(
X
_
T
(
k
)
X
_
(
k
)
)
,
c
)
.
5. A method as claimed in claim 1 , wherein said factory γ is chosen to have a fixed value below the estimated signal power.
6. A method as claimed in claim 1 , wherein said factor γ is defined by the relationship:
Υ
=
max
(
1
1000
R
(
0
)
,
1
4000
∑
L
i
=
0
x
(
k
-
i
)
x
(
k
-
i
)
,
c
)
.
7. Signal processing apparatus for processing a non-stationary signal comprising:
an adaptive filter employing an echo cancellation algorithm that creates an inverse matrix to said non-stationary signal; and
a control unit for adaptively adjusting the condition of said matrix according to the characteristics of said signal, and
wherein said adaptive echo cancellation algorithm is an affine projection algorithm including a factor γ provided to avoid numerical instability in said inverse matrix;
wherein said factor γ is adaptively adjusted in accordance with the characteristics of said signal; and
wherein said factor γ is set to have a value that is a predetermined amount below the average power of the last N set of signal components or the estimated power, where N is the order of projection of the algorithm.
8. Signal processing apparatus as claimed in claim 7 , wherein said algorithm is a proportionate normalized least-mean-squares (LMS) algorithm.
9. Signal processing apparatus as claimed in claim 7 , wherein said factor γ is defined by the relationship:
Υ
=
max
(
1
1000
R
(
0
)
,
1
4000
∑
i
=
0
L
x
(
k
-
i
)
x
(
k
-
i
)
,
c
)
.
10. Signal processing apparatus as claimed in claim 7 , wherein said factor γ is defined by the relationship:
Υ
=
max
(
1
1000
R
(
0
)
,
1
4000
∑
i
=
0
L
x
(
k
-
i
)
x
(
k
-
i
)
,
c
)
.
11. Signal processing apparatus as claimed in claim 7 , comprising a plurality of delay units, adders and multipliers, and a maximization unit to produce an output representing γ that is the maximum of two inputs, a first said input derived from said plurality of delay units, adders, and multipliers, and a second said input representing a constant.
12. Signal processing apparatus as claimed in claim 7 , further comprising:
a first input for receiving a first input signal x(k);
a first delay unit for receiving said first input signal;
a second delay unit for receiving an output of said first delay unit added to the sum of said input signal and an output of said second delay unit; and
a maximization unit have first and second inputs, said first input of said maximization unit receiving a constant value and said second input of said maximization unit receiving an input from the input to said second delay unit.