Beam Control Using the Parametric Adaptive Matched Filter STAP Approach
James
H.
Michelsl, Jaime
R.
Roman2,
and
Braham Himed’
’
Air Force Research Laboratory
Sensors Directorate
26
Electronic Parkway
Rome, New York
13441-4514
James.Michels~.rl.af.mil,
braham.himed@,rl.af.mil
Abstract
-
This paper evaluates the performance
of
B
vector autoregressive space-time adaptive processing
method, the parametric adaptive matched filter (PAMF),
in
the presence of airborne
radar
clutter and white noise
interference. Performance metrics are derived from a
PAMF weight vector representation and include the
detection test statistic, the signal-to-interference-plus-noise
ratio, and the adapted angle-Doppler beam pattern.
Results are shown for the latter metric in the presence
of
mainbeam and closely spaced sidelobe interference.
I. INTRODUCTION
This paper considers the weight vector representation
for the space-time adaptive processing (STAP) method
referred to as the parametric adaptive matched filter
(PAMF)
[I-31.
This representation can be utilized to
establish performance parameters and measures for the
PAMF in the same context as for the optimal matched
filter (MF). These include the detection test statistic, the
signal-to-interference-plus-noise ratio (SINR), and the
adapted angle-Doppler beam pattem response. The
weight vector representation can be utilized to generate
two-dimensional (2-D) adapted pattern frequency-
domain plots that can be compared with the optimal
adapted pattern plot.
Also,
weight error criteria can be
applied to evaluate the error in the PAMF weights (in
relation to the optimal weight vector). Finally, with
appropriate interpretation, analytical results derived for
the adaptive matched filter (AMF)
[4-61
may he carried
over to the PAMF. Such results cover technical issues
such as constant false alarm rate (CFAR) and probability
density function (PDF)
for
the test statistic.
The results presented herein are applicable to the
PAMF with moving-average (MA) whitening filter.
An
MA filter is the system inverse for an auto-regressive
(AR) system, thus vector AR model identification
algorithms are considered here.
11.
DETECTION
PROBLEM AND
MATCHED
FILTER
SOLUTION
We consider an airborne surveillance phased array
radar system application of space-time adaptive
processing (STAP) for ground clutter, broadband
Scientific Studies Corporation
2250 Quail Ridge
Palm Beach Gardens, FL
33418
jrromanssc@aol.com
interference cancellation and moving target detection.
Let {x(n)
1
n
=
0,
1,
...
,
N-I},
with x(n)
E
C’, be the
complex-valued sequence at the output of a linear,
equi-spaced, J-element array, after demodulation and
sampling, corresponding to the re” from a single
range cell over an N-pulse coherent processing interval
(CPI).
An
equivalent array output representation
is
the
block vector
x
E
C”, obtained as the concatenation of
the
N
array output vectors. The array output model
is
x=ae+d
(1)
where a
is
the fixed but unknown target amplitude, e
E
C”
is
the spatio-temporal target block steering vector,
and
d
E
C” is the disturbance block vector. The
disturbance process
d
consists of narrowband clutter,
broadband interference, and receiver white noise, and
is distributed as
CN(O,Rd),
where Rd=
E[ddH]
E
IS
the disturbance block covariance matrix.
Block vectors e and
d
also admit a vector sequence
representation; namely, (e(n)
E
C’I
n
=
0,
1,
...
,
N-I}
and (d(n)
E
C’
I
n
=
0,
I,
...
,
N-I], respectively.
The detection problem
is
formulated as a dual
hypothesis testing problem, wherein the objective
is
to
select between the target-ahsent and target-present
conditions. The target-absent condition (a=O)
is
referred to as the null hypothesis,
Ho,
whereas, the
target present condition
(a
#
0)
is
referred to as the
alternative hypothesis,
H
.
This problem
is
solved via
the maximum likelihood approach, assuming the
disturbance block covariance matrix is known. The
solution is a detection test statistic that
is
compared
with a threshold for hypothesis selection. The
conventional optimal solution for the known
covariance case is the linear matched filter (MF), as
formulated independently in
[4-61.
The MF detection
test statistic is of the form
c”dN
.
0-7803-7920-9/03/$17.00
@2003
IEEE
405
2003
IEEE
Radar Conference