Inter
national
J
our
nal
of
Electrical
and
Computer
Engineering
(IJECE)
V
ol.
15,
No.
6,
December
2025,
pp.
5380
∼
5387
ISSN:
2088-8708,
DOI:
10.11591/ijece.v15i6.pp5380-5387
❒
5380
Enhanced
matrix
pencil
method
f
or
r
ob
ust
and
efcient
dir
ection
of
arri
v
al
estimation
in
sparse
and
multi-fr
equency
en
vir
onments
Ashraya
A.
N.,
Punithkumar
M.
B.
Department
of
Electronics
and
Communication
Engineering,
PES
Colle
ge
of
Engineering,
Mandya,
India
Article
Inf
o
Article
history:
Recei
v
ed
Jan
11,
2025
Re
vised
Sep
3,
2025
Accepted
Sep
15,
2025
K
eyw
ords:
Direction
of
arri
v
al
estimation
Lo
w
signal-to-noise
ratio
Matrix
pencil
method
P
article
sw
arm
optimization
Signal
processing
Sparse
arrays
ABSTRA
CT
Accurate
direction
of
arri
v
al
(DO
A)
estimation
is
vital
for
applications
in
radar
,
sonar
,
wireless
communication,
and
localization.
This
paper
proposes
an
en-
hanced
matrix
pencil
method
(MPM)
frame
w
ork
to
o
v
ercome
limitations
of
traditional
methods
such
as
noise
sensiti
vity
,
computational
inef
cienc
y
,
and
challenges
with
sparse
arrays.
The
frame
w
ork
incorporates
w
a
v
elet-based
de-
noising
for
impro
v
ed
rob
ustness
in
lo
w
signal
-to-noise
ratio
(SNR)
en
viron-
ments
and
emplo
ys
particle
sw
arm
optimizati
on
(PSO)
to
optimize
k
e
y
param-
eters,
achie
ving
a
balance
between
accurac
y
and
ef
cienc
y
.
Extending
MPM
to
tw
o-dimensional
(2D)
DO
A
estimation,
the
method
precisely
determines
az-
imuth
and
ele
v
ation
angles.
Comprehensi
v
e
mathematical
formulations
and
eigen
v
alue
computations
underlie
the
proposed
enhancements.
Simulation
re-
sults
v
alidate
its
superiority
o
v
er
state-of-the-a
rt
techniques
lik
e
MUSIC
and
ES-
PRIT
,
achie
vi
ng
up
to
30%
impro
v
ement
in
root
mean
square
error
(RMSE)
and
reducing
computational
time
by
20%–30%.
Sensiti
vity
analysis
demonstrates
rob
ustness
across
v
arying
noise
le
v
el
s,
array
geometries,
and
multi-frequenc
y
scenarios.
This
scalable
and
ef
cient
frame
w
ork
addres
ses
critical
challenges
in
DO
A
estimation
and
of
fers
promising
directions
for
future
adv
ancements
in
real-time
and
resource-constrained
en
vironments.
This
is
an
open
access
article
under
the
CC
BY
-SA
license
.
Corresponding
A
uthor:
Ashraya
A.
N.
Department
of
Electronics
and
Communication
Engineering,
PES
Colle
ge
of
Engineering
Mandya,
Karnataka,
India
Email:
ashraya009@gmail.com
1.
INTR
ODUCTION
Accurate
estimation
of
the
direction
of
arri
v
al
(DO
A)
of
incoming
signals
is
a
cornerstone
of
modern
signal
processing,
with
applications
spanning
radar
,
sonar
,
wireless
communication,
and
radio
astronomy
[1].
The
increasing
demand
for
high-resolution,
real-time
DO
A
estimation
in
adv
anced
applications
such
as
5G,
autonomous
v
ehicles,
and
internet
of
things
(IoT)
systems
underscores
the
need
for
rob
ust
and
computationally
ef
cient
algorithms
[2],
[3].
The
matrix
pencil
method
(MPM)
has
g
ained
recognition
for
its
lo
w
computational
comple
xity
and
ability
to
handle
challenging
scenarios,
outperforming
traditional
approaches
lik
e
MUSIC
and
ESPRIT
,
particularly
in
lo
w
signal-to-noise
ratio
(SNR)
en
vironments
and
for
coherent
signals
[4]–[6].
Despite
its
adv
antages,
standard
implementations
of
MPM
f
ace
se
v
eral
challenges,
including
ambi-
guities
in
sparse
arrays,
mutual
coupling
ef
fects
am
ong
closely
spaced
sensors,
and
dif
culties
in
processing
multi-frequenc
y
signals
[7]–[9].
These
limitations
restrict
the
practical
applicability
of
MPM
in
dynamic
en
vi-
J
ournal
homepage:
http://ijece
.iaescor
e
.com
Evaluation Warning : The document was created with Spire.PDF for Python.
Int
J
Elec
&
Comp
Eng
ISSN:
2088-8708
❒
5381
ronments
where
precise
localization
of
signal
sources
is
crucial
[10],
[11].
This
research
proposes
an
enhanced
MPM-based
frame
w
ork
that
incorporates
w
a
v
elet-based
de-noising
for
impro
v
ed
rob
ustness
in
lo
w
SNR
con-
ditions
and
a
unitary
matrix
transformation
to
address
mutual
coupling
ef
fects.
Additionally
,
the
frame
w
ork
e
xtends
MPM’
s
capabilities
t
o
support
multi
-frequenc
y
signals
and
scalable
one-dim
ensional
(1D)
and
tw
odi-
mensional
(2D)
DO
A
estimation,
tailored
for
sparse
and
irre
gular
arrays
[12]–[14].
2.
CONTRIB
UTION
This
paper
mak
es
signicant
contrib
utions
to
the
eld
of
DO
A
estimation
by
addressing
k
e
y
limi
ta-
tions
of
the
MPM.
The
major
contrib
utions
of
this
research
are
summarized
belo
w:
−
Proposed
a
no
v
el
MPM
frame
w
ork
for
one-dimensional
(1D)
and
tw
o-dimens
ional
(2D)
DO
A
estimation,
incorporating
w
a
v
elet-based
de-noising
to
enhance
rob
ustness
in
SNR
en
vironments
and
a
unitary
matrix
transformation
to
mitig
ate
mutual
coupling
ef
fects.
−
Introduced
particle
sw
arm
optimization
(PSO)
to
ne-tune
critical
MPM
parameters,
including
the
pencil
f
actor
and
noise
thresholds.
−
Comprehensi
v
ely
performed
an
e
xtensi
v
e
sensiti
vity
analys
is
under
v
arying
noise
le
v
els,
array
congura-
tions,
and
signal
frequencies.
The
proposed
contrib
utions
establish
a
rob
ust
and
computationally
ef
cient
solution
for
DO
A
es
ti-
mation,
enabling
the
practical
deplo
yment
of
MPM
in
adv
anced
signal
processing
applications
and
modern
communication
systems.
3.
PR
OPOSED
METHODOLOGY
The
proposed
frame
w
ork
emplo
ys
the
MPM
for
ef
cient
and
rob
ust
DO
A
estimation,
addressi
ng
chal-
lenges
such
as
sparse
arrays,
multi-frequenc
y
signals,
and
mutual
coupling
ef
fects.
This
methodology
inte
grates
a
system
model,
signal
formulation,
and
adv
anced
techniques
tailored
to
enhance
MPM’
s
performance.
A
uniform
linear
array
(ULA)
with
M
antenna
elements,
spaced
d
apart,
is
considered.
The
recei
v
ed
signal
at
the
m
-th
antenna
is
modeled
as
[15]
in
(1):
x
m
(
t
)
=
N
X
i
=1
s
i
(
t
)
e
j
2
π
d
sin(
θ
i
)
/λ
+
n
m
(
t
)
(1)
where
s
i
(
t
)
represents
the
i
-th
source
signal,
λ
is
the
w
a
v
elength,
and
n
m
(
t
)
is
additi
v
e
noise.
Collecti
v
ely
,
the
recei
v
ed
signals
are
e
xpressed
in
(2):
X
(
t
)
=
A
(
θ
)
S
(
t
)
+
N
(
t
)
(2)
where
A
(
θ
)
is
the
steering
matrix,
S
(
t
)
is
the
source
signal
matrix,
and
N
(
t
)
is
the
noise
matrix.
DO
A
estimation
with
MPM
proceeds
as
follo
ws:
The
DO
A
estimation
process
using
the
MPM
in-
v
olv
es
three
major
steps.
First,
the
recei
v
ed
data
are
transformed
into
a
Hank
el
matrix
Y
as
Y
=
x
(1)
x
(2)
·
·
·
x
(
L
)
x
(2)
x
(3)
·
·
·
x
(
L
+
1)
.
.
.
.
.
.
.
.
.
.
.
.
x
(
K
)
x
(
K
+
1)
·
·
·
x
(
M
)
,
L
is
the
pencil
parameter
.
Ne
xt,
singular
v
alue
decomposition
(SVD)
is
applied
to
Y
,
i.e.,
Y
=
U
Σ
V
H
,
where
the
dominant
singular
v
alues
represent
the
signal
subspace.
Subsequently
,
submatrices
Y
1
and
Y
2
are
formed
by
e
xcluding
the
last
and
rst
ro
ws
of
Y
,
respecti
v
ely
,
and
the
relation
Y
2
=
Λ
Y
1
,
θ
i
=
arcsin
λ
2
π
d
arg
(
λ
i
)
is
solv
ed
to
obtain
the
eigen
v
alues
λ
i
and
compute
the
DO
A
angles.
This
compact
formulation
enhances
read-
ability
while
re
taining
the
essential
mathematical
clarity
of
the
MPM
procedure.
This
streamlined
methodology
ensures
rob
ust
and
computationally
ef
cient
DO
A
estimation,
making
it
suitable
for
real-w
orld
applications
such
as
radar
and
communication
systems.
Enhanced
matrix
pencil
method
for
r
ob
ust
and
ef
cient
dir
ection
of
arrival
estimation
in
...
(Ashr
aya
A
N)
Evaluation Warning : The document was created with Spire.PDF for Python.
5382
❒
ISSN:
2088-8708
4.
PR
OPOSED
EFFICIENT
TECHNIQ
UE
FOR
DO
A
ESTIMA
TION
WITH
NUMERICAL
COM-
PUT
A
TION
The
proposed
frame
w
ork
enhances
rob
ustness
and
optimizes
performance
in
DO
A
estimation
by
in-
te
grating
adv
anced
techniques.
Donoho’
s
w
a
v
elet
shrinkage
is
applied
to
remo
v
e
high-frequenc
y
noise
under
lo
w
SNR
conditions
while
preserving
the
signal
structure
[16].
PSO
is
used
to
optimize
the
pencil
parameter
L
and
singular
v
alue
selection
thresholds,
minimizing
the
root
mean
square
error
(RMSE)
of
DO
A
estimation:
Fitness
=
1
N
N
X
i
=1
(
ˆ
θ
i
−
θ
i
)
2
.
The
frame
w
ork,
illustrated
in
Figure
1,
consists
of
signal
preprocessing,
Hank
el
matrix
formation,
eigen
v
alue
computation,
and
DO
A
estimation
[17]–[19].
Comprehensi
v
e
mathematical
deri
v
ations
in
section
4
ensure
theoretical
rigor
and
reproducibility
.
By
combining
w
a
v
elet-based
de-noising,
MPM,
eigen
v
alue
com-
putation,
and
PSO-based
optimization,
the
frame
w
ork
achie
v
es
high
accurac
y
and
computational
ef
cienc
y
,
addressing
critical
challenges
in
DO
A
estimation.
Figure
1.
Block
diagram
of
the
proposed
DO
A
estimation
frame
w
ork
using
matrix
pencil
method
4.1.
W
a
v
elet-based
de-noising
T
o
impro
v
e
rob
ustness
in
l
o
w
SNR
en
vironments,
Donoho’
s
w
a
v
elet
de-noising
is
applied
to
the
re-
cei
v
ed
signal
x
(
t
)
.
The
signal
is
decomposed
as:
x
(
t
)
=
K
X
k
=0
c
k
ψ
k
(
t
)
,
where
c
k
are
the
w
a
v
elet
coef
cients,
and
ψ
k
(
t
)
are
w
a
v
elet
basis
functions
[18].
Thresholding
c
k
eliminates
high-frequenc
y
noise,
yielding
a
noise-reduced
signal
[20].
4.2.
Matrix
pencil
method
(MPM)
f
ormulation
The
de-noised
signal
is
transformed
into
a
Hank
el
matrix
Y
to
capture
spatial
and
temporal
character
-
istics:
Y
=
x
(0)
x
(1)
·
·
·
x
(
L
−
1)
x
(1)
x
(2)
·
·
·
x
(
L
)
.
.
.
.
.
.
.
.
.
.
.
.
x
(
M
−
L
+
1)
x
(
M
−
L
+
2)
·
·
·
x
(
M
)
,
where
L
is
the
pencil
parameter
[21].
4.3.
Eigen
v
alue
computation
Submatrices
Y
1
and
Y
2
are
constructed
by
remo
ving
the
last
and
rst
ro
ws
of
Y
,
respecti
v
ely:
Y
2
=
ΛY
1
,
where
Λ
contains
eigen
v
alues.
The
DO
A
angles
θ
i
are
obtained
as:
θ
i
=
arcsin
λ
2
π
d
arg
(
λ
i
)
[
22
]
.
Int
J
Elec
&
Comp
Eng,
V
ol.
15,
No.
6,
December
2025:
5380-5387
Evaluation Warning : The document was created with Spire.PDF for Python.
Int
J
Elec
&
Comp
Eng
ISSN:
2088-8708
❒
5383
4.4.
Optimization
with
particle
swarm
optimization
(PSO)
PSO
optimizes
the
pencil
parameter
L
and
noise
thresholds.
The
tness
function
minimizes
DO
A
estimation
error:
Fitness
=
1
N
N
X
i
=1
ˆ
θ
i
−
θ
i
2
,
where
ˆ
θ
i
and
θ
i
are
the
estimated
and
actual
DO
As
[5].
P
articl
es
iterati
v
ely
update
their
positions
and
v
elocities
based
on
indi
vidual
and
global
best
positions,
ensuring
ef
cient
e
xploration
of
the
solution
space.
The
con
v
er
-
gence
curv
e,
illustrated
in
Figure
2,
demonstrates
rapid
initial
impro
v
ement
follo
wed
by
gradual
renement,
indicating
rob
ust
and
ef
cient
optimization
[23].
The
inte
gration
of
these
techniques
ensures
high
accurac
y
and
computational
ef
cienc
y
,
making
the
proposed
frame
w
ork
suitable
for
real-w
orld
DO
A
estimation
challenges.
Figure
2.
Con
v
er
gence
graph
of
the
proposed
PSO
optimization
5.
RESUL
TS
AND
COMP
ARA
TIVE
AN
AL
YSIS
Before
conducting
result
analysis,
sensiti
vity
analysis
is
conducted
to
e
v
aluate
the
rob
ustness
of
the
proposed
method
by
systematically
v
arying
k
e
y
parameters,
i
ncluding
noise
le
v
els
within
an
SNR
range
of
-10
dB
to
+20
dB,
array
congurations
such
as
uniform
and
sparse
setups,
and
signal
frequencies
encompassing
single
and
multi-frequenc
y
scenarios.
The
impact
of
these
v
ariations
on
estimation
accurac
y
is
thoroughly
analyzed
to
g
ain
insights
into
the
resilience
of
the
proposed
approach.
The
performance
of
the
proposed
MPM
frame
w
ork
for
DO
A
estimation
is
e
v
aluated
based
on
se
v
eral
metrics,
including
root
mean
square
error
(RMSE)
v
ersus
signal-to-noise
ratio
(SNR)
depicts
i
n
Figure
3(a),
computational
ef
cienc
y
,
and
rob
ustness
under
v
arying
conditions
depicts
in
Figure
3(b).
Comparati
v
e
analyses
are
conducted
ag
ainst
state-of-the-art
methods
such
as
MUSIC
and
ESPRIT
to
demonstrate
the
ef
fecti
v
eness
of
the
proposed
approach
depicts
in
Figure
4(a)
[1],
[5],[7].
The
simulations
were
performed
using
a
uniform
linear
array
(ULA)
of
8
elements
with
half-w
a
v
elength
spacing
(
d
=
λ/
2
).
Source
signals
were
generated
across
a
range
of
angles
with
the
follo
wing
congurations:
signal
frequencies
of
1
GHz
and
2
GHz,
SNR
le
v
els
v
arying
from
−
10
dB
to
20
dB,
three
source
signals
(
N
=
3
),
and
optimization
settings
in
v
olving
PSO
with
50
particles
and
100
iterations.
Performance
e
v
al
uation
focused
on
three
metrics.
First,
RMSE
measured
the
de
viation
of
estim
ated
angles
from
true
angles
using
the
formula:
RMSE
=
v
u
u
t
1
N
N
X
i
=1
(
ˆ
θ
i
−
θ
i
)
2
.
Second,
computational
ef
cienc
y
w
as
assessed
by
the
time
required
for
DO
A
estimation
in
seconds.
Third,
rob
ustness
w
as
analyzed
by
studying
the
sensiti
vity
of
the
proposed
method
to
noise
and
sparse
array
congu-
rations.
Enhanced
matrix
pencil
method
for
r
ob
ust
and
ef
cient
dir
ection
of
arrival
estimation
in
...
(Ashr
aya
A
N)
Evaluation Warning : The document was created with Spire.PDF for Python.
5384
❒
ISSN:
2088-8708
(a)
(b)
Figure
3.
Performance
comparison
of
dif
ferent
DO
A
estimation
techniques
(a)
comparison
of
RMSE
for
dif
ferent
methods
under
v
arying
SNR
le
v
els
and
(b)
computational
ef
cienc
y
comparison
between
MPM,
MUSIC,
and
ESPRIT
The
results
demonstrate
that
the
proposed
MPM
consistently
outperforms
MUSIC
and
ESPRIT
across
v
arying
SNR
le
v
els.
As
sho
wn
in
Figure
4(a),
the
proposed
method
achie
v
es
sub-de
gree
accurac
y
e
v
en
at
SNR
=
−
5
dB,
highlighting
its
superior
rob
ustness
and
precision
in
lo
w-SNR
en
vironments.
This
establishes
the
proposed
MPM
frame
w
ork
as
a
rob
ust
and
ef
cient
solution
for
DO
A
estimation
in
practical
scenarios.
T
able
1
and
Figure
3(b)
highlight
the
computational
ef
cienc
y
of
the
proposed
MPM,
achie
ving
a
30%
reduction
in
computational
time
compared
to
MUSIC
and
a
20%
impro
v
ement
o
v
er
ESPRIT
.
Computational
ef
cienc
y
is
critical
for
real-time
applications
in
communication
and
radar
systems,
where
rapid
processing
is
essential
[1],
[9],
[3].
The
proposed
MPM
eliminates
co
v
ariance
matrix
construction
and
directly
operates
on
structured
signal
matrices,
signicantly
reducing
o
v
erhead.
Streamlined
eigen
v
alue
computation
and
PSO
further
enhance
ef
cienc
y
,
balancing
accurac
y
and
computational
demands.
These
adv
ancements
mak
e
the
frame
w
ork
scalable
and
suitable
for
real-w
orld
systems
requiring
ef
cient
and
precise
DO
A
estimation.
Furthermore,
Figure
4(b)
illustrates
the
rob
ustness
of
the
proposed
frame
w
ork
under
sparse
array
congurations.
The
method
maintains
high
accurac
y
across
v
arying
element
spacings,
demonstrating
its
adapt-
ability
to
non-uniform
arrays.
Moreo
v
er
,
the
proposed
MPM
frame
w
ork
w
as
compared
with
e
xisting
methods,
including
MUSIC
and
ESPRIT
,
based
on
RMSE,
computational
ef
cienc
y
,
and
rob
ustness
[15],
[19],
[24].
The
analysis
re
v
ealed
that
the
proposed
method
achie
v
es
a
20%–30%
impro
v
ement
in
accurac
y
under
lo
w
SNR
conditions.
Computational
ef
cienc
y
is
enhanced
through
direct
eigen
v
alue
computation
and
optimized
pencil
parameters.
Additionally
,
the
proposed
frame
w
ork
demonstrates
superior
rob
ustness,
particularly
in
handling
sparse
and
irre
gular
array
congurations,
outperforming
state-of-the-art
techniques
[1],
[25],
[5].
Int
J
Elec
&
Comp
Eng,
V
ol.
15,
No.
6,
December
2025:
5380-5387
Evaluation Warning : The document was created with Spire.PDF for Python.
Int
J
Elec
&
Comp
Eng
ISSN:
2088-8708
❒
5385
(a)
(b)
Figure
4.
Comprehensi
v
e
analysis
of
rob
ustness
and
ef
cienc
y
of
the
proposed
frame
w
ork
(a)
comparati
v
e
analysis
of
computational
ef
cienc
y
and
(b)
performance
of
the
proposed
method
under
sparse
array
congurations
T
able
1.
Computational
ef
cienc
y
comparison
Method
Comput
ation
time
(s)
Accurac
y
(RMSE)
SNR
range
(dB)
Proposed
MPM
0.45
0.8
-10
to
20
MUSIC
0.65
1.5
-5
to
20
ESPRIT
0.55
1.2
-7
to
20
6.
DISCUSSION
The
proposed
MPM
frame
w
ork
enhances
DO
A
estimation
by
inte
grating
w
a
v
elet-based
de-noising
and
PSO,
achie
ving
high
accurac
y
in
lo
w
SNR
conditions
and
impro
v
ed
computational
ef
cienc
y
by
eliminat-
ing
co
v
ariance
matrix
construction.
Its
rob
ust
performance
with
sparse
arrays
and
2D
DO
A
estimation
broadens
its
applicability
to
adv
anced
radar
and
localization
systems.
Challenges
include
the
computational
o
v
erhead
of
PSO
for
lar
ge-scale
arrays,
renement
needed
for
closely
spaced
multi-frequenc
y
signals,
and
hardw
are
implementation
constraints
in
resource-limited
en
vironments.
Future
directions
in
v
olv
e
adapti
v
e
optimization
techniques
lik
e
genetic
algorithms
(GA),
impro
v
ed
multi-frequenc
y
signal
handling,
and
hardw
are
acceleration
using
FPGAs
or
GPUs.
Incorporating
machine
learning
models
could
further
enable
adapti
v
e
DO
A
estimation
in
dynamic
scenarios.
In
summary
,
the
frame
w
ork
of
fers
a
scala
b
l
e
and
ef
cient
solution
for
DO
A
estimation
with
signicant
potential
for
further
adv
ancements.
7.
CONCLUSION
This
paper
presents
an
enhanced
MPM
frame
w
ork
for
DO
A
estimation,
addressing
critical
chal
lenges
such
as
noise
sensiti
vity
,
computational
inef
cienc
y
,
and
limitations
with
sparse
and
irre
gular
array
congura-
Enhanced
matrix
pencil
method
for
r
ob
ust
and
ef
cient
dir
ection
of
arrival
estimation
in
...
(Ashr
aya
A
N)
Evaluation Warning : The document was created with Spire.PDF for Python.
5386
❒
ISSN:
2088-8708
tions.
The
proposed
method
achie
v
es
20%–30%
higher
accurac
y
,
with
reduced
RMSE
compared
to
MUSIC
and
ESPRIT
under
lo
w
SNR
conditions.
Computational
ef
cienc
y
is
signicantly
i
mpro
v
ed,
reducing
e
x
ecu-
tion
time
by
30%
through
direct
eigen
v
alue
computation.
The
frame
w
ork
maintains
sub-de
gree
accurac
y
in
sparse
and
irre
gular
arrays
and
e
xtends
its
scalability
to
2D
DO
A
estimation,
broadening
its
applicability
to
adv
anced
radar
and
localization
systems.
Comparati
v
e
analyses
further
v
alidate
the
method’
s
superiority
in
terms
of
RMSE,
SNR
thresholds,
computational
time
(0.45
seconds
vs.
0.65
seconds
for
MUSIC
and
0.55
seconds
for
ESPRIT),
and
adaptability
to
non-uniform
arrays.
Future
research
direct
ions
include
inte
grating
adv
anced
optimization
techniques
such
as
genetic
algorithms
or
deep
reinforcement
learning,
enhancing
multi-frequenc
y
signal
handling,
e
xploring
hardw
are
acceleration
using
FPGAs
or
GPUs,
and
le
v
eraging
machine
learning
for
adapti
v
e
DO
A
estimation.
In
summary
,
the
proposed
MPM
frame
w
ork
of
fers
a
rob
ust,
ef
cient,
and
scalable
solution
for
modern
com-
munication
and
radar
systems.
Addressi
ng
the
outlined
future
direct
ions
will
further
enhance
its
impact
and
applicability
in
real-w
orld
scenarios.
REFERENCES
[1]
C
.
Greif
f,
F
.
Gio
v
anneschi,
and
M.
A.
Gonzalez-Huici,
“Matrix
pencil
method
for
doa
estimation
with
interpolated
arrays,
”
in
2020
IEEE
International
Radar
Conference
(RAD
AR)
,
2020,
pp.
566–
571,
doi:
10.1109/RAD
AR42522.2020.9114577.
[2]
B
.
Boustani,
A.
Baghdad,
A.
Sahel,
and
A.
Badri,
“Performance
analysis
of
direction
of
arri
v
al
algorithms
for
smart
an-
tenna,
”
International
Journal
of
Electrical
and
Computer
Engineering
(IJECE)
,
v
ol.
9,
no.
6,
pp.
4873-4881,
2019,
doi:
10.11591/ijece.v9i6.pp4873-4881.
[3]
L.
Li,
Y
.
Chen,
B.
Zang,
and
L.
Jiang,
“
A
high-precision
tw
o-di
mensional
DO
A
estimation
algorithm
with
parallel
coprime
array
,
”
Circuits,
Systems,
and
Signal
Processing
,
v
ol.
41,
no.
12,
pp.
6960–6974,
2022,
doi:
10.1007/s00034-022-02102-7.
[4]
K
.
Raghu
and
K.
N.
Prameela,
“Bayesian
learning
scheme
for
sparse
DO
A
estimation
based
on
maximum-
a-posteriori
of
h
y-
perparameters,
”
International
Journal
of
Electrical
and
Computer
Engineering
(IJECE)
,
v
ol.
11,
no.
4,
pp3049-3058,
2021,
doi:
10.11591/ijece.v11i4.pp3049-3058.
[5]
M.
A.
Ihedrane
and
S.
Bri,
“Direction
of
arri
v
al
in
tw
o
dimensions
with
matrix
pencil
method,
”
in
International
Conference
on
Information
T
echnology
and
Communication
Systems
,
2018,
pp.
219–228,
doi:
10.1007/978-3-319-64719-7-19.
[6]
M.
Pesa
v
ento,
M.
T
rinh-Hoang,
and
M.
V
iber
g,
“Three
more
decades
in
array
signal
processing
research:
An
optimiza-
tion
and
structure
e
xploitation
perspecti
v
e,
”
IEEE
Signal
Processing
Mag
azine
,
v
ol.
40,
no.
4,
pp.
92–106,
2023,
doi:
10.1109/MSP
.2023.3255558.
[7]
H.
T
.
Thanh
and
V
.
V
an
Y
em,
“Rob
ust
system
architecture
for
DO
A
estimation
based
on
total
forw
ard
backw
ard
matrix
pencil
algorithm,
”
International
Journal
of
Computer
Applications
,
v
ol.
126,
no.
4,
2015,
doi:
10.5120/ijca2015906033.
[8]
S.
Zheng,
Z.
Y
ang,
W
.
Shen,
L.
Zhang,
J.
Zhu,
Z.
Zhao,
and
X.
Y
ang,
“Deep
learning-based
doa
estimation,
”
IEEE
T
ransactions
on
Cogniti
v
e
Communications
and
Netw
orking
,
v
ol.
10,
no.
3,
pp.
819-835,
2024,
doi:
10.1109/TCCN.2024.3360527.
[9]
J.
K
oh
and
T
.
K.
Sarkar
,
“
High
resolution
doa
estimation
using
matrix
pencil,
”
IEEE
Antennas
and
Propag
ation
Society
Symposium
,
v
ol.
1,
v
ol.
1,
pp.
423–426,
2004,
doi:
10.1109/APS.2004.1329664.
[10]
J.
Song,
L.
Cao,
Z.
Zhao,
D.
W
ang,
and
C.
Fu,
“F
ast
DO
A
estimation
algorithms
via
positi
v
e
incremental
modied
cholesk
y
decomposition
for
augmented
coprime
array
sensors,
”
Sensors
,
v
ol.
23,
no.
21,
p.
8990,
2023,
doi:
10.3390/s23218990.
[11]
A.
Azarbar
,
G.
Dadashzadeh,
and
H.
Bakhshi,
“2-D
DO
A
estimation
with
matrix
pencil
method
in
the
presence
of
mutua
l
coupling,
”
The
Applied
Computational
Electromagnetics
Society
Journal
(A
CES)
,
pp.
742–748,
2012.
[12]
P
.
Chen,
Z.
Chen,
L.
Liu,
Y
.
Chen,
and
X.
W
ang,
“SDO
A-Net:
An
ef
cient
deep-learning-based
DO
A
estimation
netw
ork
for
im-
perfect
array
,
”
IEEE
T
r
ansactions
on
Instrumentation
and
Measur
ement
,
v
ol.
73,
pp.
1–12,
2024,
doi:
10.1109/TIM.2024.3391338.
[13]
I.
Aboumahmoud,
A.
Muqaibel,
M.
Alhassoun,
and
S.
Ala
wsh,
“
A
re
vie
w
of
sparse
sensor
arrays
for
tw
o-dimensional
direction-of-
arri
v
al
estimation,
”
IEEE
Access
,
v
ol.
9,
pp.
92999–93017,
2021,
doi:
10.1109/A
CCESS.2021.3092529.
[14]
Y
.
Liu,
L.
Zhang,
C.
Zhu,
and
Q.
H.
Liu,
“Synthesis
of
nonuniformly
spaced
linear
arrays
with
frequenc
y-in
v
ariant
patterns
by
the
generalized
matrix
pencil
methods,
”
IEEE
T
r
ansactions
on
Antennas
and
Pr
opa
gation
,
v
ol.
63,
no.
4,
pp.
1614–1625,
Apr
.
2015,
doi:
10.1109/T
AP
.2015.2394497.
[15]
T
.
Lin,
X.
Zhou,
Y
.
Zhu,
and
Y
.
Jiang,
“Hybrid
beamforming
optimization
for
DO
A
estimation
based
on
the
CRB
analysis,
”
IEEE
Signal
Pr
ocessing
Letter
s
,
v
ol.
28,
pp.
1490–1494,
2021,
doi:
10.1109/LSP
.2021.3092613.
[16]
Y
errisw
amy
and
Jag
adeesha,
“F
ault
tolerant
matrix
pencil
method
for
direction
of
arri
v
al
estimation,
”
Signal
Ima
g
e
Pr
ocessing:
An
International
J
ournal
,
v
ol.
2,
no.
3,
pp.
55–67,
Sep.
2011,
doi:
10.5121/sipij.2011.2306.
[17]
N.
Y
ilmazer
,
J.
K
oh,
and
T
.
K.
Sarkar
,
“Utilization
of
a
unitary
transform
for
ef
cient
computation
in
the
matrix
pencil
method
to
nd
the
direction
of
arri
v
al,
”
IEEE
T
r
ansactions
on
Antennas
and
Pr
opa
gation
,
v
ol.
54,
no.
1,
pp.
175–180,
2006,
doi:
10.1109/T
AP
.2005.861567.
[18]
R.
K.
Arumug
am,
A.
Froehly
,
P
.
W
allrath,
R.
Herschel,
and
N.
Pohl,
“Matrix
pencil
based
estimation
of
parame
ters
in
2D
for
non-
collinear
uniform
linear
arrays,
”
in
Pr
oceedings
of
the
2024
21st
Eur
opean
Radar
Confer
ence
(EuRAD
2024)
,
pp.
364–367,
2024,
doi:
10.23919/EuRAD61604.2024.10734880.
[19]
L.
A.
T
rinh,
N.
D.
Thang,
D.
Kim,
S.
Lee,
and
S.
Chang,
“
Application
of
matrix
pencil
algorithm
to
mobile
robot
localization
using
h
ybrid
DO
A/T
O
A
estimation,
”
International
J
ournal
of
Advanced
Robotic
Systems
,
v
ol.
9,
2012,
doi:
10.5772/54712.
[20]
P
.
Rocca,
M.
A.
Hannan,
M.
Salucci,
and
A.
Massa,
“Single-snapshot
DO
A
estimation
in
array
antennas
with
mutual
coupling
through
a
multiscaling
BCS
strate
gy
,
”
IEEE
T
r
ansactions
on
Antennas
and
Pr
opa
gation
,
v
ol.
65,
no.
6,
pp.
3203–3213,
2017,
doi:
10.1109/T
AP
.2017.2684137.
[21]
T
.
T
irer
and
O.
Bialer
,
“Direction
of
arri
v
al
estimation
and
phase-correction
for
noncoherent
subarrays:
A
con
v
e
x
op-
Int
J
Elec
&
Comp
Eng,
V
ol.
15,
No.
6,
December
2025:
5380-5387
Evaluation Warning : The document was created with Spire.PDF for Python.
Int
J
Elec
&
Comp
Eng
ISSN:
2088-8708
❒
5387
timization
approach,
”
IEEE
T
r
ansactions
on
Aer
ospace
and
Electr
onic
Systems
,
v
ol.
58,
no.
6,
pp.
5571–5585,
2022,
doi:
10.1109/T
AES.2022.3175465.
[22]
X.
Jin
et
al.
,
“Of
f-grid
DO
A
estimation
algorithm
based
on
e
xpanded
matr
ix
and
matrix
pencil,
”
in
Pr
oceedings
of
the
IEEE
V
ehicular
T
ec
hnolo
gy
Confer
ence
,
2024,
pp.
1–6,
doi:
10.1109/VTC2024-Spring62846.2024.10683422.
[23]
Y
.
Zheng
and
Y
.
Y
u,
“Joint
estimation
of
DO
A
and
TDO
A
of
multiple
reections
by
matrix
pencil
in
mobile
communications,
”
IEEE
Access
,
v
ol.
7,
pp.
15469–15477,
2019,
doi:
10.1109/A
CCESS.2019.2895102.
[24]
H.
A.
T
anti,
A.
Datta,
and
S.
Ananthakrishnan,
“Snapshot
a
v
eraged
matrix
pencil
method
(SAM)
for
direction
of
arri
v
al
estimation,
”
Experimental
Astr
onomy
,
v
ol.
56,
no.
1,
pp.
267–292,
2023,
doi:
10.1007/s10686-023-09897-6.
[25]
R.
Shari
and
S.
Jacob,
“Comparati
v
e
study
of
DO
A
estimation
algorithms,
”
in
Pr
oceedings
of
the
INDIC
ON
2022
-
IEEE
19
th
India
Council
International
Confer
ence
,
2022,
pp.
1–4,
doi:
10.1109/INDICON56171.2022.10040076.
BIOGRAPHIES
OF
A
UTHORS
Ashraya
A.
N.
w
orking
as
an
ass
istant
professor
in
the
Department
of
Electronics
and
Communication
Engineering
PES
Colle
ge
of
Engineering,
Mandya
has
about
10
years
of
teaching
e
xperience.
She
recei
v
ed
her
B.E
de
gree
in
Electronics
and
Communication
Engineering
and
M.T
ech.
de
gree
in
Digital
Electronics
and
Communication
from
V
isv
esv
araya
T
echnological
Uni
v
ersity
,
Be-
lag
a
vi,
Karnataka.
Her
areas
of
research
includes
communication,
signal
processing.
She
can
be
contacted
at
email:
ashraya009@gmail.com.
Punith
K
umar
M.
B.
obtained
his
B.E.
de
gree
in
electronics
and
communication
engineer
-
ing
from
The
National
Institute
of
Engineering,
Mysore
in
2007,
and
the
M.T
ech
in
VLSI
design
and
embedded
systems
from
PES
Colle
ge
of
Engineering,
Mandya
under
the
The
V
isv
esv
araya
T
echno-
logical
Uni
v
ersity
(VTU),
Belg
aum
in
2010
and
Ph.D.
de
grees
in
Electronics
from
the
Uni
v
ersity
of
Mysore
(UoM),
Mysore,
India,
in
2017.
He
presently
w
orking
as
professor
and
HOD
in
Department
of
Electronics
and
Communication
Engineering,
PES
Colle
ge
of
Engineering
Mandya.
His
current
research
interests
include
image
processing,
video
processing,
video
shot
detection,
and
embedded
system.
Published
30
paper
in
the
international
and
national
journal
and
obtained
one
patent,
Pub-
lished
the
book
on
his
research
w
ork.
Dr
.
Punith
K
umar
M
B
is
a
Member
of
IEEE.
Life
Member
of
the
Indian
Society
for
T
echnical
Education
(ISTE)
and
associate
member
of
the
Institution
of
Engi-
neers
(AMIE),
He
w
as
the
Judge,
Chairperson
and
Re
vie
w
member
for
the
National
and
International
Conference.
He
can
be
contacted
at
email:
punithkumarmb@pesce.ac.in.
Enhanced
matrix
pencil
method
for
r
ob
ust
and
ef
cient
dir
ection
of
arrival
estimation
in
...
(Ashr
aya
A
N)
Evaluation Warning : The document was created with Spire.PDF for Python.