Indonesian J
ournal of Ele
c
trical Engin
eering and
Computer Sci
e
nce
Vol. 1, No. 2,
February 20
1
6
, pp. 341 ~
348
DOI: 10.115
9
1
/ijeecs.v1.i2.pp34
1-3
4
8
341
Re
cei
v
ed Se
ptem
ber 19, 2015; Revi
se
d De
ce
m
ber
30, 2015; Accepted Janu
ary 25, 201
6
Noise Uncertainty Effect on a Modified Two-Stage
Spectrum Sensing Technique
Heba A.Tag El-Dien
1*
, Ro
kaia M. Zaki
1
, Mohsen M. Tanta
w
y
2
, Hala M. Abdel
-
Kader
1
1
Shoubr
a F
a
cu
lt
y
of Eng
i
ne
eri
ng, Benh
a Un
i
v
ersit
y
2
Nationa
l T
e
lecommunic
a
tio
n
Institute, Cairo
,
Eg
y
p
t
*Corres
p
o
ndi
n
g
author, e-ma
il:
heb
aal
la
h.sh
ahat@f
e
ng.b
u
.edu.e
g
A
b
st
r
a
ct
Detectin
g the prese
n
ce or a
b
senc
e of primary user is the
key ta
sk of cogn
itive rad
i
o
netw
o
rks.
How
e
ver, rely
i
ng o
n
si
ngl
e d
e
tector red
u
ce
s the pr
oba
bil
i
ty of detecti
on
and
incr
eases the
pro
b
a
b
il
ity o
f
miss
ed det
ecti
on. Co
mbi
n
i
ng
tw
o convention
a
l spectru
m
se
nsin
g techni
qu
es by integrati
ng their in
divi
d
u
a
l
features impro
v
es
the pro
bab
ility of
detecti
o
n
esp
e
ci
ally
un
der n
o
is
e u
n
ce
rtainty. This p
a
per i
n
trod
uces
a
mo
difi
ed
tw
o-s
t
age detecti
on
techni
qu
e
that
dep
en
ds
o
n
th
e e
nergy
d
e
te
ction
as a
first
stage
d
ue to
i
t
s
ease a
nd sp
ee
d of detectio
n
, and the pro
p
o
sed Mo
difi
e
d
Co
mbi
nati
o
n
a
l Maxi
mu
m-Mi
ni
mu
m Ei
ge
nval
u
e
base
d
detecti
o
n
as a secon
d
stage un
der
noise u
n
ce
rta
i
nty and co
mp
eres it w
i
th th
e case of usi
n
g
Maxi
mu
m-Mi
ni
mu
m Ei
ge
nval
ue an
d Co
mb
i
natio
nal Max
i
mu
m-M
i
ni
mu
m Eige
nval
ue as a
secon
d
stage.
Ke
y
w
ords
: Co
gnitiv
e
radi
o, Noise u
n
ce
rtai
nt
y, Tw
o-Stage Detector.
Copy
right
©
2016 In
stitu
t
e o
f
Ad
van
ced
En
g
i
n
eerin
g and
Scien
ce. All
rig
h
t
s reser
ve
d
.
1. Introduc
tion
Acco
rdi
ng to
Fede
ral Co
mmuni
cation
Commi
ssion
(FCC), la
rg
e amou
nt of unu
sed
spe
c
tru
m
i
s
available i
n
li
cen
s
e
d
spe
c
t
r
um
whi
c
h i
s
not effe
ctively use
d
du
e
to non
-unifo
rm
spe
c
tral d
e
m
and in time, frequen
cy and spa
c
e. T
h
is reve
als that the inad
equate spe
c
trum
manag
eme
n
t poli
c
ie
s a
r
e
the main
rea
s
on
for spe
c
trum
sca
r
city. To
overcom
e
this, th
e F
CC
approved to
allow existin
g
unli
c
en
se
d radio
se
rvic
e
s
in the
licen
sed TV
White
Space
(TVWS)
throug
h Co
g
n
itive Radio (CR) [1]. In CR, the
secon
dary users n
eed to oppo
rt
unisti
c
ally se
nse
the idle
ch
an
nels.
On
ce a
n
idle
ch
ann
e
l
is
se
n
s
e
d
, the
se
cond
ary
users
will a
c
ce
ss the
cha
nnel.
Hen
c
e, sp
ect
r
um sensi
ng
requ
est
s
the se
con
dar
y u
s
ers to efficie
n
t
ly and effectively detect th
e
pre
s
en
ce of the prima
r
y si
gnal
s, and is
a fundame
n
ta
l problem in
CR [2]. Spect
r
um sen
s
ing
can
be cla
s
sified
into two m
a
in cate
go
rie
s
, namely coope
rative d
e
tection te
ch
nique a
nd n
on-
coo
perative detection te
ch
nique. Th
e n
on-coo
p
e
r
at
ive detectio
n
can be furth
e
r divided into two
cla
s
ses: (i
) bl
ind se
nsi
ng
whi
c
h do
es n
o
t need any
i
n
formatio
n a
bout the prim
ary user’
s
sig
nal
su
ch a
s
Eige
nvalue ba
se
d
detecto
r an
d
Energy Dete
ctor
(ED), (ii)
sign
al sp
ecifi
c
sen
s
ing
whi
c
h
need
s inform
ation about th
e prima
r
y use
r
’s si
gnal
su
ch as Matched
Filter (MF) a
nd The Fe
atu
r
e
detecto
r or
Cyclostatio
nary
Feature d
e
te
ctor (CF
D
) [3]
.
ED is
simpl
e
and fa
st te
chni
que,
whi
c
h
wo
rks b
e
tter in hi
gh S
i
gnal to
Noi
s
e Ratio
(SNR),
but it i
s
n
o
t robu
st a
t
low S
N
R a
n
d
c
ann
ot diffe
rentiate
between
noi
se
an
d si
gnal
[4]. MF
is a
n
o
p
tima
l dete
c
tor in
white
Ga
ussian
noi
se,
but it ne
ed
s more info
rmation a
bout
the
transmitted si
gnal. CF
D op
erate
s
in the mid-
way between ED and
MF. In one hand it need
s less
informatio
n a
bout the pri
m
ary user’
s
sign
al
than the MF; in the other h
a
nd it has be
tter
perfo
rman
ce
than the E
D
. CF
D relie
s
on the fa
ct
t
hat mo
st sig
nals
exhibit
perio
dic feat
ure
s
,
pre
s
ent i
n
pi
lots, cy
clic p
r
efixes, m
o
d
u
lation
s,
c
a
r
r
i
ers,
and
ot
h
e
r rep
e
t
i
t
i
v
e
cha
r
a
c
t
e
ri
st
ic
s.
Becau
s
e
the
noise is
no
t perio
dic, th
e sig
nal
can
be succe
ssfully detecte
d. In [5-7], the
eigenvalu
e
b
a
se
d dete
c
tion de
sign
ed
as a blin
d
sen
s
in
g tech
nique
with hi
gh proba
bility of
detection in l
o
w S
NR
envi
r
onm
ents. So for effi
cient sensi
ng IEEE 802.22
standard
prefers two-
stage
sen
s
in
g that i
s
coa
r
se
se
nsi
ng
which
covers l
a
rge
ba
nd
wid
t
h and
small
sen
s
in
g time
and
fine sen
s
ing t
hat con
c
ent
ra
tes on lower
band
width an
d use
s
very robu
st sen
s
in
g techniq
u
e
s
like
eigenvalu
e
b
a
se
d tech
niq
ues [4, 8].
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 25
02-4
752
IJEECS
Vol.
1, No. 2, February 201
6 : 341 – 348
342
Spectrum se
nsin
g faces
some
ch
allen
ges
su
ch
as low SNR fo
r prim
ary u
s
ers, time
disp
ersion,
chann
el fadin
g
,
and
noi
se
u
n
ce
rtainty. In
this p
ape
r, th
e noi
se
un
ce
rtainty effect o
n
the modified two-stag
e co
mbination
a
l maximum-mi
nimum eig
e
n
v
alue dete
c
to
r is investig
ated.
Variou
s se
nsi
ng techni
que
s and their
chara
c
te
risti
c
s are describe
d
in [3].
In [1], a semi
blind metho
d
whi
c
h is b
a
se
d on minimu
m eigenvalu
e
of a covarian
ce matrix is p
r
opo
se
d. In [2] a
novel d
e
tecto
r
i
s
p
r
o
posed
ba
sed
on
th
e ent
r
opy
of
spe
c
tru
m
p
o
w
er d
e
n
s
ity. The t
w
o-stag
e
sen
s
in
g tech
nique
s an
d there
different
algorithm
s a
r
e di
scusse
d in [4] and [8]. The autho
rs in
[13] discusse
s different sp
ectru
m
se
nsi
ng al
go
rithm and focuse
s
on se
nsin
g al
gorithm ba
se
d on
the eigenval
u
e
s of re
ceive
d
sign
al, [13] also introdu
ce
s a Matlab
code fo
r si
mulating Tra
c
y-
widom
di
strib
u
tion fun
c
tion
. The a
u
tho
r
s in [10, 1
2
] d
e
fines differe
nt eige
nvalue
algo
rithm
s
such
as maximum
-
minimum eig
envalue an
d energy to minimum eigenv
alue algo
rith
ms. The effect o
f
noise correlat
ion on
eige
nvalue b
a
sed S
pectrum
Se
n
s
ing i
s
studie
d
analytically unde
r b
o
th the
noise o
n
ly a
nd the
si
gnal
plu
s
n
o
ise h
y
potheses in
[14]. In [15]
a u
n
ified
co
mpari
s
o
n
of t
he
perfo
rman
ce
of energy detectio
n
, maximum eige
nvalue ba
se
d detectio
n
and maximu
m-
minimum eig
envalue dete
c
tion
te
chni
q
ues
fo
r c
ent
ralize
d
data
-
f
u
sio
n
coop
erative spe
c
trum
s
e
ns
ing
u
n
der
imp
u
l
s
i
ve
no
is
e
,
is
pr
es
en
te
d
.
The p
ape
r i
s
o
r
ga
nized
as foll
owi
ng,
se
c II inve
stigate
s
the
previou
s
wo
rk in
the
detec
tion algorithms
in
CRN,
s
e
c
III dis
c
uss
e
s
t
he
s
y
s
t
em model of the
propos
ed algorithms
,
and finally the pape
r is co
nclu
ded in
se
c IV.
2. Pre
v
ious
Work
The
dete
c
tio
n
p
r
oble
m
ca
n be
summa
rized
u
s
in
g
two bin
a
ry
hyp
o
theses that
indicate
the ab
sen
c
e
and the
p
r
esence of p
r
ima
r
y user’
s
sign
al. In practi
ce
, due to th
e n
o
ise
un
ce
rtai
nty,
the e
s
timate
d noi
se
po
wer may
be
different from
the a
c
tual
noi
se
power. A
nd the
estim
a
ted
noise p
o
wer
cha
nge
d in
th
e inte
rval
σ
v
2
ϵ
[
σ
v
2
/
β
,
β
σ
v
2
] w
h
er
e
β
>1
is the
n
o
ise
fl
uctuatio
n fa
ctor
whi
c
h is n
o
rm
ally range
s from 1 to 2 dB [10].
2.1. Energ
y
Detection
The en
ergy d
e
tector i
s
the
simple
st det
ecto
r
as it d
o
e
sn’t n
eed
a
n
y informatio
n abo
ut
the prima
r
y user
sign
al. It
comp
ares the
receive
d
sig
nal po
wer to the noi
se po
wer [5].
The effect of noise un
certa
i
nty
β
on the energy detection prob
abilit
y of false alarm P
fa
and proba
bility of detection
P
d
is shown in Equation (1
) and (2) resp
ectively.
(
2
)
2
2
NN
P
β
P
Q
2N
4
N
P
β
β
v
d
vv
(
2
)
Solv
ing (1) a
nd (2
) re
sults
in:
12
2
β
1
ββ
2
12
β
fa
d
N
QP
S
N
R
PQ
SN
R
(
3
)
W
h
er
e Q
(
.)
is
th
e Q fu
nc
tio
n
,
ϒ
i
s
the
thre
shol
d valu
e,
σ
v
2
is the
noise vari
an
ce, P is receiv
ed
sign
al po
wer,
N is the num
ber of sample
s, and
β
is th
e noise fluctu
ation factor.
2.2.
Maximum-minimum Eige
nv
alue Based Detectio
n
The Maxim
u
m-minim
u
m
eigenvalu
e
b
a
se
d dete
c
ti
on (M
ME) te
chni
que i
s
o
ne of the
eigenvalu
e
bl
ind se
nsi
ng d
e
tection te
ch
nique
s. MME
improve
s
th
e perfo
rma
n
ce of detectio
n
at
low SNRs, b
u
t the improvement of
sen
s
ing
perfo
rm
ance al
so
co
mes
at a cost
of com
putati
onal
2
2
N
β
pQ
β
2N
v
fa
v
Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS
ISSN:
2502-4
752
Noi
s
e Uncert
ainty Effect o
n
a Modified
Two
-
Stage S
pectrum
Sensing
…
(Heba
A.Tag El-Die
n)
343
compl
e
xity and long
-time
pro
c
e
ssi
ng [8
]. It compare
s
the ratio b
e
twee
n maxim
u
m and
mini
mum
eigenvalu
e
s of
the re
ceived sign
al cov
a
rian
ce
matri
x
to a p
r
ed
efined th
re
shol
d value
1
as
sho
w
n in Equ
a
tion (4
).
/
M
ME
ma
x
m
i
n
T
(
4
)
The probabili
ty of detection and probability
of false alarm of the MME
can be written
according to [10, 12] as foll
owin
g:
2
1
1
1
fa
NM
L
µ
PF
v
(
5
)
2
1
1
M
L
1
v
d1
NN
Þ
Þ
/
σ
µ
P1
F
v
(
6
)
Whe
r
e F1
(.) is the tra
c
y-widom
distri
b
u
tion of t
he first orde
r, M is the num
be
r of the received
antenn
as,
L is the smoot
hing factor,
2
1
µ
NM
L
,
1/
3
11
1
1
vN
M
L
NM
L
,
1
Þ
is the max eigenvalu
e
, a
nd
Þ
ML
is the minimum
eigenvalu
e
of
the received sign
al matrix, [10, 12].
To stu
d
y the
effect of noi
se un
ce
rtainty
β
on the MM
E ROC, the probability of detection
must be
written as in (7).
2
1
1
1
1
/
1
ML
v
d
N
NÞ
Þ
µ
PF
v
(
7
)
The relation b
e
twee
n P
d
an
d P
fa
for the MME with an
d wit
hout n
o
ise uncertai
n
ty at
β
=1,
β
=1.05&
β
=1.1 respec
tively is
s
h
own in Figure 1.
Figure 1. The
ROC of MM
E with and wi
thout noise fluctuatio
n at
β
=
1
,
β
=1.05
and
β
=
1
.1
3. Sy
stem
Model
3.1.
C
o
mbinat
ional Maximum-Minimum Ei
genv
alue
Acco
rdi
ng to [4], the Combination
a
l Ma
ximum-Minim
u
m eigenval
u
e
Tech
niqu
e CMME
is another form of the ei
genvalue blind sensin
g detection techni
ques .It compares the rat
i
o
betwe
en max
i
mum eige
nvalue an
d the
differen
c
e b
e
twee
n maxim
u
m and mi
ni
mum eige
nva
l
ues
of the re
ceiv
ed si
gnal
co
varian
ce m
a
trix to a p
r
ed
efined th
re
sh
old value
2
as sho
w
n i
n
Equation (8).
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 25
02-4
752
IJEECS
Vol.
1, No. 2, February 201
6 : 341 – 348
344
/
MM
E
m
a
x
m
a
x
m
i
n
T
(
8
)
The p
r
ob
abili
ty of detectio
n
and p
r
o
bab
ility of
false alarm of the
CMME can
be
written
as follo
wing:
2
'
1
1
fa
NM
L
µ
PF
v
(
9
)
''
2
M
L
1
v
d1
NN
Þ
Þ
/
σ
µ
P1
F
v
(
1
0
)
Whe
r
e
'
2
2
/1
. The effect of
noise un
ce
rtainty
β
on t
he CM
ME p
r
odu
ce
s
probability of detection writ
ten as:
''
2
1
1
/
1
ML
v
d
NN
Þ
Þ
µ
PF
v
(
1
1
)
The relation
betwe
en P
d
a
nd P
fa
for the
CMME with
and
without n
o
ise
uncertai
n
ty at
β
=1,
β
=
1
.05&
β
=1.1 re
sp
ectively is shown in Figure 2.
.
Figure 2. The
ROC of CM
ME with
and without
noi
se fluctuation
at
β
=
1
,
β
=1.05
and
β
=
1
.1
3.2.
Proposed M
odified CMM
E
Algorithm
The mo
dified
CMME tech
nique
(MCM
ME) is a n
e
w
form of u
s
ing m
a
ximu
m and
minimum
eig
envalue
s. It compa
r
e
s
the
ratio b
e
twe
e
n
the sum
and
the differen
c
e of maximu
m
and minim
u
m eigenvalu
e
of the received sign
al co
varian
ce m
a
trix to a pre
defined th
re
shold
value
3
as
sho
w
n in Equatio
n (12
)
.
/
M
M
E
m
ax
m
i
n
m
ax
m
i
n
T
(
1
2
)
The proba
bility of detection
and proba
bili
ty of
false ala
r
m of the MCMME can
be
written
as follo
wing:
2
''
1
1
fa
NM
Lµ
PF
v
(
1
3
)
Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS
ISSN:
2502-4
752
Noi
s
e Uncert
ainty Effect o
n
a Modified
Two
-
Stage S
pectrum
Sensing
…
(Heba
A.Tag El-Die
n)
345
''
''
2
M
L
1
v
d1
NN
Þ
Þ
/
σ
µ
P1
F
v
(
1
4
)
Whe
r
e
''
3
3
1/
1
. The
effect of n
o
ise un
ce
rtainty
β
on the MCMME produces
probability of detection writ
ten as:
''
''
2
1
1
/
1
ML
v
d
N
NÞ
Þ
µ
PF
v
(
1
5
)
The relation
betwe
en P
d
a
nd P
fa
for the
MCMME
wit
hout noi
se
un
certai
nty at
β
=
1
, an
d
with noi
se un
certai
nty at
β
=1.05
a
nd
β
=
1
.1 res
p
ec
tively is
s
h
own in Figure 3.
Figure 3. The
ROC of M
C
MME with
an
d without noi
se fluctuation
at
β
=
1
,
β
=1.05 and
β
=
1
.1
Figure 4(a) and (b)
show t
he re
lation
between the P
r
obability of
detection and threshold
value
for MME, CM
ME, and M
C
MME at
β
=1
and
β
=1.0
5
respe
c
tively.We n
o
te that
to get the
sa
me
probability of
detection at high noi
se fluctuations
, the threshol
d value mu
st be
decreased. Fig.
5
sho
w
s the rel
a
tion betwe
e
n
β
and the Proba
bility of d
e
tection for M
M
E, CMME,
and MCM
M
E at
P
fa
=0.07, to ensu
r
e that
as the noi
se fluctu
atio
n increa
sed
the proba
bi
lity of detection
d
e
c
r
e
as
ed
.
(a)
(b)
Figure 4. The
relation bet
ween the Pro
b
ability
of detection an
d thre
shol
d value for MME,
CMME, and
MCMME (a)
at
β
=1, (b) at
β
=
1
.05
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 25
02-4
752
IJEECS
Vol.
1, No. 2, February 201
6 : 341 – 348
346
Figure 5. The
relation bet
ween the Pro
b
ability of detection an
d
β
for MME, CMM
E
, and MCM
M
E
at
0.07
fa
P
3.3. T
w
o
-
S
t
age
S
y
stem
This se
ction explain
s
the two-stage det
ection
al
go
rithm that exploits the me
rits of ED
and o
ne
of the p
r
eviou
s
eigenvalu
e
d
e
tection
tech
nique
s. In thi
s
system th
e
first
stage, i
.
e.,
coa
r
se
sen
s
i
ng sta
ge, te
sts the
cha
n
n
e
l usi
ng E
D
tech
niqu
e. If the de
ci
sion i
n
co
arse
se
n
s
ing
(Dc) i
s
g
r
eat
er than th
e thre
shol
d
ϒ
Dc
, then the chann
el is d
e
clared a
s
o
c
cupi
ed. Else
the
received
sign
al is sen
s
ed
by using th
e se
con
d
stag
e
,
i.e.,
fine se
nsin
g stag
e
by using M
M
E
,
CMME, or MCMME. If the dec
i
s
i
on
in fin
e
s
e
ns
in
g (D
f
)
is greater than th
e thre
shol
d value
ϒ
Df
,
then the ch
an
nel is con
s
ide
r
ed a
s
occu
pi
ed else, it is empty [4].
The overall probability of detection P
dT
and p
r
ob
abilit
y of false ala
r
m P
faT
a
r
e gi
ven as
[8]:
1
f
a
T
f
aC
f
a
C
f
aF
PP
P
P
(
1
6
)
1
dT
dC
dC
dF
P
PP
P
(
1
7
)
Substituting
(2) a
nd
(7
) in
(17
)
results t
he
ove
r
all p
r
obability of d
e
tection
of two-sta
g
e
ED-MME det
ection te
chni
que a
s
sh
own in (18
)
.
11
1
dT
P
ED
M
M
E
F
Z
F
Z
Q
Y
(
1
8
)
Whe
r
e
2
2
N
NP
β
2N
4N
P
β
β
v
vv
Y
and
2
1
1
1
/
ML
v
NN
Þ
Þ
µ
Z
v
.
Similarly from
(2) and (11)
the overall
pr
ob
ability of
detectio
n
of two-stag
e ED-CMME
detectio
n
techniqu
e as
sh
own in (19
)
.
11
1'
'
dT
PE
D
C
M
M
E
F
Z
F
Z
Q
Y
(
1
9
)
Whe
r
e
''
2
1
/
'
ML
v
N
NÞ
Þ
µ
Z
v
.
And from (2
) and (15
)
the overall p
r
obab
ility of detection of two-stag
e ED-MCM
ME
detectio
n
techniqu
e as
sh
own in (20
)
.
11
1'
'
'
'
dT
P
E
D
MC
MME
F
Z
F
Z
Q
Y
(
2
0
)
Evaluation Warning : The document was created with Spire.PDF for Python.
IJEECS
ISSN:
2502-4
752
Noi
s
e Uncert
ainty Effect o
n
a Modified
Two
-
Stage S
pectrum
Sensing
…
(Heba
A.Tag El-Die
n)
347
Whe
r
e
''
'
'
2
1
/
''
ML
v
NN
Þ
Þ
µ
Z
v
.
The compa
r
e
d
RO
C of the
Two-stag
e ED
-MME, ED-CMME and ED-M
CMME detection
techni
que
s
wi
th and
witho
u
t
noise fluctu
ation at
β
=1,
β
=1.05&
β
=1.1 are
sho
w
n in Fi
gure 6
(
a),
(b), an
d(
c)
re
spe
c
tiv
e
ly
.
(a)
(b)
(c
)
Figure 6. The
ROC of the
Two
-
sta
ge ED
-MME, ED-CMME and E
D
-M
CMME d
e
tection
techni
que
with and with
out
noise flu
c
tua
t
ion (a) at
β
=
1
, (b) at
β
=1.
05& (c) at
β
=1.1
4. Conclusio
n
In this pa
per
we h
a
ve stud
ied the effe
ct of noise un
certainty on th
e Prop
osed
modified
two-stage combinational maximum-mi
nimum eig
e
n
v
alue d
e
tecto
r
(E
D-MCM
M
E
), and
comp
are
it
with the two-stage E
D
-CMME and E
D
-MME. T
h
e
re
sults sh
owed
that
with
and without noise
fluctuation
s
, ED-M
CMME
has b
e
tter p
e
rfor
m
a
n
c
e t
han ED-CM
M
E and wo
rse than E
D
-MME.
But for a noi
se fluctuatio
n
of about 1
0
%, the prob
ab
ili
ty of detectio
n
of ED-M
CMME clo
s
e
s
to it
for ED-M
ME.
Referen
ces
[1]
S
y
e
d
Sa
jja
d A
li, Cha
ng
Liu,
Mingl
u Ji
n. Mi
nimu
m Ei
ge
nv
alu
e
Detecti
o
n
for Spectrum
Sensi
ng i
n
Cog
n
itive
Ra
di
o.
Internati
o
n
a
l
Journ
a
l
of El
ectrical
an
d
C
o
mputer Eng
i
n
eeri
ng
(IJECE)
. 201
4; 4(4)
:
623-
p 63
0, ISSN: 2088-
87
08.
[2]
Nan Z
h
ao. A
Novel
T
w
o-sta
ge E
n
trop
y-b
a
s
ed R
o
b
u
st C
oop
erative
Sp
ectrum Sens
in
g Sch
e
me
w
i
t
h
T
w
o-
bit Decis
i
on in C
o
g
n
itive
Radi
o. Dali
an,
Liao
nin
g
, 116
024, Ch
in
a. 20
11.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 25
02-4
752
IJEECS
Vol.
1, No. 2, February 201
6 : 341 – 348
348
[3]
Mohd. Sa
jid I
m
am, Sheeta
l
Ingle, Sh
abn
am Ar
a. A Revie
w
pa
per
base
d
on s
p
e
c
trum sensin
g
techni
qu
es in
Cog
n
itive R
adi
o Net
w
orks.
IISTE.
2013; 3(9).
[4]
Nares
h
Gunich
ett
y
, SM Hirem
ath, SK Patra.
T
w
o
Sta
ge spe
c
trum Sensin
g for Cogn
itive R
adi
o usi
n
g
CMME. 2015
.
[5]
T
T
r
ump, I Müürse
pp.
Dete
ction Spe
ed o
f
Respons
ive
Co
mmun
icati
o
n Jammin
g
D
e
tectors.
2
nd
Internatio
na
l
Confer
ence
o
n
C
i
rcuits, S
ystems,
Comm
unic
a
tions,
Co
mputers
an
d
Appl
icatio
ns
.
Dubr
ovnik, Cro
a
tia. 201
3.
[6]
Ja
yanta Mis
h
r
a
, Deep
ak Ku
mar Barik, Ch
Manoj Kum
a
r
S
w
a
i
n. C
y
cl
o
s
tationar
y Bas
ed Spectrum
Sensi
ng i
n
Co
gnitiv
e
Rad
i
o.
Internatio
na
l Journ
a
l of Rec
ent T
e
chn
o
lo
g
y
and En
gin
e
e
r
ing (IJRT
E
)
.
201
4; 3.
[7]
Asif Mirza, F
a
ique Bi
n Arsha
d
. Performanc
e Anal
ys
is of C
y
c
l
ostati
onar
y Se
nsin
g in C
ogn
itive Ra
d
i
o
Net
w
orks. T
e
chnic
a
l rep
o
rt, IDE-11
27. 20
11
.
[8]
Kana
ba
dee Sr
i
s
ombo
on, Aka
r
a Pra
y
ote, W
i
l
a
ip
orn L
ee.
T
w
o- stage S
pect
r
um S
ensi
ng f
o
r Cog
n
itiv
e
Radi
o und
er
N
o
ise Unc
e
rtaint
y.
Eighth
Inter
natio
nal
Co
nfe
r
ence
on
Mob
i
l
e
comp
utin
g a
nd U
b
i
quito
u
s
Net
w
orki
ng (IC
M
U). 2015.
[9]
Yan Z
h
an
g s J
un Z
h
en
g s
Hs
iao-H
w
a
Ch
en.
Co
gnitiv
e
R
a
d
i
o N
e
t
w
orks
Ar
chitectures, Pr
otocols,
an
d
Standar
ds. ISBN: 978-1-
42
00-
777
5-9.
[10]
Yong
ho
ng Z
e
n
g
, Ying-C
han
g
Lian
g. Eigenv
alu
e
base
d
Sp
ectrum Sensi
n
g Algorit
hms for Cogn
itive
Radi
o.
IEEE.
2009.
[11]
Omkar S Vaidya, Vij
a
ya M Kulkar
ni. Anal
ysis
of Ener
g
y
Detection b
a
s
ed Spectrum
Sensi
ng ove
r
Wireless Fad
i
n
g
Cha
n
n
e
ls in
Cog
n
itive R
a
d
i
o Net
w
orks.
Internati
ona
l Jour
nal of E
m
er
gin
g
T
e
chno
lo
g
y
and Adv
anc
ed
Engi
neer
in
g (IJET
AE).
2013; 3.
[12]
Yonghong Zeng, Ying-Chang Li
ang. Max
i
m
u
m-Minim
u
m Eigenvalue De
tection for Cognitiv
e
Radio.
IEEE
. 2007.
[13]
Abi
y
T
e
refe T
e
shome. F
P
GA base
d
Eig
enva
l
ue D
e
tectio
n Algorit
hm for Co
gnitiv
e
Rad
i
o. 201
0.
[14] Shree
Krishna
Sharma, Sy
meon Chatzinotas,
Bj¨orn Ottersten. Eigenv
alue Based Sensing and SNR
Estimation for
Cog
n
itive R
adi
o in Prese
n
ce
of Noise C
o
rrel
a
tion. 20
12.
[15]
Da
ya
n A Guim
ar, Raus
le
y A
A
de S
ouz
a, A
ndr´e
N Barr
eto. Performa
nc
e of C
oop
erati
v
e Eig
env
alu
e
Spectrum Sen
s
ing
w
i
t
h
a Re
alistic Rec
e
iv
e
r
Model u
nder
Impulsive N
o
i
s
e.
J. Sens. Actuator Netw
.
201
3.
Evaluation Warning : The document was created with Spire.PDF for Python.