Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
10
,
No.
3
,
June
2020
,
pp. 3
116~
3124
IS
S
N: 20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v10
i
3
.
pp3116
-
31
24
3116
Journ
al h
om
e
page
:
http:
//
ij
ece.i
aesc
or
e.c
om/i
nd
ex
.ph
p/IJ
ECE
Coopera
tive
-
hyb
rid
det
ection of p
rimary u
ser emul
ators in
cognitiv
e r
adio n
etworks
S. A. Ade
bo,
E. N
.
O
nwuka
, A.
U.
Usm
an, and
A. J
. On
uma
n
yi
Depa
rtment
o
f
T
el
e
comm
unic
at
i
on
Engi
n
ee
ring
,
Feder
al Unive
rsi
t
y
of Te
chnol
og
y
,
Niger
ia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Ma
r 8
, 2
019
Re
vised
N
ov 29
,
2019
Accepte
d
Dec
10, 201
9
Prim
ar
y
user
e
m
ula
tor
(PU
E)
at
t
ac
k
oc
cur
s
in
Cognit
ive
Rad
i
o
Networks
(CRNs
)
when
a
m
al
ic
ious
sec
on
dar
y
user
(SU
)
poses
as
a
primary
user
(PU
)
in
orde
r
to
depr
i
ve
othe
r
l
egi
t
imate
SU
s
the
righ
t
to
fre
e
spe
ct
ra
l
ac
c
ess
for
opportuni
stic
co
m
m
unic
at
ion.
In
m
o
st
ca
ses,
the
se
le
gi
ti
m
at
e
SU
s
are
unable
to
eff
e
ct
iv
ely
de
te
c
t
PU
Es
bec
au
se
the
qualit
y
of
the
signal
s
re
ceive
d
from
a
PU
E
m
a
y
be
seve
re
l
y
a
tt
enu
at
ed
b
y
ch
anne
l
fad
ing
and/
or
shadowing
.
Consequent
l
y
,
i
n
thi
s
pap
er,
we
hav
e
inv
esti
g
at
ed
the
use
of
coope
r
a
tiv
e
spec
trum
sensing
(CSS
)
to
improve
PU
E
detec
t
ion
base
d
o
n
a
h
y
bri
d
loc
a
li
z
at
ion
sche
m
e.
W
e
consid
e
red
diff
ere
n
t
pa
i
rs
of
sec
onda
r
y
users
(SU
s)
over
diffe
ren
t
re
ce
iv
ed
signal
str
engt
h
(RSS
)
val
ues
to
eva
luate
the
ene
r
g
y
eff
iciency
,
accu
racy
,
and
spee
d
o
f
th
e
new
co
oper
ative
sch
eme.
B
ase
d
on
computer
sim
ula
ti
ons,
our
f
indings
suggest
tha
t
a
PU
E
ca
n
be
eff
ectiv
e
l
y
det
e
ct
ed
b
y
a
pa
ir
of
SUs
with
a
low
Root
Mea
n
Square
Err
or
rate
of
0.
0047
eve
n
though
thes
e
SUs
m
a
y
hav
e
cl
ose
RS
S
val
ues
withi
n
the
sam
e
cl
ust
er.
Furthermore,
ou
r
sche
m
e
per
for
m
s be
tt
er
in
te
rm
s of
spee
d,
accur
acy
and
low
ene
rg
y
consum
pti
on
rates
when
compare
d
with
othe
r
PU
E
det
e
ction
sche
m
es.
Thus,
i
t
is
a
v
ia
bl
e
prop
ositi
on
to
be
tt
e
r detect
PU
Es
in CRN
s.
Ke
yw
or
d
s
:
Cl
us
te
r
Hybr
i
d
Pr
im
ary us
er e
m
ula
tor
Seco
nd
a
ry
us
er
s
Sp
ect
r
um
sen
sing
Copyright
©
202
0
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
S. A. A
debo,
Dep
a
rtm
ent o
f Te
le
com
m
un
icati
on
E
nginee
ring
,
Fede
ral U
niv
e
r
sit
y of
Tec
h
no
l
og
y,
Mi
nn
a,
N
i
ger
ia
.
Em
a
il
:
adeb
os
a
m
uel@ya
ho
o.
c
om
1.
INTROD
U
CTION
Cognit
ive
Ra
di
o
(CR)
is
a
n
i
ntell
igent
ra
di
o
that
aut
om
atical
ly
detect
s
fr
ee
c
hannels
(
cal
le
d
w
hite
sp
aces
or
s
pec
trum
ho
le
s)
a
nd
cha
nges
it
s
transcei
ver
pa
r
a
m
et
ers
to
tra
ns
m
it
op
portu
nisit
ic
al
ly
ov
er
these
wh
it
e
s
paces,
wh
il
e
vacati
ng
occ
up
ie
d
c
hannels
to
preve
nt
interfe
re
nce
to
existi
ng
pri
m
a
ry
us
e
rs
(PUs)
[1
-
5]
.
A
PU
ref
e
rs
to
the
li
censed
owne
r
of
the
s
pe
ct
ru
m
(o
r
cha
nn
el
)
w
hile
we
ref
er
to
a
network
of
CR
node
s
as
a
CR
netw
ork
(CRN)
.
CR
Ns
pro
vid
e
se
vera
l
ben
e
fits
to
wireless
com
m
unic
at
ion
s
uch
a
s
i
m
pr
oved
qual
it
y
of
serv
ic
e
by
usi
ng
f
ree
cha
nne
ls,
longer
tra
nsm
issi
on
range
ov
e
r
lo
wer
f
re
qu
e
ncy
ba
nds,
and
im
pr
ove
sp
ect
ra
util
iz
at
ion
[6
]
.
Nev
e
rtheless
,
si
m
i
la
r
to
oth
er
wireless
com
m
un
ic
at
io
n
net
w
orks,
CR
Ns
a
r
e
al
so
su
sce
ptible
to
secur
it
y chall
e
ng
e
s
[
7].
A
m
ajo
r
sec
uri
ty
chall
eng
e
in
CR
Ns
is
the
Pr
im
ary
User
Em
ulator
(PUE)
at
ta
ck
.
P
UE
re
fers
t
o
a
sit
uation
in
wh
ic
h
a
m
a
li
cio
us
sec
onda
ry
us
er
(SU)
or
CR
us
er
fei
gn
s
as
a
le
gitim
a
te
PU
in
orde
r
to
de
ny
oth
e
r
le
gitim
ate
SU
s
in
t
he
CR
N
acce
ss
to
net
work
res
ources,
le
a
ding
to
den
ia
l
of
serv
ic
e
or
ne
twork
flo
od
i
ng
[8
]
.
I
t
is
therefore
per
ti
ne
nt
f
or
l
egitim
at
e
SU
s
to
detect
pot
entia
l
PU
Es
t
o
pre
ven
t
the
m
fr
om
unde
rm
ining
the
e
ntire
CR
N.
It
is
t
he
quest
to
de
velo
p
e
ff
ect
ive
P
UE
detect
ion
m
et
ho
ds
that
m
ot
ivate
d
the
sc
hem
e
pr
opose
d
in
t
his
pa
per.
T
o
reali
ze
our
quest
,
we
co
ns
ide
re
d
the
proces
s
of
s
pe
ct
ru
m
sensing
(S
S
)
,
wh
ic
h
is
pi
vo
ta
l
to
the
su
cces
s
of
CR
te
ch
nolog
y.
SS
deter
m
ines
wh
et
her
wh
it
es
paces
ex
ist
or
not
in
a
s
ense
d
sp
ect
ra
[9,
10]
.
SS
is
al
so
a
n
esse
ntial
too
l
to
determ
ine
wh
et
her
a
P
U
E
exists
or
not
in
a
CR
N.
It
achieves
this
by
al
lowin
g
S
Us
to
loc
al
iz
e
a
po
te
ntial
PU
E
an
d
to
com
par
e
the
PU
E
’s
locat
io
n
with
the
loc
at
ion
of
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Cooper
ative
-
hy
br
id
d
et
ect
io
n of
pr
im
ar
y
us
e
r emulat
or
s i
n
cogniti
ve ra
dio netw
or
k
s
(
S. A
. A
debo
)
3117
the
le
gitim
at
e
PU
.
If
discre
pa
ncies
exist
be
tween
the
l
oca
ti
on
of
the
PUE
an
d
the
le
git
PU
,
t
hen
a
P
UE
is
consi
der
e
d
to
hav
e
been
suc
cessf
ully
detect
ed
by
an
S
U
and
the
bas
e
st
at
ion
ca
n
ea
sil
y
procee
d
t
o
is
olate
it
from
the
network
[
11]
.
H
owe
ver,
since
local
iz
at
ion
de
pend
s
sign
ific
a
ntly
on
si
gn
al
s r
ece
ived
f
r
om
a
po
te
ntial
PU
E,
it
m
ay
beco
m
e
diff
ic
ult
to
detect
PU
Es
w
hose
sign
al
s
hav
e
be
en
seve
rely
under
m
ined
by
channel
eff
ect
s
s
uch
as
m
ulti
path
fad
i
ng
a
nd
sh
a
dow
ing
[
12]
.
Co
nse
qu
e
ntly
,
co
operati
ve
sp
ect
rum
sensing
(CS
S)
ha
s
been
a
dopted
t
o
m
it
igate
su
ch
eff
ect
s
in
C
RNs.
He
re,
S
Us
ar
e
co
nf
i
gure
d
to
us
e
C
SS
to
m
ake
c
om
bin
ed
decisi
ons c
on
c
ern
i
ng the
pr
e
s
ence
of P
UEs
i
n
a CR
N [13,
14].
CSS
in
CR
N
s
can
be
cl
a
ssifie
d
int
o
di
stribu
te
d
a
nd
centrali
zed
s
pectr
um
sensing
[15
-
18]
.
In
distrib
uted
sp
ect
r
um
sensing
(S
S
),
eac
h
SU
pe
rfo
rm
s
s
pectr
um
sensi
ng
in
div
i
dual
ly
and
com
m
un
ic
at
es
the
sense
d
inf
or
m
at
ion
to
ne
ighbou
rin
g
SUs
without
com
m
on
Fu
sio
n
centre
(F
C
).
Dist
rib
uted
SS
re
quire
s
reli
able
com
mu
nicat
io
n
li
nk
s
between
th
e
neig
hbouri
ng
SU
s
an
d
inc
urs
com
m
un
ic
at
i
on
ov
e
r
head
duri
ng
sp
ect
r
um
sens
ed
data
e
xch
a
ng
e
.
Wh
il
e
i
n
centrali
ze
SS
,
a
Fusi
on
ce
ntr
e
(F
C
)
gathe
rs
sense
d
inf
or
m
at
ion
from
all
SU
s
in
the
CR
N
a
nd
us
es
t
hese
inf
or
m
at
ion
to
com
pu
te
the
sensing
sc
he
dule
of
each
S
U
over
a
par
ti
cular
ch
ann
el
,
wh
ic
h
m
akes
it
m
or
e
eff
ic
ie
nt
than
distri
bute
d
spe
ct
ru
m
sensing
[19
-
21]
.
CSS
ha
s
been
us
e
d
f
or
se
ve
r
al
purposes
i
n
CR
Ns,
f
or
exa
m
ple,
autho
r
s
in
[
16]
pro
po
se
d
a
ce
ntrali
zed
coope
rati
ng
s
ensin
g
schem
e
to
est
i
m
at
e
an
op
tim
al
nu
m
ber
of
S
Us
an
d
local
sensing
tim
e
that
gu
ara
ntees
im
pro
ve
perform
ance
in
te
rm
s
of
sensi
ng
delay
and
sp
ect
r
um
utiliz
at
ion
.
I
n
[
17
]
,
authors
m
ini
m
iz
ed
interfere
nce
to
the
P
U
w
hile
m
axi
m
iz
ing
th
e
ex
pected
t
ra
ns
m
issi
on
tim
e
in
a
CR
N.
They
ac
hieve
d
this
by
deter
m
ining
the
op
tim
a
l
decisi
on
thre
shold
f
or
a
gi
ven
false
al
arm
pr
ob
a
bili
ty
us
ing
op
ti
m
al
co
m
bi
ned
ru
le
in
ce
nt
rali
zed
co
op
e
r
at
ive
sensing
sc
hem
e.
Sim
i
la
rly
,
auth
or
s
in
[22]
pro
posed
a
r
ei
nfor
cem
ent
le
arn
i
ng
-
ba
sed
coope
rati
ve
se
ns
in
g
(RLCS)
t
o
re
duce
de
te
ct
ion
over
hea
d
an
d
i
m
pr
ov
e
detect
ion
pe
rfor
m
ance
in
CR
Ns.
Ac
cordin
g
to
[
22
]
,
an
FC
coope
rates
wit
h
ne
ig
hbourin
g
S
Us
to
deter
m
ine
an
opti
m
al
set
of
c
oope
rati
ng
SUs
w
it
h
m
ini
m
u
m
con
t
rol
traff
ic
an
d
le
s
s
sensing
de
la
y.
Essentia
ll
y,
we
note
that
a
sign
ific
ant
a
m
ou
nt
of
re
s
earch
has
bee
n
do
ne
con
ce
r
ning
the
us
e
of
CSS
in
CR
Ns
(see
w
orks
i
n
[
23
-
26]
,
howe
ver,
m
o
st
of
these
sc
hem
es
fo
cuse
d
m
ai
nly
on
m
axi
m
iz
ing
sensing
pa
ra
m
et
ers
in
CSS.
Othe
rs
we
re
con
ce
r
ned
with
opti
m
iz
ing
the
locat
io
n
of
SU
s
to
i
m
pr
ove
detect
ion
pe
rfor
m
ance [21,
26
]
.
Howe
ver,
the
r
e
has
bee
n
li
tt
le
or
no
w
ork
done
c
on
ce
rni
ng
t
he
us
e
of
CSS
t
o
e
ff
ect
ively
detect
PU
Es
i
n
CR
N
s.
Co
ns
eq
ue
ntly
,
in
this
pa
pe
r,
we
ha
ve
inv
e
sti
gated
a
cl
us
te
r
-
base
d
centrali
zed
s
pe
ct
ru
m
sensing
sc
hem
e
to
detect
P
U
Es
with
gr
eat
e
r
accuracy,
sp
ee
d
an
d
lo
we
r
en
erg
y
c
on
s
um
ption
rates.
T
o
a
chieve
this,
we
cl
us
t
ered
SU
s
i
nto
gro
up
s
w
her
e
in
SU
s
in
the
sam
e
c
luster
or
neig
hbouri
ng
cl
us
te
rs
ty
pical
ly
exp
e
rience
si
m
il
ar
sign
al
pr
opa
gation
c
ha
racteri
sti
cs,
wh
ic
h
res
ults
to
si
m
il
ar
Re
c
ei
ved
Si
gn
al
Stren
gt
h
(RSS)
for
SU
s
in
these
cl
us
te
r(
s
).
I
n
this
case,
we
co
ns
ide
r
ed
su
c
h
sim
il
a
rly
gr
ou
ped
S
Us
as
cl
os
el
y
relat
ed.
We
f
ur
the
r
intr
oduce
d
a
hybri
d
schem
e
to
bette
r
local
iz
e
P
UEs
ba
sed
on
a
com
bin
at
ion
of
the
a
ng
le
of arr
iva
l
(AoA)
a
nd
rec
ei
ved
si
gn
al
str
eng
t
h
(RS
S)
m
et
hods
.
O
ur
fin
dings
s
uggest
t
hat
our
sc
hem
e
pro
vid
es
i
m
pr
ov
e
d
perform
ance
in
detect
in
g
P
U
Es
in
CR
Ns.
T
he
rest
of
this
pap
e
r
is
orga
ni
zed
as
f
ollows
.
Sect
ion
2
pre
sent
s
the
m
et
ho
do
l
ogy
an
d
the
m
od
el
f
or
i
nv
est
i
gating
ef
fects
of
c
oope
rati
ve
sensing
on
th
e
hybri
d
L
ocal
iz
at
ion
Schem
e
fo
r
De
te
ct
ion
of
P
rim
ary
User
Em
ulator
in
CR
Ns.
In
sect
ion
3
,
we
pr
ese
nt
the
res
ults
and
d
isc
ussi
on.
Perfo
rm
ance an
al
ysi
s of
our
s
tud
y a
nd concl
us
io
n
a
re
pr
ese
nted
i
n
sect
io
ns 4 a
nd
5
res
pe
ct
ively
.
2.
METHO
DOL
OGY
In
t
his
sect
ion,
we
de
scrib
e
the
syst
em
m
o
del
in
w
hic
h
our
CS
S
ba
sed
hybr
id
sc
hem
e
is
dep
l
oyed
.
We
pr
e
sent
a
gen
e
ral
syst
em
m
od
el
of
the
CR
N,
the
CS
S
sc
hem
e
based
on
e
nergy
de
te
ct
ion
,
our
hy
br
i
d
schem
e and
t
he
ty
pical
opera
ti
on
s
of the e
nt
ire syst
em
are
then
prese
nted
i
n
Fi
gure
1.
Figure
1.
A
ty
pi
cal
CR
N
de
pi
ct
ing
S
Us
co
m
m
un
ic
at
ing
w
it
h
a
n
SB
S in
th
e presence
of a
potenti
al
PUE
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
10
, No
.
3
,
J
une
2020
:
31
16
-
3124
3118
2
.
1.
S
ys
te
m m
od
el
Fig
ure
1
dep
i
ct
s
a
ty
pical
CR
N
consi
sti
ng
of
seve
ral
S
Us
set
up
to
detect
a
PU
E
with
seve
ral
env
i
ronm
ental
ob
sta
cl
es
exis
ti
ng
in
the
en
vi
ronm
ent
that
m
ay
cause
fad
ing
an
d
sh
a
do
wing.
To
e
ff
ec
ti
vely
detect
the
P
U
E,
the
S
Us
a
dopt
CS
S
base
d
on
a
hy
br
id
s
chem
e
co
m
pr
i
sing
of
t
he
rec
ei
ved
si
gn
al
st
rength
(RSS)
a
nd
the
ang
le
of
ar
rival
(Ao
A)
loc
al
iz
at
ion
m
et
ho
ds.
Esse
ntial
ly
,
each
SU
re
cei
ves
dif
fer
e
nt
RSS
values
from
th
e
PU
E.
F
or
ex
a
m
ple,
the
sig
nal
from
the
PU
E
to
S
U1
i
n
Fig
ure
1
is
af
f
ect
ed
by
the
buil
ding
and
th
ough
no
visible
ob
sta
cl
e
m
a
y
exist
bet
ween
the
PUE
and
S
U4
or
SU5,
t
he
RS
S
values
m
ay
be
di
f
fer
e
nt
because
of
at
m
os
phe
ric
co
ndit
ion
s
a
nd
t
he
di
sta
nce
betwee
n
no
des.
We
c
on
si
der
i
n
Fi
g
ure
1
a
fusio
n
centre
(F
C)
dep
ic
te
d
as
the
seco
nda
ry
base
sta
ti
on
(S
BS)
t
hat
c
oor
din
at
es
the
CSS
schem
e
a
m
on
g
the
SUs.
In
thi
s
m
od
el
,
each
SU
senses
the
P
UE’
s
si
gn
al
an
d
repo
rts
it
s
decisi
on
to
the
F
C,
wh
ic
h
the
n
cond
ucts
data
fu
si
on
in
ord
er
t
o
m
a
ke
a
final
deci
sion.
T
his
fi
na
l
decisi
on
is
t
hen
bro
adcaste
d
to
the
SU
s
after
local
iz
a
ti
on
is
con
cl
ud
e
d wit
h t
he
ai
m
to
isolat
e the P
UE.
2
.
2
.
C
ooper
ati
ve
sensin
g sc
heme
The
CSS
sc
he
m
e
con
sist
s
of
SU
s
that
in
div
i
du
al
ly
senses
t
he
P
UE’
s
si
gnal
energy
and
t
hen
eac
h
S
U
sen
ds
it
s
local
decisi
on
to
the
FC,
w
hich
m
a
kes
the
fi
nal
de
ci
sion
. W
e
co
nsi
der
e
d
the
ene
rg
y
detect
or
(E
D)
a
s
the
s
pectru
m
sensing
m
et
ho
d
since
t
he
P
U
E’s
si
gn
al
e
ne
rg
y
is
t
he
on
ly
inf
or
m
at
ion
a
vaila
ble
to
eac
h
S
U.
Con
se
quently
, we m
od
el
ed
t
he
sig
nal en
e
rg
y
of the
P
UE re
cei
ved
at
each
it
h
SU as
:
0
1
(
)
;
()
(
)
(
)
;
i
i
ii
u
m
H
xm
s
m
u
m
H
(1)
wh
e
re
m
=
1,
2,
…
.,
N
is
th
e
tim
e
sa
m
p
le
ind
e
x
an
d
N
is
the
total
nu
m
be
r
of
sam
ples
sense
d
by
eac
h
SU,
x
is
the
sign
al
received
at
the
it
h
SU
,
wh
e
re
i
=
1,
2,
….,
K
,
the
PU
E
sig
nal
at
e
ach
SU
is
de
note
d
a
s
s
m
od
el
ed
as
a
var
ia
ble
with
zero
m
ean
an
d
var
ia
nce
2
s
,
and
ui(m
)
is
m
od
el
ed
as
Additi
ve
W
hite
Ga
us
si
an
No
ise
(
A
WGN
)
wit
h
zer
o
m
ean
a
nd
va
rianc
e
2
u
.
He
re,
K
re
presents
the
num
ber
of
S
Us
i
n
t
he
CR
N
,
H0
a
nd
H1
represe
nt
the
hy
po
t
hesis
that
descr
i
be
s
ei
ther
the
a
bs
e
nce
or
pre
sence
of
P
UE
sign
al
s
in
t
he
CR
N
resp
ect
ively
.
Each
SU recei
ves xi(m
)
an
d com
pu
te
s a tes
t st
at
ist
ic
, w
hic
h rep
rese
nts th
e sig
nal en
e
rg
y
as foll
ows
:
2
1
1
(
)
(
)
N
ii
m
T
X
x
m
N
(2)
Th
us
, we c
om
pu
te
d
t
he
lo
cal
proba
bili
ty
o
f d
et
ect
ion
D
p
at
each
SU as:
1
(
)
/
i
D
i
i
P
P
T
X
H
22
22
(
(
)
)
2
i
s
u
su
N
Q
(3)
and the
pro
ba
bi
li
t
y of
false al
arm
FA
P
at
each
S
U as:
0
(
(
)
/
)
i
F
A
i
i
P
P
r
T
X
H
2
2
()
2
iu
u
N
Q
(4)
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Cooper
ative
-
hy
br
id
d
et
ect
io
n of
pr
im
ar
y
us
e
r emulat
or
s i
n
cogniti
ve ra
dio netw
or
k
s
(
S. A
. A
debo
)
3119
The
pro
bab
il
it
y of m
issed d
et
ect
ion
M
p
can as
wel
l be calcula
te
d
as
:
1
ii
MD
pp
w
he
re
()
Q
is
the M
arcu
m
Q
-
functi
on g
i
ven as:
2
2
1
()
2
t
dt
x
Q
x
e
a
nd
i
is t
he
local
detect
io
n
th
res
ho
l
d
at
eac
h S
U ob
ta
ine
d fro
m
(
4)
as
:
21
2
2
()
i
u
F
A
i
N
u
QP
Each
S
U
t
hen
sen
ds
it
s
loca
l
detect
ion
sta
ti
sti
cs
to
the
FC,
w
hich
pla
ys
a
vital
ro
le
in
the
C
S
S
schem
e.
Ba
sed
on
the
local
sta
ti
sti
cs
recei
ved
at
the
FC
fr
om
K
par
ti
c
ipati
ng
S
Us,
t
he
FC
de
no
te
s
as
the
total
num
ber
of
S
Us
t
hat
ha
ve
dete
ct
ed
the
PUE
.
It
the
n
a
do
pts
a
decisi
on
strat
egy
descr
i
bed
accor
ding t
o
[
27]
as
:
0
1
,
i
f
,
i
f
HM
HM
(
5
)
The
FC
decide
s
on
the
fi
nal
pro
ba
bili
ty
of
detect
ion
an
d
prob
a
bili
ty
of
false
al
ar
m
based
on
M
dif
fer
e
nt
local
sta
ti
sti
cs as f
ol
lows
:
1
1
K
K
Km
m
D
D
D
m
m
P
P
P
(6
)
1
1
K
K
Km
m
F
A
F
A
F
A
m
m
P
P
P
(7
)
2
.
3
.
H
yb
ri
d
l
oc
aliz
at
ion s
c
h
eme
The
FC
us
es
t
he
se
ns
ed
inf
orm
ation
f
ro
m
each
S
U
to
l
oc
al
iz
e
the
PU
E
.
To
ac
hie
ve
this,
it
ad
opts
a
hybr
i
d
of
th
e
RSS
an
d
an
gle
of
arr
i
val
(
AoA)
m
et
ho
ds
to
detect
the
PU
E.
Fig
ure
2
il
lustrate
s
a
set
up
of
a
nu
m
ber
of
S
Us
ai
m
ing
to
de
te
ct
a
PU
E.
H
ere,
the
FC
gr
oups
the
diff
e
r
ent
SU
s
int
o
re
sp
ect
ive
pai
rs
wh
e
re
each
pair
ai
m
s
to
detect
the
PU
E.
We
desc
ribe
t
he
hybr
i
d
locat
io
n
sch
e
m
e
(H
LS
)
for
a
par
ti
cula
r
pair
a
s
fo
ll
ows
[
28]
:
in
Fig
ure
3,
le
t
x
1
,
y
1
and
x
2
,
y
2
de
no
te
t
he
res
pecti
ve
po
sit
io
ns
of
SU
1
and
SU
2
.
Sim
i
la
rly,
le
t
r
1
and
r
2
repres
ent the
ra
dii of
the cove
rag
e
ar
eas o
f
SU
1
an
d
SU
2
. Lin
e
D
c
onnects
the cent
res of
SU
1
and
SU
2
,
w
hile
∅
an
d
θ
are
the
resp
e
ct
ive
an
gles
from
wh
ic
h
the
le
gitim
a
te
PU
’s
si
gn
al
a
rr
i
ves
at
SU
1
and
SU
2
.
The
a
ng
le
s
α
1
an
d
α
2
represen
t
t
he
ang
le
s
at
w
hich
the
PU
E
’s
s
ign
al
ar
rives
at
SU
1
an
d
SU
2
.
Let
the
po
sit
io
n
of
the
le
gitim
at
e
PU
be
(
X
PU
,
Y
PU
)
and
the
posit
ion
of
t
he
PUE
be
(
X
e
,
Y
e
).
The
Eucli
dea
n
distanc
e
D
betwee
n
the
pa
ir of
par
ti
ci
pating S
Us
is
obta
ined
as
:
22
2
1
2
1
(
)
(
)
D
x
x
y
y
(8
)
1
1
1
t
a
n
ee
Y
y
X
x
(9
)
2
2
2
t
a
n
ee
Y
y
X
x
(10
)
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
10
, No
.
3
,
J
une
2020
:
31
16
-
3124
3120
1
1
2
2
2
1
12
(
t
a
n
)
(
t
a
n
)
t
a
n
t
a
n
e
x
x
y
y
X
(11
)
SU
1
SU
2
SU
3
SU
4
PU
E
SU
5
S
B
S
c
o
m
m
u
n
i
c
a
t
i
o
n
b
e
t
w
e
e
n
S
U
a
n
d
S
BS
P
U
E
s
i
g
n
a
l
Figure
2. A
setup o
f
S
Us
coo
per
at
in
g
t
o detec
t a p
otentia
l
PU
E
Con
se
quently
,
w
hen
a
pair
of
S
Us
recei
ve
the
sig
nal
from
a
pote
ntial
PU
E
,
fir
st,
each
pai
r
coope
rates
to
com
pu
te
the
locat
ion
of
t
he
transm
itter
u
sing
(
14)
a
nd
(
15)
.
The
n,
this
est
i
m
at
ed
transm
it
te
r
locat
ion
is
co
m
par
ed
with
the
know
n
loca
ti
on
of
the
le
gitim
a
te
PU
.
If
the
transm
it
t
er’
s
locat
ion
is
diff
e
rent
from
the
le
git
i
m
at
e
PU
’s
lo
cat
ion
,
the
tra
ns
m
itter
is
co
ns
ide
red
a
PUE.
Othe
rw
ise
,
it
is
con
sider
ed
a
s
a
le
gitim
at
e
P
U
a
nd
so
the
S
Us
qu
ic
kly
va
cat
e
the
spe
ct
r
um
to
avo
i
d
in
te
rf
ere
nce.
The
detect
ion
res
ul
ts
are
sent
to
the
FC
wh
e
re
final
det
ect
ion
is
con
cl
ud
e
d
base
d
on
the
decisi
on
s
trat
egy
in
(8
)
.
A
si
m
ple
strateg
y
to
cl
us
te
r
the
S
Us
in
a
CR
N
is
dep
ic
te
d
in
Fig
ure
4.
Her
e
,
the
SBS
form
s
fo
ur
dif
fer
e
nt
cl
ust
ers
wh
e
re
eac
h
SU
form
s
a
pair
a
nd
c
omm
un
ic
at
es
this
pairing
inform
at
ion
t
o
the
SBS.
Es
sentia
ll
y,
an
SU
can
f
or
m
only
on
e
pair
pe
r
ti
m
e
and
i
n
a
sit
uation
w
her
e
the
re
are
odd
num
ber
of
SUs
in
the
CR
N,
the
S
BS
si
m
ply
exclud
es
the last
S
U
that
f
ai
ls t
o f
or
m
a p
ai
r.
SU
6
SU
7
SU
8
SU
9
SU
10
SU
11
SU
1
SU
2
SU
3
SU
4
SU
5
Cl
u
s
t
e
r
1
Cl
u
s
t
er
2
Cl
u
s
t
e
r
3
Cl
u
s
t
e
r
4
C
l
o
s
e
n
ei
g
h
b
o
u
r
s
i
n
d
i
ff
er
en
t
c
l
u
s
t
er
s
C
l
o
s
e
n
e
i
g
h
b
o
u
r
s
i
n
t
h
e
s
a
me
c
l
u
s
t
e
r
S
B
S
Figure
3
.
The
two sec
ondar
y
us
ers
parti
ci
pat
ing
i
n
the
detect
ion PU
E
Figure
4. Cl
us
t
erin
g of
SUs
by
a secon
dar
y
base
sta
ti
on
(sam
e a
s FC)
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Cooper
ative
-
hy
br
id
d
et
ect
io
n of
pr
im
ar
y
us
e
r emulat
or
s i
n
cogniti
ve ra
dio netw
or
k
s
(
S. A
. A
debo
)
3121
2
.
4
.
Gener
al
op
er
at
i
on
of the co
op
er
ati
ve
-
h
yb
ri
d l
oca
l
iz
at
ion s
che
m
e (
C
-
HL
S)
Essentia
ll
y,
each
S
U
receive
s
the
RS
S
values
from
the
PU
E
t
ran
sm
it
t
er.
Using
t
he
RSS,
eac
h
S
U
est
i
m
at
es
their
resp
ect
ive
di
sta
nces
from
t
he
PUE
trans
m
itter
and
the
arr
ival
an
gle
of
the
sig
nal
us
in
g
the
HL
S
sc
he
m
e.
Our
HLS
is
then
us
e
d
to
local
iz
e
t
he
P
UE
base
d
on
t
he
distance
a
nd
the
angula
r
m
easur
em
ents.
Diff
e
ren
t
pair
s
of
S
Us
distribu
te
d
withi
n
c
lusters
in
the
CR
N
are
use
d
to
local
iz
e
the
PUE
with
the
ai
m
t
o
inc
rease
dete
ct
ion
acc
ur
acy
.
A
pair
of
S
Us
is
sel
ect
ed
by
the
SBS
us
i
ng
the
RSS
of
eac
h
S
U
receiv
e
d
withi
n
a
po
wer
i
nter
val
[
0
,
]
from
K
pa
rtic
ipati
ng
S
Us
. W
e d
escri
be
a
li
st
of
possibl
e
pairs
that
ca
n
be
sel
ect
ed
as
f
ollows:
-
Tw
o
SU
s
with
m
axi
m
u
m
RS
S:
In
this
case,
the
SBS
sel
ect
s
fr
om
a
m
on
g
al
l
c
luster
s
the
two
nodes
w
it
h
the
hi
gh
e
st R
S
S v
al
ues. Thi
s
i
m
plies t
hat the two no
des
ca
n be select
ed
fr
om
d
iffer
e
nt cl
us
te
rs
.
-
Tw
o
SU
s
wit
h
m
ini
m
u
m
RS
S:
The
SBS
sel
ect
s
two
node
s
with
the
l
owest
RSS
values
fr
om
a
m
on
g
al
l
cl
us
te
rs.
-
Tw
o
SUs
with
m
edium
RSS:
The
SBS
c
ompu
te
s
th
e
ave
ra
ge
RSS
v
al
ues
from
all
no
de
s
in
the
CR
N
a
nd
sel
ect
s the tw
o nodes
h
a
ving t
he
cl
ose
st val
ue
s to
t
he
a
ver
a
ge
RSS
value.
-
Tw
o
S
Us
with
one
ha
ving
t
he
highest
RSS
an
d
the
oth
e
r
hav
i
ng
the
lo
west
RSS:
T
he
SBS
sel
ect
s
two
nodes
w
it
h o
ne
h
a
ving the
h
i
ghest
RSS
and t
he othe
r havi
ng the
lo
west R
S
S
values
in
t
he
CR
N.
-
Tw
o
SU
s
that
are
cl
os
el
y
relat
ed:
Her
e,
the
SBS
sel
ect
s
two
node
s
with
the
two
highest
RSS
values
f
r
om
the sam
e cluste
r.
Our
ai
m
is
to
inv
e
sti
gate
the
best
pair
of
S
U
s
that
ca
n
m
os
t
eff
e
ct
ively
det
ect
the
pr
ese
nc
e
of
P
UEs
in
a
CR
N
us
in
g o
ur
C
-
H
LS sc
hem
e.
2
.
5
.
Perf
orm
ance
metric
We
evalu
at
ed
the
accuracy
of
the
P
UE
lo
cal
iz
at
ion
sche
m
e
us
ing
the
Roo
t
Me
an
S
qu
a
re
Er
ror
(RMSE)
f
un
ct
i
on d
e
fine
d
as
:
2
1
()
c
c
e
s
t
r
e
a
l
c
LL
R
M
S
E
c
(12
)
w
he
re
an
d
ar
e
the
est
im
at
e
d
a
nd
act
ual
l
oc
at
ion
of
the
P
UE,
an
d
C
de
note
s
the
num
ber
of
Mo
nte
Ca
rlo
tria
ls
ov
er
wh
ic
h
the
sim
ula
ti
on
was c
onduct
ed.
3.
RESU
LT
S
AND A
N
ALYSIS
In
t
his
sect
io
n,
we
disc
us
s
our
fin
dings
c
on
ce
r
ning
the
us
e
of
the
C
-
HL
S
ov
e
r
dif
fer
e
nt
pai
r
-
sel
ect
ion
sc
he
m
es.
Our
sim
ulati
on
was
c
on
du
ct
e
d
usi
ng
MATLAB
ve
r
sion
2017
b.
He
re,
S
Us
wer
e
r
ando
m
l
y
distrib
uted
ov
e
r
a
sp
at
ia
l
network
of
100m
x
100m
:
The
po
sit
io
n
of
a
pair
of
SUs
re
la
ti
ve
to
the
PU
E’
s
locat
ion
wer
e
var
ie
d
over
dif
fer
e
nt
Mon
te
Ca
rlo
sim
ulatio
n
a
ver
a
ge
d
over
1000
tria
ls
(i.e.
C
=
1000
in
(15
)).
The
tra
ns
m
it
po
we
r
of
the
P
UE
was
fixe
d
at
50dBm
and
pathloss
was
c
om
pu
te
d
us
in
g
the
f
ree
-
sp
ace
m
od
el
for
a
ref
ere
nce
distance
of
1m
and
loss
ex
po
nen
t
of
4,
co
nsi
der
in
g
ty
pical
urba
n
env
i
ron
m
ent
s.
Her
e,
w
e
no
te
that schem
es wi
th low
e
r
RM
S
E v
al
ues
ty
pica
ll
y im
pl
y bett
er acc
ur
acy
.
Figure
5
pr
ese
nts
the
accu
rac
y
per
f
or
m
ance
of
the
CSS
-
HL
S
us
in
g
a
pair
of
S
Us
with
m
ini
m
u
m
and
m
axi
m
u
m
,
m
i
nim
u
m
,
m
edian,
hi
gh
e
st,
an
d
cl
os
el
y
relat
e
d
RSSs.
Our
fi
nd
i
ngs
i
nd
ic
at
e
that
the
accuracy
of
the
C
-
HLS
over
di
ff
e
ren
t
pa
ir
-
sel
ect
ion
sc
hem
es
increases
as
the
pair
of
SUs
co
ntinuo
us
ly
rec
om
pu
te
s
the
locat
io
n
of
the
P
UE
ov
e
r
t
i
m
e.
As
ex
pect
ed,
Fig
ur
e
5
s
hows
that
sel
ect
ing
t
wo
SUs
w
it
h
the
highest
RSS
values
c
onve
rged
t
o
a
n
RM
SE
value
of
0.0
06
i
n
0.0
7
secs
.
Using
this
sel
ect
ion
sc
hem
e
i
m
plies
that
SU
s
that
receive
PUE
s
ign
al
s
via
the
best
c
hannels
(least
fa
ding
e
ff
ect
s)
ge
ne
rall
y
le
a
ds
to
im
pro
ved
pe
rform
ance
.
Figure
5
furth
er
s
hows
that
the
pai
r
-
sel
ect
ion
sc
hem
e
of
SU
s
with
the
le
ast
(m
ini
m
u
m
)
RSS
value
s
a
nd
the
pair
sc
hem
e
with
m
edian
RSS
values
ty
pical
ly
con
ve
r
ged
to
a
n
RM
SE
value
of
0.008
1
an
d
0.0
068
a
fter
0.08
secs
,
res
pecti
vely
.
Thi
s
i
m
plies
that
us
in
g
PU
s
w
it
h
s
m
al
l
RSS
values
(
poor
channel)
co
ndit
ion
s
ty
pical
ly
r
edu
c
es
PU
E
detect
ion
perform
ance.
The
le
ast
perform
ance
oc
curred
w
hen
us
in
g
tw
o
S
U
s
with
m
axi
m
u
m
and
m
ini
m
u
m
RSS
values
res
ulti
ng
i
n
an
RM
S
E
value
of
0.0
13
a
fter
0.09
s
ecs.
This
im
pli
es
that
us
in
g
the
m
ini
m
u
m
RSS
values
in
a
pair
c
om
bin
at
ion
m
a
y
no
t
necess
a
r
il
y
gu
aran
te
e
the
best
pe
rform
ance
since
detect
io
n
pe
rfor
m
ance
m
ay
a
lso
be
aff
ect
ed
by
poorl
y
est
i
m
a
t
ed
A
oAs,
th
us
neg
at
ively
a
ff
ect
in
g
the
pe
rfor
m
ance
of
t
he
pair
schem
e
.
An
in
te
resti
ng
fin
din
g
i
n
Fig
ur
e
5
su
ggest
s
that
two
S
Us
with
cl
os
el
y
relat
ed
RSS
s
conve
rg
e
d
the
fastest
to
a
n
RM
SE
value
of
0.0
047
a
f
te
r
0.0
2
secs.
This
pair
ac
hiev
e
d
the
hi
ghest
acc
ur
acy
at
t
he
fas
te
st
rate
beca
use
they
had
the
highest
R
SS
va
lues
f
r
om
wit
hin
t
he
sam
e
cl
us
te
r
.
Fu
rt
her
m
or
e,
s
ince
the
tw
o
S
Us
with
relat
ed
RSS
a
re
not
a
s
far
a
par
t
a
s
the
m
axi
m
u
m
RSS
schem
e
from
diff
e
re
nt cluste
rs,
t
hey ty
pical
ly
ex
pe
rience l
ess p
at
hlo
ss
leadin
g
to
b
et
te
r
perform
ance th
an othe
r
sc
hem
es.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
10
, No
.
3
,
J
une
2020
:
31
16
-
3124
3122
Figure
6
pr
e
se
nts
the
total
en
erg
y
co
nsum
e
d
by
each
pair
ing
sc
hem
e
ov
er
the
co
nver
ge
nce
tim
e
to
their
res
pecti
ve
m
ini
m
u
m
R
MSE
values
.
Our
fi
nd
i
ng
s
ind
ic
at
e
that
the
cl
os
el
y
-
relat
ed
pairi
ng
sc
hem
e
consum
ed
the
le
ast
ener
gy
be
cause
it
con
ve
rg
e
d
fastest
to
it
s
m
ini
m
u
m
RM
SE
value.
Essentia
ll
y,
F
igure
6
su
ggest
s
tw
o
i
nteresti
ng
obs
erv
at
io
ns
as
fol
lows
:
the
C
-
HS
L
sc
hem
e
c
an
be
us
e
d
ba
sed
on
the
cl
ust
erin
g
appr
oach
as
w
el
l
as
with
ou
t
cl
us
te
rin
g.
I
n
t
he
cl
us
te
rin
g
c
ase,
it
is
s
ugge
ste
d
that
SU
pa
irs
wit
h
the
hi
gh
est
RSS
val
ues
s
houl
d
be
sel
ect
ed
f
ro
m
the
s
a
m
e
cl
us
te
r,
a
s
this
pro
duce
s
i
m
pr
ove
d
pe
rfor
m
ance.
H
oweve
r,
in
a
non
-
cl
us
te
red
CR
N
,
it
is
su
ggest
e
d
that
the
m
axi
m
u
m
RSS
pairi
ng
sc
hem
e
sh
ou
ld
be
adopted
t
o
achieve
the b
e
st pe
rform
ance.
Figure
5.
Com
par
at
ive
p
e
rfo
r
m
ance of th
e
C
-
HL
S
schem
e u
sing t
wo S
Us wit
h
l
ow
est
,
m
edi
u
n,
highest,
cl
os
el
y rela
te
d
RSS
Figure
6.
Ene
r
gy Co
nsum
ption
of the C
-
HL
S schem
e
us
in
g
t
wo SUs
with lo
west,
m
edian, hi
gh
e
st,
cl
ose
ly
relat
ed
RSS
s
3.1.
Perf
orm
ance
a
na
l
ys
is
The
perform
a
nce
of
the
pr
opos
e
d
Im
pro
ved
-
hybr
i
d
D
e
te
ct
ion
of
P
rim
ary
User
E
m
ula
tors
in
Cognit
ive
Ra
di
o
Netw
o
r
ks
was
m
easur
ed
us
in
g
root
m
ean
squar
e
error
(RMS
E)
as
sh
ow
n
in
Fig
ure
5.
Si
m
il
arly
,
per
f
or
m
ances
of
s
om
e
sche
m
es
for
the
detect
i
on
of
pr
im
ary
us
er
em
ulator
s
in
cogniti
ve
rad
io
netw
orks
we
re
evaluated
us
i
ng
RM
SE
.
Th
e
per
f
orm
ance
of
the
propos
ed
Co
op
e
rati
v
e
-
hy
br
id
Detec
ti
on
of
Pr
im
ary
User
Em
ulators
in
Cognit
ive
R
adio
Netw
ork
s
is
bette
r
th
an
the
perfor
m
ances
of
t
he
hybri
d
schem
e
[2
8],
AoA
sc
hem
e
[2
9],
a
nd
RSS
schem
e
[3
0]
presente
d
in
Ta
ble
1.
N
otice
that
our
c
oope
r
at
ive
-
hybri
d
sc
hem
e d
em
on
strat
es
hi
gh
e
r
ac
c
ur
acy
than
R
SS,
A
oA
hybr
i
d of RS
S and A
oA as i
t exh
i
bits the l
ow
est
RM
SE
of
0.0
047.
More
over
,
it
exh
ibit
s
high
er
sp
ee
d
an
d
e
nergy
eff
ic
ie
nc
y
than
the
m
e
t
hods
us
ed
i
n
[
29,
30]
as
it
ta
kes
le
sser
nu
m
ber
of
it
erati
on
s
to
at
ta
in
converge
nc
e.
This
resu
lt
s
are
qu
it
e
sig
nificant
beca
use
sp
eed
and
acc
ur
acy
are
ve
ry
i
m
po
rta
nt
for
ef
fici
ent
sp
ect
r
um
util
iz
a
ti
on
.
Fu
rt
her
m
or
e,
the
need
for
energy
eff
ic
ie
ncy
ca
nnot
be
ov
e
rem
ph
a
siz
ed
in
re
al
iz
ing
co
gnit
ive
ra
dio
te
ch
nolo
gy,
c
on
si
de
rin
g
the
num
ber
of
dev
ic
es
that
will
f
l
ood
t
he netwo
rk in
futu
re.
Com
par
iso
n o
f
local
iz
at
ion s
chem
es
sh
own
in Ta
ble 1.
Table
1.
C
om
par
iso
n of
local
iz
at
ion
sc
hem
es
Detectio
n
Sche
m
e
Nu
m
b
e
r
o
f
I
te
ratio
n
s
RMSE
RSS [
2
9
]
50
0
.22
0
0
Ao
A [
3
0
]
30
0
.01
2
0
The H
y
b
rid o
f
RSS and
AoA [
2
8
]
20
0
.00
5
0
The
Co
o
p
erative
-
h
y
b
rid Sch
e
m
e
20
0
.00
4
7
4.
CONCL
US
I
O
N
In
this
pa
per
,
we
ha
ve
pr
ese
nted
a
c
oope
rati
ve
-
hybr
i
d
loca
li
zat
ion
schem
e
(C
-
H
LS)
t
o
im
pr
ov
e
P
U
E
detect
ion
in
C
RN.
The
C
-
H
SL
schem
e
was
inv
est
igate
d
co
ns
ide
rin
g
diff
e
ren
t
pair
ing
ap
proac
he
s
with
the
ai
m
to
determ
ine
wh
ic
h
pair
ac
hieves
the
best
pe
r
for
m
ance.
W
e
an
al
yz
ed
the
C
-
HLS
schem
e
ba
sed
on
the
acc
ur
acy
a
nd
ene
r
gy
co
nsum
ption
rate
of
the
sc
hem
e
a
s
a
f
unct
io
n
of
tim
e.
Ou
r
fi
nd
i
ng
s
in
dicat
e
th
at
tw
o
SU
s
with
cl
ose
ly
relat
ed
RS
S
val
ues
best
loca
li
zes
a
P
U
E
in
te
rm
s
of
accuracy,
ene
r
gy
co
nsum
ption
a
nd
sp
ee
d.
Ne
ve
rth
el
ess,
our
sc
he
m
e
m
a
y
ben
efi
t
fu
rt
her
from
e
m
plo
yi
ng
bette
r
sp
ect
ru
m
sensing
m
et
ho
ds
and
inco
rpor
at
in
g a
dap
ti
ve
th
res
hold tech
niques i
n
the
E
D,
w
hich wil
l be c
onsidere
d
i
n
f
uture
works.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Cooper
ative
-
hy
br
id
d
et
ect
io
n of
pr
im
ar
y
us
e
r emulat
or
s i
n
cogniti
ve ra
dio netw
or
k
s
(
S. A
. A
debo
)
3123
ACKN
OWLE
DGE
MENTS
This
researc
h
was
s
uppo
rted
by
the
TET
FUND
I
ns
ti
tuti
on
-
Ba
sed
Re
sear
ch
In
te
rv
e
ntio
n
(I
BR
I)
f
und
of the
Fede
ral
Un
i
ver
sit
y o
f Te
ch
no
l
og
y,
Mi
nn
a,
N
i
ger
ia
(
TET
FUN
D/F
UTMI
NNA/2
016
-
2017/6th
BR
P/15).
REFERE
NCE
S
[1]
S.
Srinu,
S.
L
.
S
aba
t
,
and
S.
K
.
Udgata
,
“
Spect
r
um
Sensing
using
Freque
nc
y
Do
m
ai
n
Ent
rop
y
E
stim
at
ion
a
nd
i
ts
FPGA
I
m
ple
m
ent
at
io
n
for
Cognit
ive Ra
d
io
,
”
Pr
oce
dia
Engi
n
ee
r
ing,
vo
l. 30, pp.
289
–
296,
2012
.
[2]
P.
Li
n,
Y.
Chen
,
P.
Chang
and
S.
Jeng,
"Coopera
t
i
ve
Spect
rum
Sensing
and
Optimiza
t
ion
on
Multi
-
Antenna
Ene
r
g
y
Dete
c
ti
on
in
Ra
y
l
ei
gh
Fading
C
hanne
l
,
"
27th
Wirel
ess
and
Optic
al
Comm
unic
ation
Confe
renc
e
(
WOCC)
,
Huali
e
n
,
pp.
1
-
5
,
2018
.
[3]
A.
J.
Onum
an
y
i
,
E.
N.
Onw
uka,
A.
M.
Aibinu,
O
.
C.
Ugw
ej
e
,
an
d
M.
J.
E.
Sal
am
i,
“
A
Modifie
d
Otsu’s
Algorit
hm
for
Im
proving
the
Perform
anc
e
of
the
En
erg
y
Dete
c
tor
in
C
ognit
ive
R
adi
o
,
”
AEU
-
Int
ernational
Journal
of
El
e
ct
ronics
and
Comm
unic
ati
ons,
vol
.
79
,
pp
.
53
–
63,
Sep
.
2017
.
[4]
C.
Xing,
Z
.
-
s.
BIE,
and
W
.
-
l.
W
U,
“
Dete
ct
ion
Eff
i
ci
en
c
y
of
Coopera
t
ive
Spe
ct
rum
Sensing
i
n
Cognit
iv
e
R
a
di
o
Network,
”
The
J
ournal
of
Ch
ina U
niv
ersities of
P
osts and
Teleco
mm
unic
ati
ons,
v
ol.
15
,
no
.
3
,
pp
.
1
–
7,
2008
.
[5]
A.
Bal
ieiro,
P.
Yos
hioka
,
K.
Dias,
D.
Cava
lcant
i,
and
C.
Cordeiro,
“
A
Multi
-
Objective
Gene
ti
c
Optimiza
ti
o
n
for
Spect
rum
Sensin
g
in
Cogn
it
iv
e
R
adi
o
,
”
Ex
p
ert
Sy
stems wit
h
Appl
i
cat
ions
,
vol
.
41
,
no.
8
,
pp
.
3640
–
3650,
2014
.
[6]
A.
J.
Onum
an
y
i
,
E.
N.
Onw
uka,
A.
M.
Aibinu, O.
C.
Ugw
eje, a
nd
M.
J.
E
.
Sal
ami,
“
A
Rea
l
Valu
ed
Neura
l
Ne
two
rk
Based
Autore
gr
essive
Ene
rg
y
Dete
c
tor
for
Cognit
ive
R
adi
o
Applic
ation
,
”
Inte
rnational
S
c
holarly
Re
searc
h
Noti
ce
s
,
vol
.
201
4,
pp
.
1
–
11
,
201
4.
[7]
A.
J.
Onum
an
y
i,
A.
M.
Abu
-
Mahfouz
,
and
G
.
P.
Hanc
ke
,
“
A
Co
m
par
at
ive
Anal
ysis
of
Local
and
Global
Adap
tiv
e
Thre
shold
Esti
m
at
ion
Techni
que
s
for
Ene
rg
y
D
e
te
c
ti
on
in
Cogni
ti
ve
R
adi
o
,”
Ph
ysic
al
Comm
uni
cat
ion
,
vol
.
29
,
no.
C,
pp.
1
–
11,
Apr.
2018.
[8]
K.
Akbari
and
J.
Abouei
,
“
Signal
Cl
assificat
io
n
for
Det
ecting
Prim
ar
y
user
Emulat
ion
Attac
k
in
C
ent
ra
li
z
e
d
Cognit
ive
Rad
io
Networks,
”
Ele
ct
rical E
ng
ineering
(
ICEE
)
,
Iranian
Confe
ren
ce o
n.
IE
EE,
pp
.
342
–
347,
2018
.
[9]
R.
Yu,
Y.
Zha
ng
,
Y.
Li
u,
S.
Gjes
sing,
and
M.
Guiza
n
i,
“
Secur
ing
Cognit
ive
Rad
io
Networks
Against
Prim
ary
user
Emulat
ion
Attac
ks
,”
IE
EE Ne
two
rk,
vol
.
29
,
no
.
4
,
pp
.
68
–
74
,
201
5.
[10]
Y.
-
C.
Li
ang
,
Y.
Ze
ng,
E
.
C.
Peh,
and
A.
T.
Hoang,
“
Sensing
-
Throughput
Tra
deof
f
for
Cognit
ive
Radi
o
Networks,
”
IEE
E
transacti
o
ns on
Wireless
C
omm
unic
ati
ons,
vol.
7
,
no
.
4
,
pp
.
1326
–
1337,
200
8.
[11]
E.
Orum
wense,
O.
O
y
er
inde,
an
d
S.
Mnene
y
,
“
Im
pac
t
of
Prim
ar
y
user
Emulation
Attacks
on
Cognit
ive
R
adi
o
Networks,”
In
te
r
nati
onal Journal
on
Comm
unic
at
ions A
nt
enna
an
d
Propagati
on
,
v
ol.
4
,
no
.
1
,
pp
.
1
9
–
26,
2014
.
[12]
A.
As
hokan
and
L.
Jac
ob,
“
Distribut
ed
cooperat
iv
e
spec
trum
sensing
with
m
ul
ti
ple
co
al
i
tions
and
non
-
ide
al
rep
orti
ng
cha
nn
el
,
”
in
S
ignal
Proce
ss
ing,
In
f
orm
ati
cs,
Comm
unic
ati
on
a
nd
Ene
rgy
S
yste
m
s
(
SPICE
S)
,
IEE
E
Inte
rna
ti
onal
Co
nfe
renc
e
on
.
I
EEE,
pp
.
1
–
6
,
2017
.
[13]
G.
Sharm
a
and
R.
Sharm
a,
“
Distribut
ed
Coo
per
ative
Spe
ct
ru
m
Sensing
Over
Diffe
ren
t
Fad
ing
Channels
i
n
Cognit
ive
Radi
o
,
”
in
Com
pute
r,
Com
m
unic
at
ions
and
Elec
troni
c
s
(Com
pte
li
x),
2
017
In
te
rnationa
l
Confe
ren
ce
on
.
IEE
E
,
pp
.
107
–
1
11,
2017
.
[14]
I.
F.
Ak
y
il
d
iz
,
B.
F.
Lo,
and
R.
Bal
akr
ishnan
,
“
Coopera
ti
ve
Spect
rum
Sensing
in
Cognit
ive
Radi
o
Networks:
A
Surve
y
,
”
Phy
s
ic
al
communic
at
ion,
vo
l. 4, no. 1, pp. 40
–
62,
2011
.
[15]
S.
A.
Alvi
,
M.
S.
Younis,
M.
I
m
ran
,
et
a
l
.
,
“
A
weight
ed
li
ne
ar
combining
sc
heme
for
coop
e
rat
iv
e
spec
trum
sensing,
”
Procedia
Computer
S
c
ie
nc
e,
vol. 32, p
p.
149
–
157
,
201
4.
[16]
D.
Tre
eumnuk, S
.
L.
Mac
don
al
d
,
and
D.
C.
Pope
scu,
“
Optimizi
n
g
per
form
anc
e
o
f
coope
ra
ti
ve
se
nsing f
or
inc
re
ase
d
spec
trum
uti
l
iz
a
t
ion
in
d
y
namic
c
ognit
ive
rad
io
s
ystems
,
”
in
2013
IEE
E
In
te
rnatio
nal
Confe
ren
c
e
Comm
unic
ati
ons
(
ICC
)
on
IEE
E
,
pp.
4656
–
4660
,
2013.
[17]
L.
Yang
,
S.
Song,
and
K.
Le
t
ai
ef
,
“
Optimizi
n
g
Spect
rum
Sensing
Eff
iciency
in
Cognit
iv
e
Ra
dio
Networks
,”
Computing,
Co
mm
unic
at
ions a
nd
Applications
Confe
ren
c
e
(
ComCom
Ap)
,
IEE
E
,
pp
.
262
–
266
,
2
012.
[18]
S.
Shrivasta
va
and
D.
Kothar
i
,
“
Su
throughpu
t
enha
nc
ement
in
a
decision
f
usion
base
d
co
oper
ative
sensin
g
s
y
stem,”
A
EU
-
Inte
rnational
Jou
rnal
of El
e
ct
roni
cs
and
Comm
uni
cat
ions,
vol
.
87
,
pp.
95
–
100
,
201
8.
[19]
G.
Verm
a,
V.
Dhage
,
and
S
.
S.
Chauha
n,
“
Anal
y
s
is
of
Com
bine
d
Dat
a
-
Dec
i
sion
Fus
ion
Sch
eme
for
Cognit
i
ve
Radi
o
,
”
2nd
Inter
nati
onal
Conf
ere
nce on
In
ve
n
t
ive
Syst
ems and
C
ontrol
(
ICISC)
,
IEE
E
,
pp
.
1324
–
1327
,
2018
.
[20]
P.
H.
C.
Souza,
D.
A.
Guim
ar˜
ae
s,
and
G.
P.
Aquino,
“
Eff
i
ci
en
t
Fus
ion
of
Spect
rum
Sens
ing
Inform
at
ion
Under
Para
m
eter
Unce
rt
ai
n
t
y
and
Im
pulsive
Noise
,”
Journal
of
Comm
unication
and
Infor
mation
Syste
ms
,
vol.
33
,
no
.
1
,
20
18.
[21]
M.
Haji
ab
adi,
H.
Khos
hbin,
and
G.
A.
Hodtani,
“
Coopera
ti
ve
S
pec
trum
Esti
m
ation
Over
La
rg
e
-
Scal
e
Cogn
it
iv
e
Radi
o
Ne
tworks
,
”
IET
S
ignal P
ro
ce
ss
ing
,
vo
l. 11,
no.
8
,
pp
.
1006
–
1014,
2017
.
[22]
B.
F.
Lo
and
I.
F.
Ak
y
i
ldi
z
,
“
Re
inforc
ement
Lea
rning
for
Coope
rat
iv
e
Sens
ing
Gain
in
Cogni
tive
Radi
o
Ad
Ho
c
Networks
,”
W
ire
le
ss
Net
works,
v
ol.
19
,
no
.
6
,
pp
.
1237
–
1250,
201
3.
[23]
H.
Li
,
X
.
Cheng
,
K.
Li
,
X.
Xing
a
nd
T.
Jing,
"U
t
il
i
t
y
-
B
ase
d
Cooper
at
iv
e
Spect
rum
Sensing
Scheduling
in
Cognit
ive
Radi
o
Ne
tworks
,"
Proceedi
ngs I
EE
E
INFOCOM
,
Tur
in,
pp.
165
-
169
,
2013
.
[24]
A.
Bal
i
ei
ro
,
P.
Yos
hioka
,
K.
D
ia
s,
C.
Cordei
ro
,
and
D.
Cavalc
ant
i
,
“
Adapt
ive
Spect
rum
Sensing
for
Cognit
iv
e
Radi
o
B
ase
d
on
Multi
-
Obje
ct
iv
e Genet
i
c
Optimis
at
ion
,
”
Elec
troni
cs
Letters
,
vol
.
4
9,
no
.
17
,
pp
.
10
99
–
1101,
2013
.
[25]
P.
Chen
g,
R.
Deng,
and
J.
Che
n,
“
Ene
rg
y
-
Eff
i
ci
en
t
Coopera
t
i
ve
Spect
rum
Sensing
in
Sensor
-
Aided
Cognit
iv
e
Radi
o
Ne
tworks,
”
IE
EE Wi
rel
ess
Comm
unic
ati
ons,
vol
.
19
,
no
.
6
,
2
012.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
10
, No
.
3
,
J
une
2020
:
31
16
-
3124
3124
[26]
F.
Ye,
X
.
Zh
an
g,
Y.
Li,
and
C
.
T
ang,
“
Fai
thworth
y
Coll
abor
a
t
ive
Spe
ct
rum
Sensing
Based
on
Credi
bi
li
t
y
an
d
Evi
den
ce
The
or
y
for
Cogni
ti
ve
Radi
o
Ne
tworks
,
”
Symmet
r
y,
vol. 9, no. 3,
p
p.
36,
2017.
[27]
W
.
Han,
J.
L
i,
Z.
T
ia
n
,
and
Y
.
Zha
ng,
“
Eff
ici
ent
Cooper
at
iv
e
Spect
rum
Sensing
with
Minim
um
Overhe
ad
i
n
Cognit
ive Ra
d
io
,
”
I
EEE
Tr
ansacti
ons on
W
irel
es
s
Comm
unic
ati
o
ns,
vol. 9, no. 10
,
pp
.
3006
–
3011
,
2010.
[28]
S.
A.
Adebo,
E.
N.
Onw
uka,
A.
U.
Us
m
an
and
A.
J.
Onum
an
y
i
,
“
A
Hy
brid
Local
i
za
t
ion
Scheme
for
Det
ec
t
ion
of
Prim
ar
y
Us
er
E
m
ula
tor
in
Cogn
it
ive
Radi
o
Net
works
,
”
Inte
rnat
ional
Journal
of
Computing
and
Digit
a
l
Syste
ms
,
vol
.
8
,
no
.
3
,
pp
.
217
-
227,
2019
.
[29]
Y.
A.
W
assim
Fass
i
Fihri,
H
assan
El
Gha
zi,
Naima
Ka
abo
uch,
B
adr
Abou
El
Ma
jd
"A
Parti
c
le
Sw
ar
m
Optimiza
ti
o
n
Ba
sed
Algorit
hm
f
or
Prim
ar
y
Us
er
Emulation
Att
a
ck
Det
ection
,
"
I
EE
E
8th
Annual
Computing
and
Comm
unic
ati
on
Workshop and
C
onfe
renc
e
(
CCWC)
,
La
s Vega
s
,
US
A,
pp.
823
-
8
27,
2018
.
[30]
F.
P.
D.
Cabri
c
,
"Coopera
ti
v
e
DoA
-
Onl
y
Local
i
za
t
ion
of
Prim
ary
Us
ers
in
Cognit
ive
R
adi
o
Net
works
,
"
EURA
SIP
Journal
on
W
ire
le
ss
Comm
unic
a
ti
ons and
N
et
wo
rking,
pp
.
1
-
14
,
2013.
BIOGR
AP
H
I
ES
OF
A
UTH
ORS
Samu
el
A.
Ad
ebo
rec
ei
v
ed
B
.
E
ng.
degr
ee
in
E
l
ec
tr
ic
a
l
and
Co
m
pute
r
Engi
n
eering
and
M.
Eng
.
degr
ee
in
Com
m
unic
a
ti
ons
Engi
n
ee
ring
from
Feder
al
Univer
si
t
y
o
f
Te
chnol
og
y
M
inna
,
Niger
ia
in
2008
and
2013
r
espe
ctively
.
He
i
s
cur
ren
tly
work
ing
towar
ds
his
Ph.D
degr
ee
in
Com
m
unic
at
ions
Engi
ne
eri
ng
at
the
Depa
rtment
of
Te
le
com
m
unic
at
ions
En
gine
er
ing,
Fede
ral
Un
ive
rsit
y
of
Te
chno
log
y
,
Mi
nna,
Niger
ia
.
Hi
s
rese
arc
h
inter
ests
inc
lude
spe
ct
rum
m
ana
gement
in
cognitiv
e
rad
io, a
nd
wire
les
s sensor ne
tworks.
Eli
z
abeth
N.
Onw
uk
a
is
a
Prof
essor
of
Tele
co
m
m
unic
at
ions
E
ngine
er
ing.
She
holds
a
Ph.D
.
in
Com
m
unic
at
ions
and
Inform
ation
S
y
st
ems
Engi
nee
r
ing,
fro
m
Tsinghua
Univer
sit
y
,
Be
ij
in
g,
People
’s
Repub
li
c
of
Chin
a;
a
Master
of
En
gine
er
ing
degr
e
e,
in
Te
l
ec
om
m
unic
at
ions;
and
a
Bac
h
el
or
of
E
ngine
er
ing
degr
e
e
from
El
ec
tr
ical
and
Com
pute
r
Engi
ne
eri
ng
De
par
tment,
Fed
eral
Univer
sit
y
of
T
ec
hnolog
y
(FU
T)
Minna
,
Nige
r
Stat
e
,
Nig
eria
.
Her
r
ese
ar
ch
int
er
est
in
cl
udes
Mobile
comm
un
ic
a
ti
ons
ne
twork,
Mobil
e
IP
n
etw
orks,
Handoff
m
ana
gement
,
P
agi
ng,
Ne
twork
int
egr
at
ion
,
R
esourc
e
m
an
age
m
ent
in
wir
el
ess
net
works
,
spec
tr
um
m
ana
gement,
and
Big
Data
Anal
y
tics.
A.
U.
Us
man,
is
a
Senior
L
ecture
r
with
the
Depa
rtment
of
Te
l
ec
om
m
unic
ation
Engi
ne
eri
ng
,
Feder
al
Univer
si
t
y
of
T
ec
hno
log
y
,
Minna
,
Nig
eria.
He
obt
ai
ned
h
is
B.
Eng
.
degr
ee
in
E
lectr
i
ca
l
&
Com
pute
r
Enginee
ring
from
t
he
sam
e
Unive
rsit
y
in
1998.
He
ac
quir
ed
M
.
Sc.
in
Elec
tri
c
al
Engi
ne
eri
ng
fro
m
Univer
sit
y
of
La
gos,
Nig
eria
and
PhD
in
C
om
m
unic
at
ion
Engi
ne
eri
ng
from
Abubaka
r
Ta
f
a
wa
Bal
ewa
U
nive
rsit
y
,
Bauchi
Niger
ia
in
2002
and
2014
respe
ctively
.
He
is
cur
ren
tly
the
Deput
y
Dea
n,
School
of
El
ec
tri
ca
l
Engi
nee
ring
and
Te
chno
log
y
.
He
has
t
eachi
n
g
expe
r
ie
n
ce
i
n
the
ar
ea
o
f
m
obil
e
r
adi
o
propa
ga
ti
on
m
odel
ing,
wir
eless
comm
unic
at
ion
s
y
stem,
wire
le
ss
net
work
resour
ce
ut
il
i
za
t
ion,
n
um
eri
ca
l
m
et
ho
ds,
and
digital
el
e
ct
roni
cs.
His
rese
arc
h
int
er
e
st
inc
lude
s
rad
i
o
propa
gat
i
on
m
odel
li
ng,
indo
or
and
outdoor
wire
le
ss
comm
unic
a
ti
on
a
nd
appl
i
ca
t
ion
of
Artifi
c
ia
l
Intelli
gent
t
ec
hni
ques
in
Eng
ine
er
in
g
.
He
has
publ
ishe
d
seve
ra
l
p
ape
rs
in
na
ti
ona
l/
in
te
r
nat
ion
al
journa
ls
and confe
r
ences.
Dr.
Ad
ei
z
a
Ja
me
s
Onu
many
i
recei
ved
his
B.
Eng.
degr
e
e
in
E
le
c
trica
l
and
Elec
t
roni
cs
Engi
ne
eri
ng
fro
m
Abubaka
r
Ta
fawa
Bal
ewa
Uni
ver
sit
y
,
Bau
chi,
Niger
ia,
in
2005,
and
his
M.E
ng
an
d
PhD
degr
ee
s
in
Com
m
unic
ati
ons
Engi
ne
eri
ng
from
Feder
al
Univer
sit
y
of
Te
ch
nolog
y
(F.U.
T),
Minna,
Nig
eria
i
n
2010
and
2014
,
respe
ctively
.
H
e
has
pub
li
shed
seve
ral
rese
arc
h
art
i
cl
es
in
pe
er
-
rev
ie
wed
journ
als
and
in
IEE
E
fl
agship
conf
ere
n
ce
s.
Dr.
Onum
an
y
i
l
e
c
ture
s
at
t
he
Depa
rtment
o
f
Te
l
ec
om
m
unic
ation
Engi
n
ee
ring
,
F.U.T,
Minna
,
Niger
ia.
He
h
a
s
won
gra
nts
at
F.U.T
,
Minn
a,
serve
d
on
seve
r
al
conf
ere
n
ce
-
or
gani
z
ing
comm
it
tees.
His
rese
a
rch
intere
sts
in
cl
ude
spe
ct
rum
sensing
in cogni
t
ive
r
adi
o
,
wir
el
e
ss
sensor ne
t
works
.
Evaluation Warning : The document was created with Spire.PDF for Python.