, Nandana Rajatheva
, rajath@ait.ac.th
y
2
: H
0
1
2
(
y
2
)
u1
2
e
2+y
2
I
u1
(
2y) : H
1
where (.) is gamma function and I
(.) is the
th
order
modied Bessel function of the rst kind [1],[3]. Hence, P
d
and P
f
over AWGN channel can be given as in (1) and (2),
P
d
= Q
u
_
_
2,
_
(1)
P
f
=
_
u,
2
_
(u)
(2)
where Q
u
(., .) is the generalized Marcum Q-function and
(., .) is the incomplete gamma function [3].
III. AVERAGE DETECTION PROBABILITY (P
d
) OVER
NAKAGAMI-m FADING CHANNEL WITH EQUAL GAIN
COMBINING
If the number of branches in the equal gain combiner is
L, then the combined signal amplitude is given by [8, eqn.
6.32, pp.285],
=
L
i=1
i
. (3)
The SNR of the composite signal () is dened by
=
Es
2
LN01
[8, eqn.6.33, pp.285]. This follows, average SNR
per branch E() = =
Es
N01
where = E(
2
i
) and E(.) is
the mathematical expectation.
By means of the alternative representation of Marcum Q-
function given in [7, eqn.4.63, pp.101], we can write
Q
u
(a
, b) = 1 e
a
2
+ b
2
2
n=u
_
b
a
_
n
I
n
(ab
) (4)
where a =
2 and b =
, b) g
i
() d
= 1 e
b
2
2
n=u
_
b
a
_
n
_
0
n
2
e
a
2
2
I
n
(ab
) g
i
()d
(5)
where we have used the fact that
_
0
g
i
()d = 1.
A. Single Branch (L = 1)
When the received signal follows Nakagami-m distribution,
PDF of the received signal amplitude is given by,
f() =
2m
m
2m1
(m)
m
e
m
2
, 0 (6)
where (.) is gamma function and m is fading gure. Hence,
g
1
() can be written as in (7).
g
1
() =
1
(m)
_
m
_
m
m1
e
, 0 (7)
By (5) and (7), P
d,1
can be expressed as,
P
d,1
= 1
e
b
2
2
(m)
_
m
_
m
n=u
_
b
a
_
n
_
0
e
m1
n
2
I
n
(ab
)d
(8)
where = 1 +
m
2
_
m
_
m
n=u
_
2
_
n
1
(n + 1)
1
F
1
_
m; n + 1;
2
_
(9)
where we have use the relation given in (10),
M
k,
(x) = x
+
1
2
e
x
2
1
F
1
_
1
2
+ k; 1 + 2; x
_
. (10)
Here, M
k,
(.) is the Whittaker function and
1
F
1
(., .; .) is the
conuent hypergeometric function of the rst kind.
This result is numerically equivalent to the result [3,
eqn.(20)] for integer m. Since m is not restricted to integers in
(9), this can be treated as a more general result for Nakagami-
m fading channel compared to the result in [3, eqn.(20)].
B. Two i.i.d. branches (L = 2)
When the received signal follows Nakagami-m distribution,
PDF of combined signal amplitude is given by [6, eqn.(4)].
Hence by following the same procedure for (7), g
2
() can be
derived as given in (11),
g
2
() =
2
2m1
e
mL
2
4m1
(2m)
2
(m)(2m+
1
2
)
_
mL
_
2m
1
F
1
_
2m; 2m+
1
2
;
mL
2
_
, 0.
(11)
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE "GLOBECOM" 2008 proceedings.
978-1-4244-2324-8/08/$25.00 2008 IEEE.
By (5) and (11), P
d,2
can be expressed as,
P
d,2
= 1
2
(m)e
b
2
2
_
L
_
2m
n=u
_
b
a
_
n
_
0
2m
n
2
1
e
_
mL
+
a
2
2
_
I
n
(ab
)
1
F
1
_
2m; 2m+
1
2
;
mL
2
_
d
(12)
where
2
(m) dened by (13),
2
(m) =
2
(2m)m
2m
2
(m)(2m+
1
2
)2
4m1
. (13)
The hypergeometric series expansion of the form
1
F
1
(.; .; .) is
given by (14) [10, pp.202],
1
F
1
(a; b; x) =
k=0
(a)
k
x
k
(b)
k
k!
. (14)
Here, (.)
k
is Pochhammer symbols [10, pp.202] which is
dened by (15).
(a)
k
=
(a +k)
(k)
(15)
Using [9, eqn.6.643-2, pp.709], (12),(13),(14) and (15), P
d,2
for m > 0 can be evaluated as,
P
d,2
= 1
n=u
k=0
_
2
_
n
_
mL
_
2m+k
2
(m, n, k)
2
4m+k2
k!
1
F
1
(2m+k; n + 1;
2
)
(16)
where
2
(m, n, k) is dened by (17) and = 1 +
mL
2
(m, n, k) =
2
(2m+k)
2
(m)(n + 1)(2m+k +
1
2
)
(17)
C. Three i.i.d. branches (L = 3)
When the received signal follows Nakagami-m distribution,
PDF of combined signal amplitude is given by, [6, eqn.(8)] and
by following the same procedure used for (7), g
3
(), 0
can be expressed as,
g
3
() =
4
(2m)e
mL
3
(m)2
4m1
n=0
(2m+n)
(2m+n +
1
2
)
(4m+ 2n)
3m+n1
(6m+ 2n)(n + 1) 2
n
_
mL
_
3m+n
2
F
2
_
2m, 4m+ 2n; 3m+n +
1
2
, 3m+n;
mL
2
_
.
(18)
The Hypergeometric series expansion of generalized hyper-
geometric function [10, pp.202] is given in (19).
p
F
q
(a
1
, ..., a
p
; b
1
, ..., b
q
; x) =
n=0
(a
1
)
n
...(a
p
)
n
(b
1
)
n
...(b
p
)
n
x
n
n!
(19)
Using the form of
2
F
2
(a
1
, a
2
; b
1
, b
2
; x) in (19) and using (5)
and (18), P
d,3
can be shown as in (20),
P
d,3
= 1
3
(m)e
b
2
2
n=u
p=0
k=0
_
b
a
_
n _
m
2
_
p+k
_
L
_
3m+p+k
(2m+p)(4m+ 2p)
(6m+ 2p)(p + 1)(2m+p +
1
2
)
(2m)
k
(4m+ 2p)
k
(3m+p)
k
_
3m+p +
1
2
_
k
k!
_
0
3m+p+k
n
2
1
e
_
mL
+
a
2
2
_
I
n
(ab
)d
(20)
where
3
(m) is dened by (21),
3
(m) =
4
(2m)m
3m
3
(m)2
4m1
. (21)
Using [9, eqn.6.643-2, pp.709] and the identity given in (10),
P
d,3
for m > 0 can be computed as,
P
d,3
= 1
n=u
p=0
k=0
_
2
_
n
_
mL
_
3m+p+k
3
(m, n, p, k)
2
4m+p+k3
k!
1
F
1
_
3m+p +k; n + 1;
2
_
(22)
where
3
(m, n, p, k) is dened by (23) below and = 1+
mL
3
(m, n, p, k) =
(2m+p)(2m+k)(3m+p)
3
(m)(n + 1)(p + 1)(2m+p +
1
2
)
(3m+p +
1
2
)(4m+ 2p +k)
(3m+p +k +
1
2
)(6m+ 2p)
(23)
IV. FOUR OR MORE I.I.D. BRANCHES (L 4)
When the received signal follows Nakagami-m distribution,
PDF of combined signal amplitude of four branch EGC is
given by [6, eqn.(9)] and by following the same procedure used
for (7), g
4
() can be expressed as given in (24). Following a
similar procedure in deriving (22), P
d,4
for m > 0 can be
evaluated as given in (25) where
4
(m, n, p, q, k) is dened
by (26) and = 1 +
mL
.
The PDF of combined signal amplitude of Nakagami-m
faded L 4 branch EGC is given in [6, eqn.(10)]. By
following the same line of arguments for (7), g
L
() can be
evaluated as given in (27). Following similar procedure in
deriving (22), P
d,L
for m > 0 can be evaluated as given in
(28) where = 1 +
mL
. Replacing k
1
by p, k
2
by q, p by k
and L = 4 in (28), we can easily verify (25).
Probability of false alarm depends only on the hypothesis
H
0
. Hence the statistics of each sample taken from combined
signal follows central Chi-square distribution and further de-
cision variable also follows the central Chi-square distribution
with 2u degree freedom (
2
2u
) [1],[3]. This leads to the same
P
f,i
for i = 1, 2, ..., L as given by (2).
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE "GLOBECOM" 2008 proceedings.
978-1-4244-2324-8/08/$25.00 2008 IEEE.
g
4
() =
8
(2m)m
4m
e
mL
4
(m)2
4m1
p=0
q=0
(2m+p)(2m+q)(4m+ 2p +q)
(p + 1)(q + 1)(2m+p +
1
2
)(8m+ 2p + 2q)
4m+p+q1
2
qp
m
p+q
_
L
_
4m+p+q
2
F
2
(2m, 6m+ 2p + 2q; 4m+p +q +
1
2
, 4m+p +q;
mL
2
)
(24)
P
d,4
= 1
n=u
p=0
q=0
k=0
_
2
_
n
_
mL
_
4m+p+q+k
4
(m, n, p, q, k)
2
4m+p+kq4
k!
1
F
1
_
4m+p +q +k; n + 1;
2
_
(25)
4
(m, n, p, q, k) =
(2m+p)(2m+q)(2m+k)(4m+p +q)(4m+ 2p +q)(4m+p +q +
1
2
)(6m+ 2p + 2q +k)
4
(m)(q + 1)(p + 1)(n + 1)(2m+p +
1
2
)(4m+p +q +k +
1
2
)(6m+ 2p + 2q)(8m+ 2p + 2q)
(26)
g
L
() =
2
L1
(2m) m
Lm
e
mL
2
4m1
L
(m)
k1=0
k2=0
...
kL2
L2
i=1
_
(2m+k
i
)
(1 +k
i
)
_
(4m+ 2k
1
+k
2
) (6m+ 2k
1
+ 2k
2
+k
3
) ...
_
2m(L 2) + 2
L3
i=1
k
i
+k
L2
_
_
2m(L 1) + 2
L2
i=1
k
i
_
(2m+k
1
+
1
2
) (6m+ 2k
1
+ 2k
2
) (8m+ 2k
1
+ 2k
2
+ 2k
3
) ...
_
2Lm+ 2
L2
i=1
k
i
_
_
2m(L 2) + 2
L3
i=1
k
i
+k
L2
_
m
(
L2
i=1
ki)
2
(
L2
i=2
kik1)
(Lm+
L2
i=1
ki1)
_
L
_
(Lm+
L2
i=1
ki)
2
F
2
_
2m, 2m(L 1) + 2
L2
i=1
k
i
; Lm+
L2
i=1
k
i
+
1
2
, Lm+
L2
i=1
k
i
;
mL
2
_
, L 4
(27)
P
d,L
= 1
n=u
k1=0
k2=0
...
kL2
p=0
L2
i=1
_
(2m+k
i
)
(1 +k
i
)
_
(4m+ 2k
1
+k
2
)(6m+ 2k
1
+ 2k
2
+k
3
) ...
_
2m(L 3) + 2
L4
i=1
k
i
+k
L3
_
2
_
2m(L 2) + 2
L3
i=1
k
i
+k
L2
_
(2m+k
1
+
1
2
)(6m+ 2k
1
+ 2k
2
)(8m+ 2k
1
+ 2k
2
+ 2k
3
) ...
_
2Lm+ 2
L2
i=1
k
i
_
(2m+p)
_
Lm+
L2
i=1
k
i
_
_
Lm+
L2
i=1
k
i
+
1
2
_
_
2m(L 1) + 2
L2
i=1
k
i
+p
_
L
(m) (n + 1)
_
mL +
L2
i=1
k
i
+p +
1
2
_
2
_
n
_
mL
_
(Lm+
L2
i=1
ki+p)
2
(4m+k1
L2
i=2
kiL+p)
p!
1
F
1
_
p +Lm+
L2
i=1
k
i
; n + 1;
2
_
, L 4
(28)
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE "GLOBECOM" 2008 proceedings.
978-1-4244-2324-8/08/$25.00 2008 IEEE.
0 5 10 15 20
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Average SNR(dB)
P
d
Fading m=1.0
EGC L=2 m=0.5
EGC L=2 m=1.0
EGC L=2 m=1.5
EGC L=2 m=2.0
EGC L=2 m=2.5
EGC L=2 m=3.0
Fig. 1. P
d,i
for i=1,2 when P
f
= 0.001 and u = 1.
V. NUMERICAL RESULTS
We describe the performance of the receiver by means of
P
d,i
vs. SNR curves for pre-specied P
f,i
value. First, energy
threshold of the receiver () is calculated for the specied P
f,i
value. Then for that , P
d,i
is evaluated.
Fig.1 shows the P
d,i
performance improvement of dual
branch EGC receiver over different values of m when u = 1
and P
f,i
= 0.001. As it is expected, higher values of m
and SNR result in better performance. However, the highest
P
d,i
improvement is observed over lower range of m values.
Fig.2 plots the similar analysis including the three branch EGC
receiver.
Fig.3 shows the P
d,i
performance variation for u = 1, 4
when m = 2, L = 1, 2, 3 and P
f,i
= 0.01. This plot shows the
performance reduction for higher value of u in each case, i.e.
no-diversity fading channel and L = 2, 3. Comparing suitable
curves of Fig.1, Fig.2 and Fig.3, it is possible to observe the
performance reduction of P
d,i
for higher P
f,i
requirement.
However, the SNR requirement for higher detection probability
is moderate in all cases.
VI. CONCLUSION
We consider the problem of primary user detection in
cognitive radio over the Nakagami-m fading channel with
the equal gain combining diversity receiver. Expressions are
derived for exact probability of detection when the number
of diversity branches are 1, 2, 3 and L 4. Interestingly,
all the expressions could be expressed in terms of conuent
hypergeometric function of the rst kind
1
F
1
(.; .; .). These
results could be readily used in deciding the number of
diversity branches and the energy threshold value to achieve
a specied false alarm rate of equal gain combiner energy
detector receiver in cognitive radio.
REFERENCES
[1] H. Urkowitz, Energy detection of Unknown Deterministic Signals,
Proc IEEE, vol. 55, no. 4, pp.523-531,Apr. 1967.
0 5 10 15 20
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Average SNR(dB)
P
d
Fading u=1 m=1
Fading u=1 m=3
EGC L=2 u=1 m=1
EGC L=2 u=1 m=3
EGC L=3 u=1 m=1
EGC L=3 u=1 m=3
Fig. 2. P
d,i
for i=1,2,3 when P
f
= 0.01 and u = 1.
0 5 10 15 20
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Average SNR(dB)
P
d
Fading u=1 m=2
Fading u=4 m=2
EGC L=2 u=1 m=2
EGC L=2 u=4 m=2
EGC L=3 u=1 m=2
EGC L=3 u=4 m=2
Fig. 3. P
d,i
for i = 1, 2, 3 when P
f
= 0.01.
[2] V.I. Kostylev, Energy detection of a signal with random amplitude,
IEEE Int. Conf. ICC 2002, vol 3, pp.1606 - 1610, Apr-May 2002.
[3] Fadel F. Digham, Mohamed-Slim Alouni and Marvin K. Simon, On the
Energy Detection of Unknown Signals Over Fading Channels, IEEE Int.
Conf. ICC03, vol 5, pp. 3575 - 3579, May 2003.
[4] Fadel F. Digham, Mohamed-Slim Alouni and Marvin K. Simon, On the
Energy Detection of Unknown Signals Over Fading Channels, IEEE
Trans. Commun., vol 55, no.1,pp.21-24, Jan. 2007.
[5] Ashish Pandharipande and Jean-Paul M.G.Linnartz, Performance analy-
sis of primary user detection in a multiple antenna cognitive radio,
IEEE Int. Conf. ICC 07, pp. 6482 - 6486, June 2007.
[6] Prathapasinghe Dharmawansa, Nandana Rajatheva and Kazi
Ahmed, On the distribution of the sum of Nakagami-m random
variable, IEEE trans. on commun., vol. 55, no. 7, pp.1407-1416, july
2007.
[7] Marvin K.Simon and Mohamed-Slim Alouni, Digital Communication
Over Fading Channels, 2nd Edition.
[8] Gordon L. Stuber Principles of Mobile Communication, 2nd Edition,
Kluwer Academic Publishers.
[9] I.S. Gradshteyn and I.M. Ryzhik, Table of Integrals, Series, and
Products, 7th edition, 2007.
[10] Harry Bateman, Higher Transcendental Functions, Volume I, 1953.
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE "GLOBECOM" 2008 proceedings.
978-1-4244-2324-8/08/$25.00 2008 IEEE.