Performance Analysis of Decode-and-Forward

Cooperative Diversity Using Differential EGC over

Nakagami-m Fading Channels
Salama S. Ikki, Student member IEEE and Mohamed H. Ahmed, Senior Member IEEE
Faculty of Engineering and Applied Sciences
Memorial University of Newfoundland
St. Johns, NL, Canada A1B 3X5
e-mail: s.s.ikki@mun.ca, mhahmed@mun.ca
AbstractCooperative diversity is a promising technology for
future wireless networks. In this paper, we derive the average bit
error rate (BER) and outage probability (P
out) for differential
equal gain combining (EGC) in cooperative diversity networks.
The considered network uses adaptive decode-and-forward (DF)
relaying over independent non-identical Nakagami-m fading
channels. In adaptive DF relaying among M relays, that can
participate, only C relays (C M), with good channels to
the source, decode and then forward (retransmit) the source
information to the destination. Then, the destination combines
the direct and the indirect signals using differential EGC. We
rst derive a simple exact expression for the equivalent SNR at
the destination. Second, we derive the expressions of the PDF and
the MGF of this equivalent total SNR at the destination. Then the
MGF is used to determine the error and outage probabilities of
adaptive DF with an arbitrary number of relays. Furthermore,
we found (in terms of MGF) the SNR moments, the average
signal-to-noise ratio (SNR) and the amount of fading. Computer
simulations are used to validate our analytical results. Results
show the signicant performance improvement due to the use
of the adaptive DF cooperative diversity. Also, results show
that the performance of the adaptive DF differential EGC is
comparable to the adaptive DF maximum ratio combining (MRC)
performance.
Index TermsCooperative diversity; Rayleigh fading; error
performance; outage probability; differential equal gain combin-
ing.
I. INTRODUCTION
Cooperative-diversity networks technology is a promising
solution for the high data-rate coverage required in future wire-
less communications systems. There are two main advantages
of this technology; the low transmit RF power requirements,
and the spatial diversity gain. Different cooperation diversity
protocols have been proposed for wireless networks [1]. The
basic idea is that in addition to the direct transmission from
the transmitter to the receiver, there are other nodes, which
can be used to enhance the diversity by relaying the source
signal to the destination.
Performance analysis of cooperative-diversity networks
yielded many interesting results including information theo-
retic metrics, outage signal-to-noise ratio (SNR) probability,
and average error probability expressions over Rayleigh-fading
channels [2][4]. In [1], authors have proposed a variety of
low-complexity cooperative protocols using the three-node
case (source, destination and relay). These protocols have
been utilized using amplify-and-forward (non-regenerative)
relaying and decode-and-forward (regenerative) relaying. Also
the outage probability, using high-SNR approximations, has
been analyzed under Rayleigh fading channels in [1]. The
authors in [5], [6] have proposed the user cooperation concept
and considered practical issues related to its implementation.
Furthermore, the authors in [2] have presented an overview
of cooperative diversity networks and compared their perfor-
mance with that of direct transmission and normal relaying
networks. In [7][9], the authors have determined the outage
probability of adaptive DF cooperative networks with identical
and non-identical Rayleigh fading channels. In particular, the
authors in [7] have determined an upper bound for the outage
probability while the authors in [8] have determined a closed
form expression for the outage probability. Furthermore, in [9],
the authors have found the exact form expression of the error
probability under independent non-identical Rayleigh fading
channels.
Most of the previous publications assume that the destina-
tion has perfect knowledge of the channel state information
(CSI)
1
of all links. While in some scenarios, e.g. slow fading
environment, the CSI can be acquired using the pilot symbols
or training sequences, it may not be possible in some sys-
tems, particularly with fast fading channels. Furthermore, it is
questionable how the destination can obtain the source-relay
channel information perfectly without noise. Moreover, the
computational overhead for the channel estimation increases
with the number of relaying nodes.
Differential modulation has been well accepted as a modu-
lation technique that provides a good tradeoff between receiver
complexity and performance since CSI is not necessarily
required for demodulation. Differential Phase Shift Keying
(DPSK) modulation has been investigated in [10] for a two-
hop amplify-and-forward relaying system over Rayleigh fad-
ing channels. Recently, a framework of non-coherent co-
1
Full channel state information includes the channel amplitude and phase
information.
978-1-4244-2517-4/09/$20.00 2009 IEEE
operative communications has been proposed in [11] for
decode-and-forward scheme employing frequency shift keying
(FSK) modulation. In [12][15], a differential scheme has
been proposed for two-user and multi user cooperating nodes
that employ the amplify-and-forward relaying protocol. In all
these papers, the bit error-rate (BER) performance is provided
for Rayleigh fading channels. For a decode-and-forward and
single relay only the authors in [16], have determined a
tight error probability approximation. Also in their paper,
error probability upper bound and lower bound are found.
Some analysis for optimal power allocation also is considered
in [16]. Finally, a specic two-user differential cooperation
scheme was proposed in [17]. Nevertheless, the scheme de-
pends on the synchronization among users in the networks,
and the scheme provides a limited transmission rate.
Most of the previous publications have been devoted to
study the cooperative-diversity networks over Rayleigh-fading
channels. However, it is widely known that the Nakagami-m
distribution is a generalized distribution, which can be used to
model different fading environments. It has the advantage of
including the Rayleigh and the one-sided Gaussian distribution
as special cases. Furthermore, it can be used to model fading
conditions that are less or more severe than those modeled by
the Rayleigh distribution.
In this paper, we analyze the performance of adaptive
DF differential EGC cooperative-diversity networks over in-
dependent non identical Nakagami-m fading channels. We
determine closed-form expressions for the probability density
function (PDF), and moment generating function (MGF) of the
instantaneous SNR of the indirect link (S R
i
D link).
Since it is difcult to nd the PDF of the combined signal at
the destination, we nd, instead, the MGF of the combined
signal SNR. By using the MGF of the combined signal SNR,
we obtain closed-form expressions for the average bit error
rate, SNR outage probability and the moments of the SNR.
The remainder of this paper is organized as follows. Sec-
tion II introduces the system and model under investigation.
Sections III and IV evaluate the end-to-end error rate and
the outage probability, respectively. Results are presented and
discussed in Section V. Finally, concluding remarks are given
in Section VI
II. SYSTEM MODEL
As shown in Fig. 1, a source node (S) and a destination
node (D) communicate over a channel with a at Nakagami-
m fading coefcient (h
S,D
). A number of cooperating nodes
R
i
, (i = 1, 2, , M) relay the signal to provide the destina-
tion with multiple copies of the original signal. The channel
coefcients between S and R
i
(h
S,Ri
) and between R
i
and
D (h
Ri,D
) are also at Nakagami-m fading coefcients. In
addition, h
S,D
, h
S,Ri
and h
Ri,D
are mutually-independent
and non-identical. We also assume here, without any loss of
generality that the additive white Gaussian noise (AWGN)
terms of all links have zero mean and equal variance (N
0
).
Suppose that DMPSK modulation is used, i.e., the in-
formation is conveyed in the phase difference between two
Fig. 1. Illustration of a multi-branch decode-and-forward cooperative-
diversity network.
consecutive symbols; then the modulated information at the
source can be described as
m
= exp(j
m
) where {
m
}
L1
m=0
is a set of L information bases. In the case of MDPSK,
m
can be specied as
m
= 2m/L where m = 0, 1, . . . , L1.
The source differentially encodes the information symbol as
[13], [14]
s(n) =
m
s(n 1) (1)
where n is the time index, and s(n) is the differentially
encoded symbol with unit energy to be transmitted at time
n. Then, the source transmits s(n) with energy E
s
to the
destination and the relays. The corresponding received signals
at the destination and the i
th
relay (i = 1, 2, . . . , M) can be
expressed as
y
S,D
(n) =
_
E
s
h
S,D
s(n) + z
S,D
(n)
y
S,Ri
(n) =
_
E
s
h
S,Ri
s(n) + z
S,Ri
(n) (2)
where h
S,D
and h
S,Ri
represent the channel fading coefcients
from the source to the destination and from the source to
the i
th
relay, respectively. The terms z
S,D
(n) and z
S,Ri
(n)
are the additive white Gaussian noise at the destination and
the i
th
relay, respectively. The relays differentially decode
the transmitted symbol from the source. Two consecutive
received signals, y
S,Ri
(n) and y
S,Ri
(n 1), are required to
recover the transmitted information at each symbol period. By
assuming that the channel coefcient h
S,Ri
is almost constant
over two symbol periods, the relay differentially decodes based
on the decision rule
m = arg max
m=0,1,...,L
Re
_
(
m
y
S,Ri
(n 1))

y
S,Ri
(n)
_
(3)
We dene the decoding set (C) as the set of relays with the
ability to fully decode the source message correctly. That is,
the relay node is said to belong to the decoding set provided
that the channel between the source and the relay node is
sufciently good to allow for successful decoding. Then, the
destination combines the direct and the indirect links using
differential EGC.
In standard differential detection, successful differential
decoding requires that the encoder differentially encodes each
information symbol with the previously transmitted symbol.
In this way, if the information symbols are sent every time
slot, then the information symbol to be transmitted at time
(n) is differentially encoded with the transmitted symbol at
time (n 1). However, in the differential adaptive decode-
and-forward scheme, the information symbols at the ith relay
are transmitted only if they are correctly decoded. Therefore,
the transmission time of the previously transmitted symbol can
be any time before the current time. We denote such previous
transmission time as (n
i
) for
i
, i.e., n
i
is the latest
time that the ith relay correctly decodes the symbol before
time n. In order to perform successful differential decoding,
the ith relay needs to store the transmitted symbol at time
n
i
. Note that needing a memory at the relay will not
increase the complexity of system compared to the standard
differential system. The only difference is that the memory
in this scheme stores the transmitted symbol at time n
i
instead of time n1 as does the standard differential scheme.
The differentially re-encoded signal at the relay can be written
as
r
i
(n) =
m
s (n
i
) (4)
where r
i
(n) is the differentially encoded symbol at the relay
at time n. It should be noted from (1) and (4) that the
differentially encoded symbols at the relays and at the source
convey the same information symbol
m
.
Finally, at the destination, the received signal from the
source and that from the relays (that are in the decoding set C)
are combined together, and then the combined output is jointly
differentially decoded. Note that, for successfully decoding,
the differential detector at the destination requires to store
the previously received signal for each relay on the decoding
set C. With an assumption that the channel coefcients stay
almost constant for several symbol periods, the signal in the
memory can be used for efcient differential decoding at the
destination. Based on the multi-channel differential detection
in [18], the combined signal before being differentially de-
coded is
y
DF
= a
s
y

S,D
(n 1)y
S,D
(n)
+

iC
a
i
y

Ri,D
(n
i
)y
Ri,D
(n) (5)
where a
s
and a
i
are the combining weights given by a
s
=
a
i
= 1/N
0
and n
i
represents the time index of the latest
signal in memory, i.e., is the most recent received signal from
the ith relay. Note that a
s
and a
i
, maximize SNR of the
combiner output when the destination is able to differentially
decode the signals from both source and relays. Based on the
combined signal in (5), the decoder at the destination jointly
differentially decodes the transmitted information symbol by
using the following decision rule:
m = arg max
m=0,1,...,L
Re {

m
y
DF
} (6)
Finally, it should be noted here that the equivalent output
SNR from the differential EGC combiner can be written as
[18]

DF
=
S,D
+

iC

Ri,D
(7)
where
S,D
= E
s
h
2
S,D
/N
0
, is the instantaneous SNR be-
tween S and D,
S,Ri
= E
s
h
2
S,Ri
/N
0
, is the instantaneous
SNR between S and R
i
and
Ri,D
= E
s
h
2
Ri,D
/N
0
and

S,Ri
= E
s

h
2
S,Ri
/N
0
are is the instantaneous and average
SNR between R
i
and D.
III. ERROR PROBABILITY ANALYSIS
Since the decoding set C is not constant and varied from
time to time, it would be difcult to use equations (5) and (7)
to nd a closed form expression for the error probability of the
adaptive decode-and-forward cooperative networks especially
over non-identical fading channels. To bypass this difculty,
the cooperative-diversity network in Fig. 1 can be visualized as
a system that has effectively M +1 paths between the source
and destination. Let the path i represent the S R
i
D
indirect (cascaded) link, where i = 1, , M. the equivalent
output SNR at the destination can be represented as a random
variable
i
that will take account of both the source to the ith
relay link and the ith relay to destination link. Therefore,
i
has PDF as
f
i
(x) =f
i|Ri Decodes Incorrectly
(x) Pr (R
i
Decodes Incorrectly)
+f
i|Ri Decodes Correctly
(x) Pr (R
i
Decodes Correctly)
(8)
The probability that the link will be non-active (off) is equiv-
alent to the error probability at R
i
which can be written as
[18]
A
i
=
1

_
(L1)
L
0
M
S,R
i
_
g
PSK
1 +

1 q
PSK
cos ()
_
d (9)
where M
S,R
i
(s) is he Moment Generating Function
2
(MGF)
of
S,Ri
and g
PSK
= sin
2
(/L). It is known that the MGF of

S,Ri
can be written as
_
1 + s
S,R
i
mS,R
i
_
mS,R
i
where m
S,Ri
is
the fading parameter of
S,Ri
. For the special case of Binary
DPSK wherein g
PSK
= 1 and L = 2 we can write
A
i
=
1
2
_
m
S,Ri
m
S,Ri
+
S,Ri
_
mS,R
i
(10)
When R
i
decodes incorrectly, the received SNR at the des-
tination by R
i
will be 0 (since there is no retransmission
from R
i
because in this case the relay will be off and
does not belong to the decoding set C). Therefore, the
conditional PDF f
i|Ri Decodes Incorrectly
(x) can be written as
2
The MGF of a random variable X is dened as M
X
(s) =
E(exp(sX)) where E is the statistical average operator.
f
i|Ri Decodes Incorrectly
(x) = (x). The probability that the ith
link decodes the source message correctly is 1A
i
. Since the
destination in this case will receive another copy of the source
signal by R
i
with a SNR
Ri,D
(because in this case the relay
will be on and belongs to the decoding set C), the conditional
PDF f
i|Ri Decodes Correctly
(x) can be written as
f
i|Ri Decodes Correctly
(x) =
_
m
Ri,D

Ri,D
_
mR
i
,D

mR
i
,D1
(m
Ri,D
)
exp
_

m
Ri,D

Ri,D
_
(11)
Therefore, the PDF of
i
can be expressed as
f
i
(x) =A
i
(x) + (1 A
i
)
_
m
Ri,D

Ri,D
_
mR
i
,D

mR
i
,D1
(m
Ri,D
)
exp
_

m
Ri,D

Ri,D
_
(12)
Now, the equivalent output SNR from the differential EGC
combiner can be rewritten as

DF
=
S,D
+
M

i=1

i
(13)
Note that the expression of
DF
in (13) is equal to the
expression given in (7). However, (13) is more analytically
tractable than (7) and as a result, this facilitates the derivation
of the SNR statistics (MGF).
If we can nd the MGF of
DF
, we can use this MGF to
calculate the error probability. Since all the links are assumed
to be independent; hence, the MGF of
DF
is given by
M
DF
(s) = M
S,D
(s)
M

i=1
M
i
(s) (14)
where M
S,D
(s) =
_
1 + s
S,D
mS,D
_
mS,D
is the MGF between
S and D path and M
i
(s) , i = 1, , M is the MGF of the
ith indirect link variable
i
and it can be easily derived from
(12) as
M
i
(s) = A
i
+ (1 A
i
) M
R
i
,D
(s) (15)
where M
R
i
,D
(s) = (1 + s
Ri,D
/m
Ri,D
)
mR
i
,D
is the MGF
between R
i
and D path. Finally, the error probability can be
written as:
P(e) =
1
2
2M+2
_

()
1 + 2 sin() +
2
M
S,D
()
M

i=1
M
i
()d (16)
where =
_
a
2
+ b
2
+ 2ab sin
_
/2, and 0
+
(a/b) =
1 where a and b are constants that depend on the
modulation size, specically, a = 10
3
and b =

2 for
DBPSK modulation, while a =
_
2

2 and b =
_
2 +

2
for DQPSK modulation [18]. The values for larger modulation
sizes can be obtained by using the result in [18]. Finally,
() =
M+1

i=1
_
2M + 1
M + 1 i
_

_
_

i+1

i+1
_
cos
_
(i 1)( +

2
)
_

i+2

i
_
cos
_
i( +

2
)
__
(17)
where
_
2M+1
M+1i
_
=
(2M+1)!
(M+1i)!(M+i)!
. While the expression
found in (16) is exact, it still requires numerical evaluation
of the integral. By using the same method in [18, eq. 9.113]
and after some algebraic manipulation, a closed form upper
bound on the error performance can be written as
P(e)
1
2
2M+1
M+1

i=1
(2M + 1)!
(M + 1 i)!(M + i)!

_
_
3
2
+
1
i1

i1
(1 )
_
M
S,D
()
M

k=1
M
i
()

_
1
2
+
1
i1

i1
(1 )
_
M
S,D
()
M

k=1
M
i
()
_
(18)
where = b
2
(1 )
2
/2 and = b
2
(1 + )
2
/2.
Finally, it should be mentioned here that although (18)
has multiple summations/multiplications, all these summa-
tions/multiplications are nite and easy to handle numerically.
IV. OUTAGE PROBABILITY AND SNR MOMENTS
In addition to the error probability, the MGF can be used
to nd the moments, average SNR and amount of fading. The
moments of
DF
can be found as

n
= (1)
n
_

n
s
n
_
M
S,D
(s)
M

i=1
M
i
(s)
__
s=0
(19)
By setting n = 1 in (19), the average equivalent SNR can
be obtained. The rst two moments of
DF
can be used to
evaluate the amount of fading (or sometimes called fading
gure) (AF) at the destination [18, Chapter 1]. The AF is
used as a unied measure of the severity of the fading and is
dened as the ratio of the variance to the square mean of
DF
and can be expressed in a closed form as AF =
2
/
2
1
1.
The mutual information between the source and destination,
when DF scheme is used, can be written as
I
DF
=
1
M + 1
log
2
_
1 +
S,D
+

iC

Ri,D
_
=
1
M + 1
log
2
_
1 +
S,D
+
M

i=1

i
_
(20)
where C is the decoding set of the relays that will decode and
forward. The reason of the
1
M+1
factor is that we need M +1
time slots (or orthogonal channels) for transmitting the data.
From the outage probability denition and utilizing the total
probability law we can write the outage probability as
P
out
= Pr (I
DF
R) = Pr
_

DF
2
(M+1)R
1
_
(21)
where R is some xed spectral efciency determined by the
designer depending on the application. Hence, the outage
probability can be written as
P
out
=
_

1
_
M
DF
(s)
s
__
s=2
(M+1)R
1
(22)
where () denotes the inverse Laplace transform. The in-
verse Laplace transform can be easily performed by resort-
ing to the efcient Euler summation-based technique [18].
Furthermore all mathematical software Like MAPLE@ and
MATHEMATICA@ can perform the inverse Lapalace trans-
form numerically.
V. NUMERICAL RESULTS AND SIMULATIONS
In this section, we show the numerical results of the BER
and outage probability for DBPSK modulation. We plot the
performance curves of the average BER and outage probability
versus the SNR of the transmitted signal (E
s
/N
0
dB). Our
analytical results are compared with Mont-Carlo simulation
results for verication. We compare the cooperative-diversity
network with a conventional non-cooperative-system that in-
volves direct transmission only from the source to the desti-
nation. For fair comparison, we set the sum of the transmitted
energy from both the source and the relay for the cooperative-
system equal to that of the non-cooperative-system.
Fig. 2 shows the error probabilities when 1, 2 and 3 relay
nodes are used and E
_
h
2
S,D
_
= E
_
h
2
S,Ri
_
= E
_
h
2
Ri,D
_
= 1.
Observe that the results obtained using the closed-form ex-
pression derived in this paper and the simulation results are
in excellent agreement. Also increasing the number of relays
always improves the error probability for all the region of SNR
and obliviously improves the diversity order. We can see that
the differential cooperative scheme achieves higher diversity
orders as M increases. For example, at an error probability of
10
4
, we observe about 2 3 dB gain as M increases from
2 to 3.
Fig. 3 shows the outage probabilities when 1, 2 and 3 relay
nodes are used. In Fig. 3, we have chosen R = 1 bit/sec/Hz.
Again observe the exact match between the simulation and the
analytical results, pointing out the validity of our proposed
analysis. Interestingly enough, the presence of many nodes
does not necessarily improve the outage probability as also
observed in [8], [9]. Hence, we may conclude that depending
on the SNR vlue, an optimal number of relaying nodes should
be utilized, when the DF relaying protocol is employed
VI. CONCLUSION
We have analyzed the performance of the multi-branch
adaptive decode-and-forward scheme for cooperative-diversity
networks operating over independent non-identical distributed
Nakagami-m fading channels. Important performance metrics
0 5 10 15 20 25
10
12
10
10
10
8
10
6
10
4
10
2
10
0
E
s
/N
0
dB
E
r
r
o
r
P
r
o
b
a
b
i
l
i
t
y
E(hS,D)
2
= 1, E(hS,Ri
)
2
= 1, and E(hRi,D)
2
= 1


Analytical
Simulation
M = 0
M = 1
M = 2
M = 3
Fig. 2. Error probability of adaptive DF differential EGC with M =
0, , 3.
0 5 10 15 20 25
10
7
10
6
10
5
10
4
10
3
10
2
10
1
10
0
E
s
/N
0
dB
O
u
t
a
g
e
P
r
o
b
a
b
i
l
i
t
y
E(hS,D)
2
= 1, E(hS,Ri
)
2
= 1, and E(hRi,D)
2
= 1


Analytical
Simulation
M = 0
M = 3
M = 2
M = 1
Fig. 3. Outage probability of the differential adaptive decode-and-forward
EGC with M = 0, , 3.
have been addressed. Capitalizing on an extracted MGF for-
mula of the output SNR at the destination, novel closed-
form expressions for outage probability, error probability,
amount of fading and the moments of the equivalent SNR
at the destination were obtained. Computer simulation results
veried the accuracy and the correctness of the proposed
analysis. We show that the diversity order of the adaptive DF
differential EGC scheme is M +1 when M is the number of
relays. We observe about 23 dB gain at an error probability
of 10
4
when M increases from 2 to 3.
REFERENCES
[1] J. N. Laneman, D. N. C. Tse, and G.W.Wornell, Cooperative diversity
in wireless networks: efcient protocols and outage behavior, IEEE
Trans. on Inf. Theory, vol. 50, pp. 3062-3080, Dec. 2004.
[2] E. Zimmermann, P. Herhold, and G. Fettweis, On the performance of
cooperative relaying protocols in wireless networks, European Trans-
actions on Telecommunications, vol. 16, pp. 5-16, Jan. 2005.
[3] J. Boyer, D. Falconer, H. Yanikomeroqlu, Multihop diversity in wireless
relaying channels, IEEE Trans. on Commun., vol. 52, pp.1820- 1830,
Oct. 2004.
[4] J. Boyer, D. Falconer, and H. Yanikomeroglu, A theoretical charac-
terization of the multihop wireless communications channel without
diversity, in Proc. Int. Symp. Personal, Indoor, and Mobile Radio
Communications, San Diego, CA, Oct. 2001.
[5] A. Sendonaris, E. Erkrip, and B. Aazhang, User cooperation diversity-
Part I: System description, IEEE Trans. on Commun., vol. 51, pp. 1927-
1938, Nov. 2003.
[6] A. Sendonaris, E. Erkrip, and B. Aazhang, User cooperation diversity-
Part II: Implementation aspects and performance analysis, IEEE Trans.
on Commun., vol. 51, pp. 1939- 1948, Nov. 2003.
[7] Y. Zhao, R. Adve, and T. J. Lim, Outage probability at arbitrary SNR
with cooperative diversity, IEEE Commun. Letter, vol. 9, pp. 700-702,
Aug. 2005.
[8] N. C. Beaulieu and J. Hu, A closed-form expression for the outage
probability of decode-and-forward relaying in dissimilar Rayleigh fading
channels, IEEE Commun. Letter, vol. 10, pp. 813-815, Dec. 2006.
[9] Jeremiah Hu and Norman C. Beaulieu, Closed-Form Expressions for
the Outage and Error Probabilities of Decode-and-Forward Relaying
in Dissimilar Rayleigh Fading Channels, Proc. IEEE International
Conference on Communication (ICC07), Glasgow, Scotland, June 2007.
[10] M. O. Hasna and M.-S. Alouini, Performance analysis of two-hop
relayed transmissions over Rayleigh fading channels, in Proc. IEEE
Vehicular Tech. Conf. (VTC), vol. 4, pp. 1992-1996, Nov. 2003.
[11] D. Chen and J. N. Laneman, Noncoherent demodulation for cooperative
diversity in wireless systems, in Proc. IEEE Global Comm. Conf.
(GLOBECOM), Dallas, TX, pp. 31-35, Nov. 2004.
[12] Q. Zhao and H. Li, Performance of Differential Modulation with
Wireless Relays in Rayleigh Fading Channels, IEEE Commun. Letters,
vol. 9, pp. 343-345, April. 2005.
[13] T. Himsoon, W. Su, and K. J. R. Liu, Differential transmission for
amplify-and-forward cooperative communications, IEEE Signal Pro-
cessing Letters, vol. 12, pp. 597-600, Sept. 2005.
[14] T. Himsoon, W. Su, and K. J. R. Liu, Differential Modulation for Multi-
Node Amplify-and-Forward Wireless Relay Networks, in Proc. IEEE
WCNC, Mar. 2005.
[15] S. Ikki and M. H. Ahmed , Performance of Cooperative Diversity Using
Equal Gain Combining (EGC) over Nakagami-m Fading Channels,
IEEE Trans. on Wrls. Commun., To appear, 2008.
[16] T. Himsoon, Siriwongpairat W.P., W. Su and Liu, K.J.R. , Differential
Modulation With Threshold-Based Decision Combining for Cooperative
Communications, IEEE Trans. On Signal Processing., vol. 55, pp.
3905-3923, July. 2007.
[17] P. Tarasak, H. Minn, and V. K. Bhargava, Differential modulation for
two-user cooperative diversity systems, in Proc. IEEE Global Comm.
Conf. (GLOBECOM), Dallas, TX, Nov. 2004.
[18] M. K. Simon and M.-S. Alouini, Digital Communication over Fading
Channels. New York: Wiley, 2000.
[19] I. S. Gradshteyn and I. M. Ryzhik, table of integrals, series and products
San Diego, CA: Academic Press, 5th edition, 1994.