IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 13, NO.
3, MARCH 2014
Achieving Delay Diversity in
Asynchronous Underwater Acoustic (UWA)
Cooperative Communication Systems
Mojtaba Rahmati and Tolga M. Duman, Fellow, IEEE
Abstract—In cooperative UWA systems, due to the low speed
of sound, a node can experience significant time delays among
the signals received from geographically separated nodes. One
way to combat the asynchronism issues is to employ orthogonal
frequency division multiplexing (OFDM)-based transmissions at
the source node by preceding every OFDM block with an
extremely long cyclic prefix (CP) which reduces the transmission
rates dramatically. One may increase the OFDM block length
accordingly to compensate for the rate loss which also degrades
the performance due to the significantly time-varying nature of
UWA channels. In this paper, we develop a new OFDM-based
scheme to combat the asynchronism problem in cooperative
UWA systems without adding a long CP (in the order of the
long relative delays) at the transmitter. By adding a much
more manageable (short) CP at the source, we obtain a delay
diversity structure at the destination for effective processing and
exploitation of spatial diversity by utilizing a low complexity
Viterbi decoder at the destination, e.g., for a binary phase shift
keying (BPSK) modulated system, we need a two-state Viterbi
decoder. We provide pairwise error probability (PEP) analysis
of the system for both time-invariant and block fading channels
showing that the system achieves full spatial diversity. We find
through extensive simulations that the proposed scheme offers a
significantly improved error rate performance for time-varying
channels (typical in UWA communications) compared to the
existing approaches.
Index Terms—Asynchronous communication, cooperative sys-
tems, underwater acoustics, OFDM.
I. I NTRODUCTION
C OOPERATIVE UWA communications which refers to a
group of nodes, known as relays, helping the source to
deliver its data to a destination is a promising physical layer
solution to improve the performance of UWA systems [1], [2],
[3]. In a UWA cooperative communication system, the time
differences among signals received from geographically sepa-
rated nodes can be excessive due to the low speed of sound in
water. For example, if the relative distance between two nodes
with respect to another one is 500 m, then their transmissions
Manuscript received March 26, 2013; revised September 12, 2013; accepted
December 13, 2013. The associate editor coordinating the review of this paper
and approving it for publication was J. Wu.
This publication was made possible by NPRP grant 09-242-2-099 from
the Qatar National Research Fund (a member of Qatar Foundation). The
statements made herein are solely the responsibility of the authors.
M. Rahmati is with the School of Electrical, Computer and Energy
Engineering (ECEE) of Arizona State University, Tempe, AZ 85287-5706,
USA (e-mail: mojtaba@asu.edu).
T. M. Duman is with the Department of Electrical and Electronics
Engineering, Bilkent University, Bilkent, Ankara, 06800, Turkey (e-mail:
duman@ee.bilkent.edu.tr). He is on leave from the School of ECEE of
Arizona State University.
Digital Object Identifier 10.1109/TWC.2014.020414.130538
1536-1276/14$31.00 
c 2014 IEEE
1367
experience a relative delay of 333 ms. Considering, for in-
stance, that in an OFDM-UWA cooperative communication
scheme with 512 sub-carriers over a total bandwidth of 8 kHz,
the OFDM block duration is only 64 ms, the excessive delay
of 333 ms becomes problematic. Furthermore, UWA channels
are highly time varying due to the large Doppler spreads
and Doppler shift effects (or Doppler scaling) [4]. Therefore,
a practical non-centralized UWA cooperative communication
system is asynchronous with large relative delays among
the nodes and sees highly time-varying frequency selective
channel conditions.
Our focus in this paper is on asynchronous cooperative
UWA communications where only the destination node is
aware of the relative delays among the nodes. Existing sig-
naling solutions for asynchronous radio terrestrial cooperative
communications rely on quasi-static fading channels with
limited delays among signals received from different relays
at the destination, e.g., see [5] and references therein, in
which every transmitted block is preceded by a time guard
not less than the maximum possible delay among the relays.
Therefore, we cannot directly apply them for cooperative
UWA communications. Our main objective is to develop new
OFDM based signaling solutions to combat the asynchronism
issues arising from excessively large relative delays without
preceding each OFDM block by a large CP (in the order of
the maximum possible relative delay).
In systems employing OFDM, e.g., [6], [7], the existing
solutions are effective when the maximum length of the
relative delays among signals received from various nodes are
less than the length of an OFDM block which is not a practical
assumption for the case of UWA communications. In [6], a
space-frequency coding approach is proposed which is proved
to achieve both full spatial and full multipath diversities. In [7],
OFDM transmission is implemented at the source node and
relays only perform time reversal and complex conjugation.
A trivial generalization of existing OFDM-based results to
compensate for large relative delays may be to increase the
OFDM block lengths. The main drawback in this case is that
inter carrier interference (ICI) is increased due to the time
variations of the UWA channels. Another trivial solution is to
increase the length of the CP. This is not an efficient solution
either, since it dramatically decreases the spectral efficiency
of the system.
There are several single carrier transmission based solutions
reported in the literature as well, e.g., [1], [8], [9], [10],
[11], [12]. In [1] a time reversal distributed space time block
code (DSTBC) is proposed for UWA cooperative commu-
1368 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 13, NO. 3, MARCH 2014

1 ()
nication systems under quasi-static multipath fading channel
conditions. In [8], a DSTBC transmission scheme by decode 1
and forward (DF) relaying is proposed which achieves both ()

()
full spatial and multipath diversities. A distributed space time
trellis code with DF relaying is proposed in [9], [10] which 2
under certain conditions can achieve full spatial and multipath
2 ()
diversities. In [11], [12], space time delay tolerant codes are
Fig. 1. Relay channel with two relays.
proposed for decode and forward relaying strategies where
in [11] a family of fully delay tolerant codes and in [12]
1 1 2 3 4
a family of bounded delay tolerant codes are developed for 2
1 2 3 4
asynchronous cooperative systems. time

In this paper, we focus on OFDM based cooperative UWA

Fig. 2. The structure of the received OFDM blocks from two different relays
communication systems with full-duplex AF relays where all of the proposed delay diversity scheme for a relative delay of D seconds.
the nodes employ the same frequency band to communicate
with the destination. We assume an asynchronous opera- a delay diversity structure is obtained. For illustration, Fig. 2
tion and potentially very large delays among different nodes shows the received OFDM block structure of the proposed
(known only at the destination). We present a new scheme delay diversity scheme from two different relays with a relative
which can compensate for the effects of the long delays among delay of D seconds. In Fig. 2, D is in a range that each
the signals received from different nodes without adding an block relayed through the relay R1 is overlapped with its
excessively long CP. We demonstrate that we can extract preceding block relayed through the relay R2 . E.g., under
delay diversity out of the asynchronism among the cooperating quasi-static fading scenario, each subcarrier of a received
nodes. The main idea is to add an appropriate CP (much block is a summation of the corresponding subcarriers from
shorter than the long relative delays among the relays) to two successively transmitted blocks which results in a delay
each OFDM block at the transmitter side to combat multipath diversity structure [13].
effects of the channels and obtain a delay diversity structure
at the destination.
A. Signaling Scheme
The paper is organized as follows. In Section II, the system
model and the structure of the OFDM signals at the source, At the transmitter, we employ a conventional OFDM trans-
relays and destination are presented. The proposed signaling mission technique with N subcarriers over a total bandwidth
scheme which includes appropriate CP addition at the source of B Hz. We consider successive transmission of M data
and CP removal at the destination is explained in Section III. blocks of length N symbols. In discrete baseband signaling
Furthermore, it is shown that the proposed scheme gives a form, the m-th (m ∈ {1, . . . , M }) data vector (in time)
delay diversity structure at the destination for large relative is denoted by X m = [X0m , . . . , XN m
−1 ]
T
and the samples
delays among the relays. In Section IV, we present another of the m-th transmitted OFDM block are represented by
transmission scheme which is also useful for delay values less xm = IFFT(X m ) = [xm 0 , . . . , xN −1 ] , where (.)
m T T
de-
than one OFDM block and provides the same delay diversity notesthe transpose operation. Therefore, we have xm n =
N −1 m j 2πk
√1 X e N n . After adding a CP of length N
structure for longer delay values. In Section V, the PEP N k=0 k CP
analysis of the system under both quasi-static and block fading to xm , the CP-assisted transmission block x̄m results. By
channel models is provided. In Section VI, the performance digital to analog (D/A) conversion of x̄m with sampling period
of the proposed scheme is evaluated through some numerical Ts = B1 seconds, we obtain the continuous time signal x̄m (t)
examples. Finally, conclusions are given in Section VII. with time duration of T = (N + NCP )Ts seconds which can
be written as
II. S YSTEM AND S IGNAL M ODELS N −1
1  m j N2πk
We consider a full-duplex AF relay system with two relays, x̄m (t) = √ Xk e Ts t R(t), (1)
N k=0
shown in Fig. 1, in which there is no direct link between source
n = x̄ (nTs ), R(t) = u(t + NCP Ts ) − u(t − N Ts )
(S) and destination (D), and the relays help the source deliver where xm m
its data to the destination by using the AF method. No power and u(t) denotes the unit step function. Furthermore, for the
allocation strategy is employed at the relay nodes and they use continuous time transmitted signal x̄(t), we can write x̄(t) =
 M
m=1 x̄ (t − (m − 1)T ).
fixed power amplification factors. Note that the model can be m
generalized to a system with an arbitrary number of relays and At the i-th relay (i ∈ {1, 2}), the signal ȳi (t) is received,
a direct link between source and the destination, and optimal hence the part of ȳi (t) corresponding to the m-th transmitted
power allocation can be used in a straightforward manner. We block, i.e., ȳim (t) = ȳi (t + (m − 1)T )R(t), can be written as
assume that the channels from the source to the relays and the  ∞
relays to the destination are time-varying multipath channels ȳi (t) =
m
x̄m (t − τ )hm
i (t, τ )dτ + z1,i (t)
m

where hi (t, τ ) and gi (t, τ ) represent the source to the i-th −∞

relay and the i-th relay to the destination channel responses  ∞ 
+ x̄m (t − (m − m)T − τ )hm i (t, τ )dτ , (2)
at time t to an impulse applied at time t − τ , respectively.
m =m −∞
Before presenting the system model of the new relaying   
scheme, we would like to give an example to demonstrate how ISI
RAHMATI and DUMAN: ACHIEVING DELAY DIVERSITY IN ASYNCHRONOUS UNDERWATER ACOUSTIC (UWA) COOPERATIVE COMMUNICATION . . . 1369


m
where z1,i (t) = z1,i (t + (m − 1)T )R(t), hm i (t, τ ) = hi (t + 1 1 2 3 4
(m − 1)T, τ )R(t) and z1,i m
(t) are independent complex Gaus- 2  1 2 3 4

sian random processes with zero mean and power spectral time
2
density (PSD) of σ1,i . By taking only the resolvable paths
Lh Fig. 3. The structure of the received signal.
into account, we can write hi (t, τ ) = l=1i hi,l (t)δ(τ −τhi ,l ),
where Lhi denotes the number of resolvable paths from the 0
source to the i-th relay, hi,l (t) are independent zero-mean

Alignment

N-FFT
(for different i and l) complex Gaussian wide-sense stationary ( ) ( ) Remove

S/P

P/S
A/D
(WSS) processes with a total envelope power of σh2 i ,l (i.e., CP

independent time-varying Rayleigh fading channel tap gains) −( − 1) − ( − ) + −1 −1

Lhi 2
assuming that l=1 σhi ,l = 1, and τhi ,l ≥ 0 denotes the
delay of the l-th resolvable path from the source to the i-th Fig. 4. The structure of the receiver.
relay. Assuming τhi ,Lhi ≤ NCP Ts , i.e., the length of the CP
overhead is greater than the delay spread of the channel (the
main job of the CP to guarantee robustness against multipath), spread of the overall channel experienced at the destination
Lhi m through Ri .
and defining I1,im
(t) = l=1 hi,l (t)x̄m−1 (t + T − τhi ,l ), we
can rewrite (2) as
III. D ELAY D IVERSITY S TRUCTURE
Lhi
 To achieve a delay diversity structure and overcome ISI at
ȳim (t) = i,l (t)x̄ (t − τhi ,l ) + I1,i (t) + z1,i (t).
hm m m m
(3) the destination, we need to add an appropriate CP at the source
l=1
and perform CP removal at the destination.
We assume that the signal passing through the second relay
is received D seconds later than the signal passing through the A. Appropriate CP Length
first relay and we also assume τhi ,1 = τgi ,1 = 0 for i = {1, 2}, In a conventional OFDM system, if we have a window of
i.e., the delay spread of the channel hi (gi ) is τhi ,Lhi (τgi ,Lgi ). length (N +L)Ts seconds corresponding to one OFDM block,
Therefore,
√ by denoting the amplification factor of the i-th relay then by removing the first L samples of the considered window
by Pi , for the received signal at the destination ȳ(t), we have and feeding the remaining N samples to the FFT block, the ISI
 ∞ is completely removed. Therefore, in our scheme, to guarantee
ȳ(t) = P1 ȳ1 (t − τ )g1 (t, τ )dτ robustness of the system against ISI, we need to have an
−∞ overlap of length (N +L)Ts seconds between two blocks
 ∞
+ P2 ȳ2 (t − D − τ )g2 (t, τ )dτ + z2 (t), (4) received from two different relays at the destination. Fig. 3
−∞ shows the structure of the received signal at the destination for
where z2 (t) is a Gaussian random processes with zero mean the case that the blocks relayed by R2 are received D seconds
Lg later than the blocks relayed by R1 , where T < D < 2T
and PSD of σ22 . By employing gi (t, τ ) = l=1i gi,l (t)δ(τ −
τgi ,l ) in (4), we obtain and d = mod(D, T ) with d = mod(D, T ) denoting the
remainder of division of D by T . To obtain the appropriate

Lg1
overlap structure, we need to have T − d ≥ (N + L)Ts or
ȳ(t) = P1 g1,l (t)ȳ1 (t − τg1 ,l ) d ≥ (N + L)Ts or both which results in T ≥ 2(N + L)Ts ,
l=1 i.e., NCP ≥ N + 2L.
Lg2

+ P2 g2,l (t)ȳ2 (t − τg2 ,l − D) + z2 (t). B. Received Signal at the Destination
l=1
The baseband signaling structure of the receiver is shown
Lg1 √ in Fig. 4, where ȳ m = [ȳ0m , . . . , ȳN +NCP −1 ] denotes the sam-
m
Defining z(t) = z2 (t) + l=1 P1 g1,l (t)z1,1 (t − τg1 ,l ) +
Lg2 √ pled vector of the received signal in the m-th signaling interval
l=1 P2 g2,l (t)z1,2 (t − τg2 ,l ) which represents a Guassian
random process conditioned on known gi,l (t) for all i and l, and b is the starting point of the m-th FFT window which is
we can write decided by the destination based on the delay value D. Since
Lh1 NCP ≥ N + 2L, by defining ȳ m (t) = ȳ(t + (m − 1)T )R(t),

Lg1
√ 
ȳ(t) = P1 g1,l (t) h1,q (t−τg1 ,l )x̄(t−τg1 ,l −τh1 ,q )+z(t) we can write
l=1 q=1
Lg2 Lh2
ȳ m (t) = I1m (t) + I2m (t) + z m (t)
 √  Lh1
+ P2 g2,l (t) h2,q (t−D−τg2 ,l )x̄(t−D−τg2 ,l −τh2 ,q ). 
Lg1

l=1 q=1 + P1 g1,l (t)
m
1,q (t − τg1 ,l )x̄ (t − τg1 ,l − τh1 ,q )
hm m

l=1 q=1
Note that without conditioning on gi,l (t), z(t) represents
 Lg2 
Lh2
a complex random process with zero mean and PSD of
σ 2 = P1 σ1,1
2 2
+ P2 σ1,2 + σ22 . Therefore, we define the received + P2 g2,l (t)
m
hm−BD
2,q (t − dr − τg2 ,l )×
as P1σ+P
l=1 q=1
signal

to noise ratio (SNR)

 2
2
. We also define L =
maxi (τhi ,Lh +τgi ,Lgi ) × x̄ m−BD
(t − dr − τg2 ,l − τh2 ,q ) ,
Ts
i
, where τhi ,Lhi + τgi ,Lgi is the delay
1370 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 13, NO. 3, MARCH 2014

time time
+

3

1 ()
 +

= 2, = −( − )
2 
1 ()
−
= 1, =

()
Fig. 5. Example of different situations for BD and dr .

()

where Fig. 6. Different possible FFT windowings for different ranges of d (a) d ≥
(N + 2L)Ts , (b) (N + L)Ts ≤ d < (N + 2L)Ts , (c) N Ts < d <
I1m (t) = (N + L)Ts , and (d) d ≤ N Ts
Lg1 Lh1
 √ 
m
P1 g1,l (t) 1,q (t − τg1 ,l ) × x̄
hm m−1
(t + T − τg1 ,l − τh1 ,q ) C. Appropriate CP Removal at the Destination
l=1 q=1
To take FFT at the destination, we need to choose the
and
FFT window by appropriate CP removal. Since the received
Lg2 L
√  
h2
OFDM blocks are not synchronized, we align the receiver
I2m+BD (t) = P2 m+BD
g2,l (t) hm2,q (t−dr −τg2 ,l )×
FFT window with one of the relays. By precise alignment,
l=1 q=1
  an overlap of length (N +L)Ts seconds between the OFDM
x̄m−1 (t+T −dr −τg2 ,l −τh2 ,q )+x̄m+1 (t−T −dr −τg2 ,l −τh2 ,q )
blocks received through R1 and R2 can be achieved which
represent the ISI, and BD and dr , as shown in Fig. 5, denote is determined with the value of d. Note that an overlap of
the effective OFDM block delay and effective residual delay at least N +L samples is necessary to guarantee robustness
observed at the destination, respectively. For BD and dr , we of the transmission against ISI. As shown in Fig. 6, for
have  D d ≥ (N + L)Ts , the receiver FFT window is aligned with
 T  , d ≤ (N + L)Ts R2 and for d < (N + L)Ts it is aligned with R1 . The only
BD = , (5)
DT , d > (N + L)Ts effect of unaligned FFT windowing in time at the destination,
and  as long as appropriate CP removal is done, is phase shift at
d , d ≤ (N + L)TS the frequency domain included in the definition of GH mi .
dr = , (6)
d−T , d > (N + L)Ts
D. Detection by Viterbi Algorithm
respectively (note that when m−BD < 0, x̄m−BD (t) = 0 for
all values of t). More precisely, BD represents the number of For the time-invariant channel scenario the noise samples
block delays between two received OFDM blocks which have Zkm are independent complex Gaussian random variables for
at least an overlap of length (N + L)Ts seconds (necessary to all m and k and i.i.d. for any specific k. Therefore, for time-
combat the ISI). As discussed in Section III-A, by choosing invariant channel conditions, N parallel Viterbi detectors with
NCP ≥ N + 2L, achieving the appropriate overlap between M BD states (assuming M-PSK modulation) can be employed
the received OFDM blocks is guaranteed. By appropriate CP for ML detection of the transmitted symbols, where the k-
removal (whose details are explained in Section III-C), y m is th Viterbi detector gets Y k as input to detect the transmitted
obtained as y m = [ȳ m (bTs ), . . . , ȳ m ((b + N − 1)Ts )]. By tak- symbols over the k-th subcarrier. On the other hand, for the
ing FFT of y m , we have Y m = [Y0m , . . . , YNm−1 ] = FFT(y m ) time-varying channel scenarios, the received noise samples at
where Ykm are given in (7) at the top of the next page, which each OFDM block are dependent complex Gaussian random
can be written as variables conditioned on known channel state information.
Note also that the noise samples corresponding to different
Ykm = GH m
1 (k)X
m
+ GH m−BD
2 (k)X m−BD + Zkm , FFT windows at the destination are independent but not nec-
essarily identically distributed. However, by approximating the
where GH m i (k) = [GHi [k, 0], · · · , GHi [k, N − 1]] with
m m

GHi [k, k ] given in (8) at the top of the next page and Zkm =
m received noise samples Zkm as independent complex Gaussian
b+N −1 m random variables, Viterbi algorithm can be employed as a
√1
j2πn

N n=b z (nTs )e− N k conditioned on channel state

information are complex Gaussian random variables with zero detector at the destination and extract delay diversity out of
mean. Hence, by defining X m = 0N for m < 1 and m > M the asynchronous system.
and X m = [X0m , X1m , · · · , XN The complexity of the Viterbi algorithm for the time varying
−1 ] for 1 ≤ m ≤ M , we can
m T

write case is prohibitive due to the ICI effects. Therefore, we

implement a suboptimal detector in which we ignore the ICI
Y m = GH m 1 X
m
+ GH m−BD
2 X m−BD + Z m , (9) effects and assume that Zkm are i.i.d. for any given subcarrier
 T k, and employ the same structure as in the time-invariant case.
T T
where GH m i = GH m i (0) , . . . , GH i (N − 1)
m
. In Hence, Y k = [Yk1 , . . . , YkM ] is given to the k-th Viterbi
fact, GH mi represents the effective S − Ri − D channel seen decoder where [Y0m , . . . , YNm−1 ] = diag(Y m ) and diag(A),
by the destination in frequency domain which depends on both with A being a square matrix, denotes a vector of diagonal
S − Ri − D channel and the position of the FFT window. elements of A.
RAHMATI and DUMAN: ACHIEVING DELAY DIVERSITY IN ASYNCHRONOUS UNDERWATER ACOUSTIC (UWA) COOPERATIVE COMMUNICATION . . . 1371

N −1 −1
1  m −j 2πn k
b+N
1 2πn
Ykm = √ yn e N = √ ȳ m (nTs )e−j N k
N n=0 N n=b
−1 
b+N  L g1

Lh1

1
=√ P1 g1,l (nTs )
m
hm1,q (nTs − τg1 ,l )x̄ (nTs − τh1 ,q − τg1 ,l ) + z (nTs )
m m
N n=b l=1 q=1


Lg2

Lh2
−j 2πn
+ m
P2 g2,l (nTs ) hm−BD
2,q (nT s − dr − τg2 ,l )x̄m−BD
(nT s − d r − τh2 ,q − τ )
g2 ,l e N k. (7)
l=1 q=1

√
b+N −1 Lg Lh1
P1  1 m 
j 2πn 
−j 2πk

N Ts (τg1 ,l +τh1 ,q ) ,
GH1m [k, k  ] = g1,l (nTs ) 1,q (nTs − τg1 ,l ) e
hm N (k −k) e
N q=1
n=b l=1
√ −1 
b+N Lg2

Lh2
P2
(nTs − dr − τg2 ,l ) ej N Ts [n(k −k)Ts −k (dr +τg2 ,l +τh2 ,q )] ,
2π  
GH2m−BD [k, k  ] = m
g2,l (nTs ) hm−BD
2,q (8)
N q=1
n=b l=1

m
IV. A M ODIFIED A MPLIFY AND F ORWARD R ELAYING then GH1 [k, k  ] and GH1m [k, k  ] have the same density
S CHEME function as well.
m
In Section III, we presented a new scheme which achieves For the block fading scenario, where GH1 (k, k  ) = 0 and
 
the delay diversity structure for BD > 0; however, for GH m2 (k, k ) = 0 for k
= k and all m, we arrive at
BD = 0, the scheme does not provide spatial diversity. To mm ∗
Ykm = GH1,k XN −k + GH2,k + Zkm
m−BD m−BD
Xk (11)
address this limitation, we present a slightly modified version
m
of the proposed scheme in this section which achieves the for k
= 0, and Y0m = GH1,0 X0m ∗ + GH2,0
m−BD m−BD
X0 + Z0m
delay diversity structure for large values of the relative delay for k = 0.
D, i.e., BD ≥ 1, and also provides diversity for small values For BD = 0, if we focus on Ykm and YNm−k−1 (k
= 0, N2 ),
of D, i.e., BD = 0. then we have
We still employ full duplex amplify and forward relay   
Ykm Xkm Zkm
nodes. Similar to the scheme described in Section III, the = Ck
m
+ (12)
YNm−k ∗ XNm ∗
−k
m ∗
ZN −k
second relay simply amplifies and forwards its received signal.
The only modification is at the first relay in which instead with  m 
m
of forwarding the received signal unchanged, a complex GH2,k GH1,k
k =
Cm m ∗
m ∗
.
conjugated version of the received signal is amplified and GH1,N −k GH2,N −k
forwarded to the destination. At the receiver, if the signal
Therefore, based on the optimal maximum likelihood (ML)
from the second relay is received D seconds later than the m ∗
detection criteria (assuming [Zkm , ZN −k ] as white Gaussian
signal from the first relay, by following the same steps as in
noise), for the optimal detector, we obtain
Section III-B, we can write    
  m∗ m  m Xkm
P1  −j 2πn
b+N−1
[X̂km , X̂N
m
−k ] =argmax Re Y , YN −k C ∗
Ykm = GH m−BD
2 (k)X m−BD +Zkm + e N
k
× Xkm ,XN
m
−k
k k XNm
−k
N 
1  m∗ m  mH m
n=b m
  Xk
Lg1 Lh1
 − Xk , XN −k C k C k m ∗ ,
m
g1,l (nTs ) ∗ m
1,q (nTs −τg1 ,l ) x̄ (nTs −τh1 ,q −τg1 ,l )
hm ∗ 2 XN −k
l=1 q=1 which offers spatial diversity. Note that for k = 0 and
m m
= GH1 (k)X̃ + GH m−BD
2 (k)X m−BD + Zkm , (10) k = N2 , no diversity is provided; however, in detection of
m m m the remaining sub carriers spatial diversity is extracted out of
where GHi (k) = [GHi [k, 0], · · · ,GHi [k, N − 1]] with the proposed system. The worst case is to not occupy the sub-
√ carriers k = 0 and k = N2 for data transmission which results
P1   m
b+N−1 Lg1
m
GH1 [k, k ] = g1,l (nTs ) × in a very small loss in rates, e.g., in an OFDM transmission
N
⎡ n=b l=1
⎤ with N = 1024 subcarriers, the system experiences a rate loss
Lh1
 2πn (k −k) −j 2πk
 of less than 0.2%.
×⎣ ∗ (τg1 ,l +τh1 ,q ) ⎦
1,q (nTs − τg1 ,l ) e
hm j N e N Ts ,
q=1
On the other hand, for BD > 0, the received signal
preserves the delay diversity structure presented in Section II.
m
GH m−BD
2 (k) is as given in (7), X̃ = Obviously, for k ∈ {0, N2 }, Y0m depends only on Xkm and
[X0∗ , XN
∗
−1 , . . . , X ∗
1 ], and Z m
k has the same statistical Xkm−BD and the delay diversity structure is similar to the
∗
properties as in (7). Obviously, since hm 1,q (nTs − τg1 ,l ) and previous scheme. For k
= 0 and N2 , by focusing on Xkm and
h1,q (nTs − τg1 ,l ) have the same probability density function,
m
XNm
−k and considering the block-fading case, we have
1372 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 13, NO. 3, MARCH 2014

scenario, we assume that the channels from the source to the

⎡ ⎤ ⎡ 1 1 ∗ ⎤ relays have significantly higher SNRs than the channels from
Yk1 GH1,k XN−k
⎥ ⎢ ∗⎥
∗ the relays to the destination. We first give the considered block
⎢ YN−k ∗
1+BD 1+BD
Xk1+BD + GH2,N−k
1 ∗ 1
⎢ 1+2BD ⎥ ⎢ GH1,k XN−k ⎥
⎢Y ⎥=⎢ 1+2BD ∗ ⎥ fading channel model. We then provide a discussion on the
⎦ ⎢ Xk1+BD ⎥
1+2BD 1+BD
⎣ k ⎣ GH1,k XN−k + GH2,k ⎦ discrete noise samples at the destination under the block fading
.. ..
. . channels and at the end provide the PEP analysis for which
⎡ ⎤ similar to the quasi-static channel conditions, we assume that
Zk1
∗ no coding is employed over the subcarriers and focus on the
⎢ ZN−k
1+BD ⎥
⎢ ⎥
+ ⎢ Z 1+2BD ⎥ . spatial diversity analysis of the system.
⎣ k ⎦
.. 1) Block Fading Frequency-Selective Channel Model:
. Here, we follow the same channel model and procedure used
Therefore, we expect to achieve the same performance as in [14] in which the PEP performance analysis of space-
the scheme introduced in Section II. Moreover, by this new time coded OFDM multi-input multi-output (MIMO) system
scheme we are also able to extract diversity out of the system over correlated block fading channels has been considered.
for BD = 0. The main difference between the system model in [14] and
the one in this paper is in the effective channel model and
V. PAIRWISE E RROR P ROBABILITY A NALYSIS the noise experienced at the destination. In fact, we need to
Design of the space-time codes is out of the scope of make some simplifying approximations to be able to derive
this work; however, we present the PEP performance analysis a closed form upper bound on the PEP of the system. By
of the system under quasi-static and block fading frequency approximating the received noise samples Zkm as complex
selective channel conditions which can be useful in a diversity Gaussian random variables (see Section V-B2), we provide a
order analysis of the proposed scheme and possible space-time PEP analysis under the block fading channel scenario in which
code designs. In the following, we present the PEP analysis for channel coefficients are fixed during each block transmission
the quasi-static and block fading frequency selective channels, and change block by block based on the following Fourier
respectively. expansion relation [15] (for ease of presentation we assume
that Lgi = Lhi = L, τhi ,l = τgi ,l = τl and all the channels
A. Quasi-Static Frequency-Selective Channels experiencing the same Doppler frequency shift fd )
Lt −1

In this section, we consider the PEP performance analysis  2

2πn(mT )
hmi,l (t) = hi,l ≈ αi,l [n]ej M T ,
m
for ML detection presented in Section III-D. We provide (14)
Lt −1
the result under the condition that the channels from the n=− 2
Lt −1
and
source to the relays have significantly higher SNRs than the  2
2πn(mT )
channels from the relays to the destination, i.e., σ12  σP2i .
m
gi,l (t) = gi,l
m
≈ βi,l [n]ej MT , (15)
1,i 2
L −1
We assume that the channels are quasi-static Rayleigh fading, n=− t2
i.e., the channel gains in time domain are random variables in which αi,l [n] and βi,l [n] are independent circularly
but fixed for the transmission of M consecutive OFDM symmetric complex Gaussian random variables with
σ2 σ2
blocks. We denote hm i,l (nTs ) = hi,l and gi,l (nTs ) = gi,l for
m
zero mean and variance of Lhit,l and Lhit,l , respectively,
n = {0, . . . , N − 1} and m = {1, . . . , M }, where hi,l and gi,l with Lt = 2fd M T + 1 . hm m
i,l and gi,l can also be
are zero mean circularly symmetric complex Gaussian random represented as hi,l = αi (l) wt (m) and gi,l = βi (l)T w t (m),
m T m

variables with variances of σh2 i ,l and σg2i ,l , respectively, with where wt (m) = [e−j2πMfd T , . . . , 1, . . . , ej2πMfd T ]T ,
Lhi 2 Lgi 2  T
l=1 σhi ,l = 1 and l=1 σgi ,l = 1. For the PEP of the under αi (l) = αi,l [− Lt2−1 ], . . . , αi,l [ Lt2−1 ] , and
 Lt −1 Lt −1 T

consideration scenario, we obtain (see Appendix A), β i (l) = βi,l [− 2 ], . . . , βi,l [ 2 ] .
  Since all the channels are block fading, for GH1m [k, k  ] and
P X k → X k ≤
      GH2m [k, k  ], we have
8σ24 P1 4 4 P2 4 4
log 1+ 2 sk −fk log 1+ 2 sk −fk , Lg1 L 
P1 P2 (s4k −fk4 ) 4σ2 4σ2 √  h1
m m −
2πjk (τh ,q+τg ,l )
1 1
GH1m [k, k ]= P1 g1,l h1,q e N Ts δ(k−k ),
(13)
l=1 q=1
 
 M  
where fk2 =  m=BD+1 (Xkm − Xkm )(Xkm−BD − Xkm−BD )∗ 
Lg2
√  k τg ,l
−2πj NkT dr m −2πj N T2
M GH2m [k, k ]= P2 e s g2,l e s

and s2k = m 2
m=1 |Xk − Xk | . We observe from (13)
m
l=1
that the system achieves the diversity order of 2. For Lh2
 k τh ,q

−2πj 2
instance, for P1 = P2 , we have SN R = 2P 1
and hm
2,q e
N Ts δ(k − k ).
σ22
  −2
  2 q=1
P X k → X k ≤ 32SN R
(s4 −f 4 )
log 1 + SN8 R s4k − fk4 .
k k GHim [k, k] can also be written as GH2m [k, k] =
k
m −2πj N Ts dr
m
G2,k H2,k e and GH1 [k, k] m
= m
G1,k H1,k m
B. Block Fading Frequency-Selective Channels L 2πkτ
−j N Tsl
In this section, we analyze the PEP performance of the with Hi,k =
m
l=1 hi,l e = hi (m)w f (k),
T
L 2πkτ
−j N Tsl
proposed scheme under block fading frequency selective Gi,k =
m
l=1 gi,l e = g i (m)w f (k) where
T

channels. Similar to the analysis for the quasi-static fading hi (m) = [hm i,1 , · · · , h m T
i,L ] , g i (m) = [gi,1 , · · · , gi,L ]
m m T
RAHMATI and DUMAN: ACHIEVING DELAY DIVERSITY IN ASYNCHRONOUS UNDERWATER ACOUSTIC (UWA) COOPERATIVE COMMUNICATION . . . 1373

2πkτ1 2πkτL
and w f (k) = [e−j N Ts , . . . , e−j N Ts ]T . On the other hand, as i.i.d. complex Gaussian random variables with zero mean
by defining W t (m) = diag{wt (m), . . . , wt (m)}LLt ×L , and variance of σ 2 to match with the sub-optimal detector
v i = [αTi (1), . . . , αTi (L)]T , q 1 = [βT1 (1), . . . , β T1 (L)]T we considered for the general time-varying channel case in
k
and q 2 = [β T2 (1), . . . , β T2 (L)]T e−2πj N Ts dr we obtain Section III-D.
= v Ti W t (m)w f (k) and Gm 3) PEP Analysis: Conditioned on known channel state in-
i,k = q i W t (m)w f (k).
m T
Hi,k
formation at the receiver, for the considered Viterbi detector1,
2) Discussion on the Statistics of the Noise Samples: For we have
the block fading scenario, we have hmi,l (t) = hi,l and gi,l (t) =   2
m m
M+BD

m
gi,l , hence we can write 
X k = arg min  m−BD 
Yk − GH1,k Xk − GH2,k Xk
m m m m
 ,
X k m=1
1 −1
b+N 
Lg1

Zkm = √ z2m (nTs )+ m m
P1 g1,l z1,1 (nTs − τg1 ,l ) where Xkm = 0 for m < 1 or m > M , and GH1,k
m
=
N k
m −2πj N Ts dr
Gm1,k H1,k and GH2,k = G2,k H2,k e
n=b l=1 m m m
. Therefore,

Lg2
j2πn under the assumption that σ12 are sufficiently larger than σP2i ,
+ P2 g2,l z1,2 (nTs − τg2 ,l ) e− N k.
m 1,i 2
we obtain (see Appendix B)
l=1 "  rk  2 #
 H 
Since the noise samples from the OFDM block durations  1 − 12 c=1 λk,c U k,c q (k)
m and m  (m
= m ) are independent then obviously P (X k → X k ) ≤ E e 4σ2

2
E{Zkm Zkm } = 0. Furthermore, for E{|Zkm |2 }, we can write
1$
rk
 b+N−1 b+N−1 
1   ≤ P EPk,c , (16)
E{|Zkm |2 } = E z2m (nTs ) 2 c=1
N n=b v=b
 where

Lg1
√ 
Lg2
√  
+ m m
P1 g1,l z1,1 (nTs − τg1 ,l ) + m m
P2 g2,l z1,2 (nTs − τg2 ,l ) 2σ22
P EPk,c = EV
l=1 l=1 2σ22 + λk,c V
 Lg1 % &
 √  2σ2
m∗
× z2 (vTs ) + m ∗
P1 g m∗1,i z1,1 (vTs − τg1 ,i ) πp 2σ22 λk,c σχ,k,p
2
2
2 2σ22
= 2 2
e σq,p
E1 2 2
i=1 σχ,k,p σq,p λk,c λk,c σχ,k,p σq,p
  p∈S0

Lg2
√
+ m ∗
P2 g 2,i z1,2 (vTs − τg2 ,i ) e
m∗ −
j2π(n−v)
N
k
. 
J 
(2σ22 )Nj (−1)Nj −1
+ 2 2 )Nj
×
i=1
j=1 p∈Sj
(Nj − 1)!(λk,c σχ,k,p σq,p
By using the facts that zj,i (t) are zero mean independent  2
2σ2
% &
Gaussian random processes (as a result E{z1,1 (t)z1,2 (t )} = 0 λk,c σ2 2 2σ 2
2
for all t and t ) and E {zj,i (t)zj,i (t )} = σj,i
2
δ(t − t ), we × e χ,k,p
σq,p
E1 2 2
λk,c σχ,k,p σq,p
obtain % &k
Nj −1
 
 b+N−1
b+N−1  2
−λk,c σχ,k,p 2
σq,p
1 + (k − 1)!
E{|Zkm |2 } = σ22 δ(n − v) 2σ22
. (17)
N n=b v=b k=1

 
Lg1 Lg1
m m∗ 2
+P1 g1,l g1,i σ1,1 δ(nTs − τg1 ,l − vTs + τg1 ,i ) VI. S IMULATION R ESULTS
l=1 i=1 To provide numerical examples we assume that the to-
  
tal occupied bandwidth is 8 kHz (over the frequency band
Lg2 Lg2
j2π(n−v)
m m∗ 2 −
+P2 g2,l g2,i σ1,2 δ(nTs −τg2 ,l −vTs +τg2 ,i ) e N
k
from 12 kHz to 20 kHz). We define fd as the Doppler
l=1 i=1
frequency shift observed at the destination in Hz, and σhi =
which leads to [σhi ,0 , . . . , σhi ,Lhi ]. We assume Pi = 1, τhi ,l = τgi ,l = lTs =
2

Lg1

Lg2 125l μs, and σi,1 = 2σ22 (i ∈ {1, 2}).
E{|Zkm |2 } = σ22 + P1 σ1,1
2 m 2
|g1,l 2
| +P2 σ1,2 m 2
|g2,l | In Figs. 7 and 8, we compare the performance of the pro-
⎡ l=1 l=1
⎤ posed scheme with the performance of the scheme proposed
 in [7] for different values of fd Ts under two different scenar-
2
+ 2σ1,1 Re⎣ m m∗
g1,l g1,i e − j2πk
N f1 ⎦ ios where quadrature phase-shift keying (QPSK) modulated
f1 ,τg1 ,l −τg1 ,i =f1 Ts symbols are transmitted over N subcarriers. The parameters
⎡ ⎤
of the two scenarios are reported in Table I in which to

2
+ 2σ1,2 Re ⎣ m m∗
g2,l g2,i e − j2πk
N f2 ⎦, make a fair comparison, both schemes are set to the same
f2 ,τg2 ,l −τg2 ,i =f2 Ts data transmission rate. We generate time varying Rayleigh
fading channel tap gains following the Jakes’ model [16]. We
where f1 and f2 only take positive integer values. Since Zkm chose [7] for comparison since it also considers an OFDM
are independent for a specific k but not identically distributed,
the optimal ML detection can be obtained by employing 1 For the considered block fading channels, the optimal ML detection

Viterbi detector over the normalized received signals according is obtained by normalizing the received signals over each subcarrier with
variance of its corresponding noise; however, we present the result for the
to E{|Zkm |2 }. However, in the following, we provide the PEP case that there is no normalization to match with the Viterbi detection of the
analysis by approximating the received noise samples Zkm general time-varying case.
 1374 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 13, NO. 3, MARCH 2014

TABLE I
PARAMETERS OF TWO DIFFERENT SCENARIOS USED TO COMPARE THE PROPOSED SCHEME WITH THE SCHEME IN [7].
Scenario Scheme N M D σhi = σgi (j, i ∈ √
{1, 2}) NCP T (ms) Data Rate (kb/s)
Proposed Scheme 512 100 1039Ts [1, 0.8, 0.6]/√2 522 129.25 7.9226
S1
Scheme from [7] 1024 2 1039Ts [1, 0.8, 0.6]/ 2 1044 258.5 7.9226
Proposed Scheme 256 100 527Ts [0.8, 0, 0.6] 266 65.25 7.8467
S2
Scheme from [7] 512 2 527Ts [0.8, 0, 0.6] 532 130.5 7.8467
Scenario S1 Scenario S2
100 100
+ + + + + + + + + + + + + + + +
   
+ + + + + + +

+ + + 
+ + +
10−1 10−1 +
 
 
+ +
 

 

 


10−2 10−2





BER

BER


10−3 10−3

Results from [7] Results from [7]

Proposed Scheme Proposed Scheme
10−4 + fd Ts = 10−3 10−4 + fd Ts = 10−3

fd Ts = 10−4 
fd Ts = 10−4
fd Ts = 0 fd Ts = 0

10−5 10−5
20 22 24 26 28 30 32 34 20 22 24 26 28 30 32 34
SNR (dB) SNR (dB)

Fig. 7. Comparison between the performance of the proposed scheme with Fig. 8. Comparison between the performance of the proposed scheme with
the scheme proposed in [7] under the scenario S1 . the scheme proposed in [7] under the scenario S2 .
100
based cooperative transmission with full-duplex AF relays. 
Upper Bound (13)
Simulation Results
However, in [7], the relays perform time reversal and symbol
10−1
complex conjugation as well. In Figs. 7 and 8, we observe
that for the time-invariant channel case (fd Ts = 0), the  



performance of both schemes are identical. However, for 10−2




time varying scenarios, the proposed scheme outperforms the 

scheme proposed in [7]. The reason is that, for the range of the 
PEP

10−3 

relative delays considered, to attain the same data rate for both 

schemes, the scheme proposed in [7] transmits longer OFDM 

blocks (larger N ) and as a result more ICI is experienced 10−4 

over the received subcarriers. Obviously, by increasing fd , i.e., 

faster fading conditions, more ICI are experienced over the

10−5 
subcarriers and the performance becomes worse. We observe
that in the SNR range considered, the bit error rate (BER)
of the fast fading scenario (fd Ts = 10−3 ) reaches an error 10−6
0 5 10 15 20 25 30
floor. We expect that for higher SNR values, the slow fading SNR (dB)
scenario (fd Ts = 10−4 ) converges to an error floor as well.
In both cases, we employ a suboptimal Viterbi decoder to Fig. 9. Comparison between the upper bound (13) and actual PEP for
X k = 1M and X k = [−1, 1T M −2 , −1]
T under quasi-static frequency
detect the transmitted signal, which assumes that the noise selective channels.
samples over the same subcarrier of different blocks are i.i.d.
and ignores the ICI effects to reduce complexity.
In Fig. 9, we compare the PEP performance of the proposed X k = 1M , where 1M represents an all one vector, and
scheme with the derived upper bounds for the quasi-static X k = [1TM −1 , −1, 1TM ]T . Note that the considered case is
2 2
frequency selective channel. We consider transmission of the worst case scenario, i.e., gives the maximum PEP among
M = 10 OFDM blocks with N = 64 binary phase-shift all the possible pairwise error events (s4k − fk4 is minimized).
keying (BPSK) modulated subcarriers over multipath channels We observe in Fig. 9 that by increasing SNR we achieve a
with σhi = σgi = [1,0.8,0.6]
√
2
where there is a relative delay tighter upper bound on PEP. Furthermore, as we expect for
of 145 Ts seconds between the signals received from the high SNR values, the diversity order of the system is 2 which
two relay nodes, i.e., BD = 1. Furthermore, we assume is due to the delay diversity structure of the system.
2 2 σ2
that P1 = P2 = 1, σ1,1 = σ1,2 = 102 . Therefore, the In Fig. 10, we compare the derived upper bound on
SN R of the system at the receiver is 3σ5 2 . We consider P (X k → X k ), for X k = 110 and several different X k as
2
RAHMATI and DUMAN: ACHIEVING DELAY DIVERSITY IN ASYNCHRONOUS UNDERWATER ACOUSTIC (UWA) COOPERATIVE COMMUNICATION . . . 1375

Case I Case II Case III Case IV

X k [−1, 1T9 ]T [−1, 1T8 , −1]T [−1T5 , 1T5 ] [−1, 1, −1, 1, −1, 1, −1, 1, −1, 1]T
D(X k , X k ) 2 4 6 10
TABLE II
D IFFERENT CASES CONSIDERED IN F IG . 10 WITH X k = 110 .

100
+ Case I the performance is similar, for time varying cases (typical in
+ + +


+ +
+

Case II UWA communications) the proposed scheme is significantly
  + + 
Case III

 + + superior.
 + + 
Case IV

 

+ + +


 + + +
 

+ + A PPENDIX A
 + +
10−5
 
 + +
 


+ + PEP A NALYSIS FOR Q UASI -S TATIC







F REQUENCY-S ELECTIVE C HANNELS
PEP


 




Under the quasi-static channel conditions, for the channel
 





gains in frequency domain, we have


10−10


Lg1 L 
 

 √  h1
−
2πjk (τh ,q+τg ,l )
1 1

 
GH1m [k, k ]= P1 g1,l h1,q e N Ts δ(k−k ),



l=1 q=1



 and
⎡ ⎤





Lg2


10−15

k τg ,l
GH2 [k, k  ] = P2 ⎣ g2,l e−2πj N Ts ⎦ ×
2
0 5 10 15 20 25
SNR (dB) l=1
⎡ ⎤
Lh2
Fig. 10. Comparison between the upper bound (34) for X k = 110 and  k τh ,q
k
×⎣ h2,q e−2πj N Ts ⎦ e−2πj N Ts dr δ(k − k  ),
2
X k as given in Table II.
q=1

Lg kτg ,l

where for a fixed k, Gi,k = l=1i gi,l e−2πj N Ts and Hi,k =

i
given in Table II under the block fading channel conditions
considered in Section V-B1 with fd Ts = 0.01. We consider the Lhi kτh ,q
−2πj N Tis
q=1 hi,q e are independent complex Gaussian ran-
same transmission specs as considered in the study given in
dom variables with zero mean and unit variance. Hence, for
Fig. 9. As we expected, larger D(X k , X k ) (as defined in (28))
the received signal on the k-th subcarrier, we have
results in a higher diversity order and a better performance
which shows the importance of designing appropriate codes Ykm = GH1 [k, k]Xkm + GH2 [k, k]Xkm−BD + Zkm , (18)
to extract the maximum possible diversity out of the system.
where conditioned on gi,l for all l ∈ {1, . . . , Lgi } and i ∈
VII. C ONCLUSIONS {1, 2}, Zkm are i.i.d. complex Gaussian random variables with
We developed a new OFDM transmission scheme for UWA zero mean and variance of σ22 +P1 |G1,k |2 σ1,1
2
+P2 |G2,k |2 σ1,2
2
.
cooperative communication systems suffering from asynchro- The above relation for BD > 0 is a delay diversity struc-
nism among the relays by considering possibly large relative ture which can be used to extract spatial diversity out of
delays among the relays (typical in UWA systems) and time- the relay system shown by the PEP analysis. If we de-
varying frequency selective channels among the cooperating fine Y k = [Yk1 , . . . , YkM+BD ], Z k = [Zk1 , . . . , ZkM+BD ],
nodes. The main advantage of the proposed scheme is in GH(k) = [GH1 [k, k], GH2 [k, k]] and
managing the asynchronism issues arising from excessively  1
Xk · · · Xk1+BD · · · XkM ··· 0
large delays among the relays without adding time guards (or Xk = ,
0 ··· Xk1 · · · XkM−BD · · · XkM
CP in OFDM-based transmissions) in the order of the max-
imum possible delay, which increases the spectral efficiency we can write Y k = GH(k)X k + Z k . Note that our focus
of the system and improves the performance in time-varying is on extracting spatial diversity out of the asynchronous
channel conditions compared with the existing solutions in cooperative system which is attained in the form of the delay
the literature. In fact, we showed that independent of the diversity. In fact, we assume no explicit channel coding is em-
maximum possible delay between the relays, by adding an ployed across different subcarriers and as a result no multipath
appropriate CP at the transmitter and appropriate CP removal diversity is attained; however, it does not mean that the system
at the receiver, a delay diversity structure can be obtained at does not achieve multipath diversity. Now, let us focus on a
the receiver, where a full-duplex AF scheme is utilized at the given subcarrier, e.g., k-th subcarrier, where conditioned on
relays. Through numerical examples, we evaluated the per- given gi,l for l ∈ {1, . . . , Lgi } and i ∈ {1, 2}, Z k is a complex
formance of the proposed scheme for time-varying multipath white Gaussian
 vector with zero mean and autocorrelation
 ma-
channels with Rayleigh fading channel taps, modeling UWA trix of σ22 + P1 |G1,k |2 σ1,1 2
+ P2 |G2,k |2 σ1,2
2
I M+BD with
channels. We compared our results with those of the existing I M denoting the M by M identity matrix. Therefore, the
schemes and found that while for time invariant channels, Viterbi algorithm proposed in Section III-D can be used as
1376 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 13, NO. 3, MARCH 2014

an ML detection scheme on symbols transmitted over the k- from [18, p. 294. Eq. 2. 15. 1. 2.] we can write
th subcarrier. Furthermore, by employing ML detection at the " # " a2 b2 b2 γ 4
#
d2 2 2 2
− k2 e−βk a2 b2 α221b22+1k
destination
 for the conditional
 PEP over the k-th subcarrier, Edk e 2 = Ea2 ,b1 ,b2
4σ
e k1
P X k → X k |GH(k) which shows the probability of de- α2k b21 + 1
" #
ciding in favor of X k at the receiver while X k is the actual ∞ −βk2 a22 b22 a22 b21 b22 γk4
e 2
e αk b1 +1 2a2 e−a2 da2
2 2
transmitted symbol conditioned on the channel realizations, = Eb1 ,b2
0 α2k b21 + 1
we have  
d2 1
  1 − 2 k
= Eb1 ,b2 ,
P X k →X k |GH(k) ≤ e 4(σ2 +P1 |G1,k | σ1,1 +P2 |G2,k | σ1,2 ), 1+α2k b21 +b21 b22 α2k βk2 +βk2 b22 −b21 b22 γk4
2 2 2 2

2 (21)
(19)
Furthermore it follows from the inequality
where in deriving the last inequality the Chernoff bound is α2k βk2 ≥ γk4 that 1+α2 b2 +b2 b2 α2 1β 2 +β 2 b2 −b2 b2 γ 4 ≤
employed [17, p. 58], dk = ||GH(k)(X k − X k )|| and ||e|| 1
k 1 1 2 k k k 2 1 2 k
   which leads to
denotes the Euclidean length of vector e. Therefore, we have α2 β 2 −γ 4 2 2 4
α β −γ
k k k α2k b21 +1 k k k βk2 b22 +1
α2 β 2 α2 β 2
  1 d2
− 4σk2
k k k k

P X k → X k ≤ Edk {e }, (20) (since b1 and b2 are independent)

2 ⎧ ⎫ ⎧ ⎫
⎧ ⎫ ⎪ ⎪ ⎪ ⎪
d2 ⎪ ⎪ ⎪ ⎪
where Edk {.} denotes the expected value with respect to the ⎨ − k2
4σ
⎬ ⎨ 1 ⎬ ⎨ 1 ⎬
≤Eb1 Eb2
random variable dk . Under the assumption that σ12  σP2i , E dk e 2
⎩ ⎭ ⎪
⎪ 2 2 4 ⎪ ⎪ 2 2 4 ⎪
1,i 2 ⎭ ⎪
⎩ αk β2k−2γk α2 b2 +1⎪ ⎩ αk β2k−2γk β 2 b2 +1⎪
⎭
k 1 k 2
d2k d2k αk βk αk βk
we obtain σ2  σ22
.
Furthermore, due to the definition of the βk
⎛ ⎞

1
P1  ⎜ ⎟
M αk α2 β 2 −γ 4 βk
= e k k k E1 ⎝ ⎠×
Euclidean distance, by defining α2k = |X m − Xkm |2 , α2k βk2 − γk4 2 2 4
αk αk βk − γk
4σ22 m=1 k ⎛ ⎞
P2  m
M 
αk
α2 β 2 −γ 4⎟ ⎜ αk
βk2 = 2 |X − Xkm |2 , and × e βk k k k ⎠ E1 ⎝
4σ2 m=1 k 2 2
βk αk βk − γk 4

     
√ α2k βk2 − γk4 α2k βk2 − γk4
P1 P2  1
M 2 2
≤ 2 2 log 1+αk log 1+βk ,
γk2 e−jφγ,k = (Xkm −Xkm )(Xkm−BD −Xkm−BD )∗, αk βk − γk4 α2k βk2 α2k βk2
4σ22 (22)
m=BD+1

we can write where E1 (a) denotes the)exponential integral function which

∞ −x
d2k GH(k)(X k − X k )(X k − X k )H GH H (k) is given as E1 (a) = a e x dx and the last inequality
= follows sine e E1 (a) ≤ log(1 + a1 ). Invoking the result
a
4σ22 4σ22 M
|GH1 [k, k]|2 |GH2 [k, k]|2 of (22) in (20), and defining s2k = m m 2
m=1 |Xk − Xk | 
= α2k + βk2 2

 M m m−BD ∗ 
and fk =  m=BD+1 (Xk − Xk )(Xk
m m−BD
− Xk ) 
 P1 P2  yields
|GH1 [k, k]| −jφ1,k |GH2 [k, k]| jφ2,k 2 −jφγ,k
+2Re √ e √ e γk e .  
P1 P2 P X k → X k
     
We also have |GHi [k, k]| = Pi |Hi,k ||Gi,k |,√where |Hi,k | ∼ 8σ24 P1 4 4 P2 4 4
√ ≤ log 1+ 2 sk −fk log 1+ 2 sk −fk ,
Rayleigh( 22 ) and |Gi,k | ∼ Rayleigh( 22 ), i.e., |Hi,k | P1 P2 (s4k −fk4 ) 4σ2 4σ2
(23)
and |Gi,k | are Rayleigh distributed random variables, and
φi,k are uniformly distributed random variables over [0, 2π]. under the assumption that the channels from the source to
If we define ai = |Hi,k | and bi = |Gi,k |, then the relays have higher SNR ratios than the channels from the
d2k relays to the destination.
4σ22
= α2k a21 b21 + βk2 a22 b22 + 2a1 a2 b1 b2 γk2 cos(φ), where φ =
−j(φ1,k −φ2,k +φγ,k )
∠e is uniformly distributed over [0, 2π]. A PPENDIX B
Therefore, we have PEP A NALYSIS FOR B LOCK FADING
" # F REQUENCY-S ELECTIVE C HANNEL
d2
− k2
' 2 2 2 2 2 2 2
(
Edk e 4σ2 =Eai ,bi ,φ e−(αk a1 b1+βk a2 b2+2a1 a2 b1 b2 γk cos(φ)) For the Viterbi detector, we have
+BD 
M 2
  2π   m 
1 X k = arg min Yk − GH1,k
m m−BD 
Xk − GH2,k Xk
m m
= Ea1 ,a2 ,b1 ,b2
2 2 2 2 2 2 2
e−(αk a1 b1 +βk a2 b2 +2a1 a2 b1 b2 γk cos(φ))dφ   ,
Xk
2π 0 m=1
' 2 2 2 2 2 2
( where Xkm = 0 for m < 1 or m > M , and GH1,k
m
=
= Ea1 ,a2 ,b1 ,b2 e−(αk a1 b1 +βk a2 b2 ) I0 (2a1 a2 b1 b2 γk2 )
a
k
m −2πj N Ts dr
  ∞  Gm m
1,k H1,k and
m
GH2,k = Gm 2,k H2,k e . Therefore,
similar to (19) under the assumption that σ12 are sufficiently
2 2 2 2 2 2 2
=Ea2 ,b1 ,b2 e−βk a2 b2 e−αk a1 b1I0(2a1 a2 b1 b2 γk2 )2a1e−a1da1 1,i
0 Pi
larger than σ22
, we can write
where the equality a holds due to the definition of the modified 
Bessel function of the first kind I0 (.). Following the result d2 X k ,X k
1 −
P (X k → X k |H i , Gi , i ∈ {1, 2}) ≤
2
4σ2
e ,
2
RAHMATI and DUMAN: ACHIEVING DELAY DIVERSITY IN ASYNCHRONOUS UNDERWATER ACOUSTIC (UWA) COOPERATIVE COMMUNICATION . . . 1377

M+BD m m m
where d2 (X k , X k ) = m=1 |G1,k H1,k dk + Since dk (m)dH
k (m) is a rank one matrix whenever Xk
=
m
m m−BD 2 m  m−BD
Gm H d
2,k 2,k k | and by defining W t,f (m, k) = Xk and/or Xk m−BD

= Xk and a zero matrix other-

diag{W t (m)w f (k), W t (m)w f (k)}, dmk = X m
k − X k
m
and wise, we can write
dk (m) = [dm m−BD T
k , dk ] , we can write rk = rank(DA (X k , X k )) ≤ min(D(X k , X k ), 2LL2t ),
M
+BD
d2 (X k , X k )= w Tf (k)W Tt (m)[v 1 q T1 , v 2 q T2 ]W t,f (m, k)dk (m) and as a result r = min rk ≤ min(Def f (k), 2LL2t )
m=1
X k ,X k
∗ ∗ with Def f (k) = min  D(X k , X k ). Conditioned on known
× dH
k (m)Wt,f (m, k)[v 1 q 1 , v 2 q 2 ] W t (m)wf (k).
H T T H
X k ,X k
On the other hand, we have channel coefficients, it follows from (27) that
  2
 H 
wTf (k)W Tt (m)[v 1 q T1 , v 2 q T2 ] = 1 − 4σ12 rc=1
k λ
k,c U k,c q (k)
  P (X k → X k |H i [k], Gi [k]) ≤ e 2 ,
2
wTf (k)W Tt (m) [αT1 (1), . . . , αT1 (L)]Tq T1 ,[αT2 (1), . . . , αT2 (L)]Tq T2
⎡ ⎤ where U k,c denotes the c-th column of U k . Furthermore, if we
w Tt (m)α1 (1)q T1 , wTt (m)α2 (1)q T2 L −j 2πkτl
⎢ .. .. .. ⎥ define μi (k) = [μk,i,1 , . . . , μk,i,Lt ] = l=1 e
N Ts α (l),
i
= wTf (k) ⎣ . . . ⎦ since αi,l [n] are i.i.d. complex Guassian random variables with
wt (m)α1 (L)q 1 , wt (m)α2 (L)q 2
T T T T
zero mean, μk,i,p are also i.i.d. complex Gaussian random
  L σh2 i ,l

L
2πkτ
−j N T l
L
2πkτ
−j N T l variables with zero mean and variance of σμ,i 2
= l=1 L .
= q1 T
e s α (l)w t (m), q
T
1
T
2 e s α (l)w t (m) .
T
2 t

l=1 l=1
Furthermore, by denoting q 1 as q 1 = [q1 , . . . , qLLt ]H and q 2
as q 2 = [qLLt +1 , . . . , q2LLt ]H , we can write
By defining
L  q(k) = [q1 μk,1,1 , . . . , q1 μk,1,Lt , . . . , qLLt μk,1,1 , . . . ,
 2πkτ
−j N T l

L 2πkτ
−j N T l
Ai (k)=diag e s αT
i (l), . . . , e s αT
i (l) , . . . , qLLt μk,1,Lt , qLLt +1 μk,2,1 , . . . , qLLt +1 μk,2,Lt ,
l=1 l=1 LLt ×LL2
. . . , q2LLt μk,2,1 , . . . , q2LLt μk,2,Lt ].
t
(29)
and W α,t (m) = diag{wt (m), . . . , w t (m)}LL2t ×LLt , we ob-
tain Therefore, U H
k,c q(k) can be written as

wTf (k)W Tt (m)[v 1 q T1 , v 2 q T2 ] 

LLt 
Lt
  ∗
k,c q(k) =
UH Uk,c,(p−1)L q μ
t +t p k,1,t
= q T1 A1 (k)W α,t (m), q T2 A2 (k)W α,t (m) . (24) p=1 t=1


L
2πkτl t 
LL Lt
Note that Ai (k)W α,t (m) = e−j N Ts αT1,1 (l)w t (m)I Lt L . + ∗
Uk,c,LL 2 +(p−1)L +t qLLt +p μk,2,t
t t
l=1
H  p=1 t=1
Furthermore, by defining q(k) = q T1 A1 (k), q T2 A2 (k) , 2 
 LLt
and W A,t (m) = diag{W α,t (m), W α,t (m)}, we can write = q(i−1)LLt +p χk,(i−1)LLt +p , (30)
i=1 p=1
wTf (k)W Tt (m)[v 1 q T1 , v 2 q T2 ] = q(k)H W A,t (m). (25)

Lt
∗
Therefore, defining where χk,(i−1)LLt +p = Uk,c,(i−1)LL 2 +(p−1)L +t μk,i,t
t
t
D A (X k , X k ) are complex Gaussian
t=1
random variables with
M
+BD 2
zero mean and variance of σχ,k,(i−1)LL t +p
=
= W A,t (m)W t,f (m, k)dk (m)dH
k (m)W t,f (m, k)W A,t (m),
H H
L t

|Uk,c,(i−1)LL2t +(p−1)Lt +t |2 σμ,i

2
m=1
. To calculate the PEP
yields to t=1
of the system, we make two simplifying assumptions:
2
(X k , X k ) D A (X k , X k )q(k).
H
d = q(k) (26)
Since D A (X k , X k ) is a positive semi-definite matrix, we can • UH H
k,i q(k) and U k,j q(k) are independent for different
write values of i and j (i
= j). (Note that for q k being
D A (X k , X k ) = U k Λk U H
k , (27) a complex Gaussian random vector with zero mean,
UH H
k,c q k and U k,j q k are independent complex Gaussian
where U k is a unitary matrix and Λ = random variables.)
diag{λk,1 , . . . , λk,r , 0, . . . , 0} with λk,i being the positive • χk,p are i.i.d. complex Gaussian random variables with
eigenvalues of DA (X k , X k ). We define D(X k , X k ) as 2
zero mean and variance of σχ,k,p .
the number of values of m where either Xkm
= Xk or
m

 m−BD  2LL 
Xkm−BD

= Xk , i.e., l=p+1 |ql χk,l | cos(φl )
t
For a fixed c, by defining θp = arccos Rp
D(X k ,X k ) = M + BD 
 2LL


  t 
and Rp =  ql χk,l  in which φl is the angle between the
M+BD
δ(Xkm −Xk )δ(Xkm−BD −Xk
m m−BD
− ). (28) l=p+1 
m=1
complex valued random vectors ql and χk,l and uniformly
distributed over [0, 2π], we can write
1378 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 13, NO. 3, MARCH 2014

* 2
σχ,k,p 2
σq,p
where πp = l∈S0 ,l=p σχ,k,p
2 2 −σ 2
σq,p 2 . Therefore, we can
χ,k,l σq,l
2LL 2
 t  write
2    
|U k,c q(k)| = 
H
qp χk,p 
  2σ22
p=1
⎧ ⎫ E|q1 |,...,|q2LLt | "2LLt
2LL ⎨2LL ⎬ 2σ22 + λk,c 2
σχ,k,p |qp |2
t 2
t 2LL
t !
p=1
= |qp χk,p | + 2Re qp χk,p ql∗ χ∗l 2σ2 2
⎩ ⎭ = EV
p=1 p=1 l=p+1 2σ22 + λk,c V
 # ∞
2LL
t 2LL
t πp 2σ22 − 2 v 2
2σ
=
σ
= |qp χk,p |2 + 2 |qp χk,p |Rp cos(φp − θp ). 2σ 2
σ 2 2σ 2
+ λ
e χ,k,p q,p

χ,k,p q,p 0 2 k,c v

p=1 p=1 p∈S 0

J  # ∞
1 2σ22 v Nj −1 − σχ,k,p
v
Due to the fact that φp are uniformly distributed over [0, 2π], + e
2 2
σq,p
.
|χp | are Rayleigh distributed and by following the same j=1 p∈Sj
2
(Nj −1)!(σχ,k,p 2 )Nj
σq,p 0 2σ22 + λk,c v
procedure as in (21), we can write
" λ R2 # By using the integral calculated in [19, p. 325. Eq. 2. 3. 6. 14.]
− k,c 0
4σ22
and due to
)∞ the definition of the exponential integral function
−z
Eφ1 ,|χk,1 | e E1 (a) = a e z dz, we obtain
 
 λk,c 2LL t
p=1 |qp χk,p |
2 λk,c
2LLt
p=2 |qp χk,p |Rp cos(φp −θp ) 2σ22
− 2 − 2 P EPk,c = EV
=E|χk,1 | e 4σ2
e 2σ2
2σ22 + λk,c V
  2 % &
λk,c |q1 χk,1 |R1  πp
2σ2
2σ22 λk,c σχ,k,p
2 2 2σ22
× I0 = 2 2 λ
e σq,p
E1 2 2
2σ22 p∈S0
σχ,k,p σq,p k,c λk,c σχ,k,p σq,p
λk,c R2
− 1
λ2 σ 2 |q |2 R2
k,c χ,k,1 1 
J 
(2σ22 )Nj (−1)Nj −1
2σ22 e 4σ2 1
2
= e 4σ2 (λk,c σ2
2
|q |2 +2σ2 )
χ,k,1 1 2 + 2 2 )Nj
×
2
λk,c σχ,k,1 |q1 |2 + 2σ22 j=1 p∈Sj
(Nj − 1)!(λk,c σχ,k,p σq,p
R2  2
2σ2
% &
2σ22
λ
2σ22
k,c 1
−
2λk,c σ2 2
|q |2 +4σ2 λk,c σ2 2
= 2 e χ,k,1 1 . (31) × e χ,k,p
σq,p
E1
λk,c σχ,k,1 |q1 |2 + 2σ22 2
λk,c σχ,k,p 2
σq,p
Nj −1 % & k

Therefore, by taking the expected value with respect to  2
−λk,c σχ,k,p 2
σq,p
φ1 , |χk,1 |, . . . , φ2LLt , |χk,2LLt |, we arrive at + (k − 1)! (33)
2σ22
λ R2 ! k=1
− k,c2 0
4σ
Eφ1 ,|χk,1 |,...,φ2LLt ,|χk,2LLt | e 2 Finally, for the PEP of the system under block fading channel
conditions, we obtain
2σ22 2σ22 + λk,c σχ,k,1
2
|q1 |2 "  2 #
=  rk  H 
2σ22 2 2 2 2
+ λk,c σχ,k,1 |q1 | 2σ2 + λk,c (σχ,k,1 |q1 | + σχ,k,2
2 2
|q2 |2 ) 1 − 1
λ U q (k) 
P (X k → X k ) ≤ E e 4σ2
2 c=1 k,c k,c
"
2σ22 + λk,c 2LL p=1
t −1 2
σχ,k,p |qp |2 2
··· "
2σ22 + λk,c 2LL 1$
2 rk
p=1 σχ,k,p |qp |
t 2
≤ P EPk,c , (34)
2σ22 2 c=1
= "2LLt (32)
2σ22 + λk,c p=1 2
σχ,k,p |qp |2
where P EPk,c are given in (33).
To obtain the above expected value, we first define V =
2LLt 2 2 σq,p
p=1 σχ,k,p |qp | in which |qp | ∼ Rayliegh( 2 ) and |qp |
√
are independent for all p, then obtain the expected value R EFERENCES
  the new defined random variable.
over  Let us define S0 =
 2 [1] M. Vajapeyam, S. Vedantam, U. Mitra, J. C. Preisig, and M. Stojanovic,
pσχ,k,p σq,p
= σχ,k,l σq,l ∀l
= p , i.e., there are |S0 | differ-
2 2 2 “Distributed space-time cooperative schemes for underwater acoustic
communications,” IEEE J. Oceanic Eng., vol. 33, no. 4, pp. 489–501,
2 2 2 2 2 2
ent values of σχ,k,p σq,p such that σχ,k,p σq,p
= σχ,k,l σq,l and Oct. 2008.
∀l
= p. Furthermore, assume that there are J distinct values [2] Z. Han, Y. L. Sun, and H. Shi, “Cooperative transmission for underwater
acoustic communications,” in Proc. 2008 IEEE Int. Conf. Commun., pp.
a2j for j ∈ {1, . . . , J} for which there are p and l (p
= l and 2028–2032.
2 2 2 2 2
p, l ∈ {1, . . . , 2LLt })such
 that σχ,k,p σq,p  = σχ,k,l σq,l = aj . [3] K. Tu, T. M. Duman, J. G. Proakis, and M. Stojanovic, “Cooperative
 2 MIMO-OFDM communications: receiver design for Doppler-distorted
We also define Sj = pσχ,k,p σq,p 2
= a2j for j ∈ {1, . . . , J} underwater acoustic channels,” in Proc. 2010 Asilomar Conf. Signals,
Syst., Comput., pp. 1335–1339.
and Nj = |Sj |. It follows from [17, p. 876] and the definition [4] M. Stojanovic, “Underwater acoustic communications,” Wiley Ency-
of the Gamma distribution that clopaedia of Electrical and Electronics Engineering, vol. 22, pp. 688–
698, 1999.
 πp − 2 v 2
pV (v) = e σ σ
χ,k,p q,p [5] H. M. Wang and X.-G. Xia, “Asynchronous cooperative communication
2 2
σχ,k,p σq,p systems: a survey on signal designs,” SCIENCE CHINA Inf. Sciences,
p∈S0 vol. 54, no. 8, pp. 1547–1531, 2011.

J 
v Nj −1 − 2 v 2
[6] Y. Li, W. Zhang, and X.-G. Xia, “Distributive high-rate space-frequency
+ 2 2
e σ σ
χ,k,p q,p , codes achieving full cooperative and multipath diversities for asyn-
j=1 p∈Sj
(Nj − 1)!(σχ,k,p σq,p ) j
N chronous cooperative communications,” IEEE Trans. Veh. Technol.,
vol. 58, no. 1, pp. 207–217, 2009.
RAHMATI and DUMAN: ACHIEVING DELAY DIVERSITY IN ASYNCHRONOUS UNDERWATER ACOUSTIC (UWA) COOPERATIVE COMMUNICATION . . .
[7] Z. Li, X.-G. Xia, and M. H. Lee, “A simple orthogonal space-time
coding scheme for asynchronous cooperative systems for frequency
selective fading channels,” IEEE Trans. Commun., vol. 58, no. 8, pp.
2219–2224, 2010.
[8] X. Li, “Space-time coded multi-transmission among distributed trans-
mitters without perfect synchronization,” IEEE Signal Process. Lett.,
vol. 11, no. 12, pp. 948–951, Dec. 2004.
[9] Z. Zhong, S. Zhu, and A. Nallanathan, “Distributed space-time trellis
code for asynchronous cooperative communications under frequency-
selective channels,” IEEE Trans. Wireless Commun., vol. 8, no. 2, pp.
796–805, Feb. 2009.
[10] ——, “Delay tolerant distributed linear convolutional space-time code
with minimum memory length under frequency-selective channels,”
IEEE Trans. Wireless Commun., vol. 8, no. 8, pp. 3944–3949, 2009.
[11] M. O. Damen and A. R. Hammons, “Delay-tolerant distributed-TAST
codes for cooperative diversity,” IEEE Trans. Inf. Theory, vol. 53, no. 10,
pp. 3755–3773, 2007.
[12] M. Nahas, A. Saadani, and G. Rekaya, “Bounded delay-tolerant space
time block codes for asynchronous cooperative networks,” IEEE Trans.
Wireless Commun., vol. 10, no. 10, pp. 3288–3297, 2011.
[13] D. Gore, S. Sandhu, and A. Paulraj, “Delay diversity codes for frequency
selective channels,” in Proc. 2002 IEEE Int. Conf. Commun., vol. 3,
2002, pp. 1949–1953.
[14] B. Lu, X. Wang, and K. Narayanan, “LDPC-based space-time coded
OFDM systems over correlated fading channels: performance analysis
and receiver design,” IEEE Trans. Commun., 2002.
[15] S. G. Wilson, Digital Modulation and Coding. Prentice-Hall, 1996.
[16] W. C. Jakes, Microwave Mobile Communication. Wiley, 1974.
[17] J. G. Proakis and M. Salehi, Digital Communications, 5th ed. McGraw-
Hill, 2007.
[18] B. Y. A. Prudnikov, A. P. and O. I. Marichev, Integrals and Series.
Gordon and Breach Science Publishers, 1986, vol. 2.
[19] ——, Integrals and Series. Gordon and Breach Science Publishers,
1986, vol. 1.
1379
Mojtaba Rahmati (S’11) received his B. S. degree
in electrical engineering in 2007 from University
of Tehran, Iran, his M. S. degree in telecommuni-
cation systems in 2009 from Sharif University of
Technology, Tehran, Iran and his Ph.D. in electrical
engineering in 2013 from Arizona State University,
Tempe, AZ. He is currently a research engineer at
Mobile Solutions Lab, Samsung Information Sys-
tems America. His research interests include wire-
less and mobile communication, digital communi-
cations, digital signal processing and information
theory.
Tolga M. Duman Tolga M. Duman (S’95–M’98–
SM’03–F’11) is a Professor of Electrical and Elec-
tronics Engineering Department of Bilkent Univer-
sity in Turkey, and is on leave from the School of
ECEE at Arizona State University. He received the
B.S. degree from Bilkent University in Turkey in
1993, M.S. and Ph.D. degrees from Northeastern
University, Boston, in 1995 and 1998, respectively,
all in electrical engineering. Prior to joining Bilkent
University in September 2012, he has been with
the Electrical Engineering Department of Arizona
State University first as an Assistant Professor (1998-2004), then as an
Associate Professor (2004-2008), and starting August 2008 as a Professor.
Dr. Duman’s current research interests are in systems, with particular focus
on communication and signal processing, including wireless and mobile
communications, coding/modulation, coding for wireless communications,
data storage systems and underwater acoustic communications.
Dr. Duman is a Fellow of IEEE, a recipient of the National Science Foun-
dation CAREER Award and IEEE Third Millennium medal. He served as an
editor for IEEE T RANSACTIONS ON W IRELESS C OMMUNICATIONS (2003-
08), IEEE T RANSACTIONS ON C OMMUNICATIONS (2007-2012) and IEEE
C OMMUNICATIONS S URVEYS AND T UTORIALS (2002-07). He is currently
the coding and communication theory area editor for IEEE T RANSACTIONS
ON C OMMUNICATIONS (2011-present) and an editor for Elsevier Physical
Communications Journal (2010-present).