1300 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 12, NO. 3, MARCH 2013
Segment Training Based
Individual Channel Estimation in
One-Way Relay Network with Power Allocation
Shun Zhang, Feifei Gao, Changxing Pei, and Xiandeng He
Abstract—In this paper, we design a segment training based
individual channel estimation (STICE) scheme in the classi-
cal three-node amplify-and-forward (AF) one-way relay network
(OWRN). The linear minimum mean-square-error (LMMSE)
channel estimator is used to obtain a good initialization, and
an iterative maximum a posteriori (MAP) channel estimator
is developed to improve the estimation accuracy. We then
investigate the underlying power allocation at the relay node
both to minimize the mean-square-error (MSE) of the individual
channel estimation and to maximize the average effective signal-
to-noise ratio (AESNR) of the data detection. The closed-form
Bayesian Cram
´
er-Rao Bound (CRB) is also derived to evaluate
the proposed algorithm. Finally, numerical results are provided
to corroborate the proposed studies.
Index Terms—Individual channel estimation, one-way relay,
Cram
´
er-Rao Bound, power allocation, maximum a posteriori.
I. INTRODUCTION
T
HE cooperative relaying recently has been considered as
a promising technique to enhance the transmission capac-
ity and to exploit the spatial diversity in wireless communica-
tion systems [1], [2]. Several techniques, such as distributed
space time coding (DSTC) [3], [4], relay beamforming [5], [6],
relay selection [7], [8], incremental relay [9], [10], etc., have
been proposed to capture the advantages of the relay networks.
In all these works, the perfect channel state information (CSI)
was assumed at some or at all network nodes.
For one-way relay network (OWRN), the estimation of
the composite channels (called co-channels here) from source
Manuscript received May 30, 2012; revised October 29, 2012; accepted
November 27, 2012. The associate editor coordinating the review of this
manuscript and approving it for publication was X. Zhu.
This work was supported in part by the National Basic Research Program
of China (973 Program) under Grant 2012CB316102, 2013CB336600, by
the National Natural Science Foundation of China under Grant 61072067,
61201187, by the State Key Laboratory of Integrated Services Networks
under the Grant ISN 1001004, by the National Science and Technology
Major Project of China under Grant 2009ZX03007-003, by the open research
fund of National Mobile Communications Research Laboratory, Southeast
University under Grant 2011D02, and by Tsinghua University Initiative
Scientific Research Program under Grant 2012THZ02157.
S. Zhang, C. X. Pei, and X. D. He are with the State Ke y
Laboratory of Integrated Services Networks, Xidian Uni versity, Xian
710071, China (e-mail: zhangshunsdu@gmail.com, chxpei@xidian.edu.cn,
xdhe@mail.xidian.edu.cn). S. Zhang is also with the Department of Automa-
tion, Tsinghua University, Beijing, China, as a Guest student.
F. Gao (corresponding author) is with the Tsinghua National Laboratory
for Information Science and Technology, Beijing, China, and is with National
Mobile Communications Research Laboratory, Southeast University, Nanjing,
China, and is also with the School of Engineering and Science, Jacobs
Univ ersity, Bremen, Germany (e-mail: feifeigao@ieee.org).
Digital Object Identifier 10.1109/TWC.2013.013013.120768
to relay and then to destination has been well studied. For
example in [11], the authors designed two different channel
estimators, i.e., the cascaded and the disintegrated channel
estimators. In [12], the authors investigated the impact of the
imperfect cascaded co-channel estimation on the symbol error
probability for amplify-and-forward (AF) OWRN. In [13], the
optimal maximum likelihood (ML) and the minimum mean-
square-error (MMSE) co-channel estimators for decode-and-
forward (DF) OWRN were derived, where the total power
constraint for all relays and the individual power constraint
for each relay were taken into consideration. In [14], the
authors presented a complete study on the optimal co-channel
estimation for the DSTC-AF-based OWRN. In [15], two
low-complex set-membership (SM) algorithm b ased channel
estimators, i.e, the SM normalized least mean square (SM-
NLMS) and the SM recursive least square (RLS) (SM-RLS)
estimators, were developed for the AF-based multi-hop multi-
node cooperative wireless senor networks.
However, purely knowing the co-channels is insufficient
to support the optimal system design in certain scenarios.
For example, in the relay beamforming scheme, individual
channels (called in-channels here) from source to relay and
from relay to destination are required to design the relay’s
operation and to let the destination predict the relay’s operation
[5], [6]. Similarly, if the relay performs the subcarrier pairing
[16], [17], the destination also needs the in-channel knowledge
in order to know the pairing strategy used at the relay.
1
Some works about the in-channel estimation for the multi-
input multi-output (MIMO) AF-based OWRN have been done
recently. In [18], the authors designed two in-channel estima-
tors under Bayesian framework, i.e., linear minimum mean-
square-error (LMMSE) in-channel estimator and expectation-
maximization (EM) based maximum a posteriori (MAP) in-
channel estimator. Unfortunately, the above two algorithms
could only be applied for the asymmetric channel scenario:
the source→relay lin k experienced Rician fading while the
relay→ destination link was subject to Raleigh fading. Mor e-
over, the scalar ambiguity cannot be eliminated for the above
estimators. In [19], the authors proposed a segment training
based in-channel estimation (called STICE here), where the
in-channel knowledge could be obtained at the destination
through two consecutive segments. In the first segment, the
in-channels from the relay to the destination were estimated
1
Feedback the pairing strategy from the relay to source is possible but
drastically increases the amount of overhead.
1536-1276/13$31.00
c
2013 IEEE