Professional Documents
Culture Documents
Joint Channel Estimation and Data Detection For OFDM Based Cooperative System
Joint Channel Estimation and Data Detection For OFDM Based Cooperative System
Telecommunication Systems
Modelling, Analysis, Design and
Management
ISSN 1018-4864
Telecommun Syst
DOI 10.1007/s11235-019-00622-3
1 23
Your article is protected by copyright and
all rights are held exclusively by Springer
Science+Business Media, LLC, part of
Springer Nature. This e-offprint is for personal
use only and shall not be self-archived in
electronic repositories. If you wish to self-
archive your article, please use the accepted
manuscript version for posting on your own
website. You may further deposit the accepted
manuscript version in any repository,
provided it is only made publicly available 12
months after official publication or later and
provided acknowledgement is given to the
original source of publication and a link is
inserted to the published article on Springer's
website. The link must be accompanied by
the following text: "The final publication is
available at link.springer.com”.
1 23
Author's personal copy
Telecommunication Systems
https://doi.org/10.1007/s11235-019-00622-3
Abstract
Orthogonal frequency division multiplexing based cooperative system using Alamouti space–time block coding at relay
node represents an alternative solution to achieve better connectivity and significant enhancement to the data rate in wireless
fading channels. But, these advantages cannot be achieved without an efficient estimation of the channels which becomes
more challenging for cooperative communications. This paper addresses the joint channel estimation and data detection for
cooperative communication systems. Indeed, equispaced pilot symbols are used by maximum likelihood (ML) algorithm
to derive channels estimator, and then equalizers are calculated and applied to improve receiver data detection. The main
contribution of our work is the development of the ML estimator, the corresponding Cramer–Rao lower bound, mean square
error, signal to interference plus noise ratio, outage probability, bit error probability and the use of simulations to demonstrate
the superior performances of the proposed methods.
Keywords Cooperative · OFDM · Estimation · SINR · CRLB · MSE · BEP · ML · Outage probability
123
Author's personal copy
M. Besseghier et al.
Insert Pilots
Mapping
CP
Data S /
The rest of the paper is organized as follows. In the next
IFFT
/ S
P
section, we represent the system model of OFDM based
cooperative system. Joint channel estimation and data detec-
tion algorithms are derived in Sect. 2, while MSE and CRLB
for channels estimator are proposed in Sect. 4. Based on
the approximation of SINR distribution we develop a tight Relay
Decoding
S/CP
Data P /
FFT
matrix with the vector x on its main diagonal, while tr(X) cor- / P
S
responds to the trace of matrix X. For a vector, · denotes Estimation
the Euclidean norm. The (N × N ) null matrix is represented
by 0 N , and IN signify the (N × N ) identity matrix. {·} and (b) Baseband Model of OFDM based cooperative
J{·} indicate the real and imaginary parts and E{·} represents system
the expectation operator.
Fig. 1 a Cooperative system structure. b Baseband model of OFDM
based cooperative system
2 System model
Next, the inverse F F T (I F F T ) is applied to the data
Let us consider an OFDM based cooperative system with symbol vector giving thus the OFDM symbol qi
three nodes source, destination and relay nodes equipped T
qi (0), . . . , qi (N − 1) . To overcome the ISI generated by
with single antenna for each one as shown in Fig. 1a. The time-dispersive channel, each time-domain OFDM symbol
destination node will combine the received signals coming is extended by the so-called cyclic prefix of L cp samples
from source and relay nodes. The N subcarriers formed the
duration. The L cp + N samples are transmitted sequen-
OFDM signal are split in two subsets of Nα and Nβ sub- tially over the channel to relay and destination nodes. At the
carriers attributed to pilot
and data symbols respectively.
For received nodes, the samples of the cyclic prefix after serial to
that, we define α α1 . . . α Nα and β β1 . . . β Nβ as parallel converted are removed, and the remaining N samples
pilot and data indexes with car d{α ∪ β} N . The M-PSK are computed employing an N − F F T , and the recovered
or M-QAM modulations are used as mapping for Nβ data data symbols zi are restored in serial order after processing
symbols, while Nα pilot symbols are considered as non-zero (estimation, Alamouti decoding) and demultiplexed [27].
symbols. Thus, the i-th transmitted OFDM symbol from the source
The serial data stream is mapped to data symbols employ- node is expressed as
ing M-PSK or M-QAM modulation scheme, and the resulting
symbol stream si is demultiplexed with pilot symbols pi into
a vector of N data symbols xi [xi (0), . . . , xi (N − 1)]T . qi FH xi (1)
123
Author's personal copy
Joint channel estimation and data detection for OFDM based cooperative system
where xi Tα pi + Tβ si is (N × 1) vector containing data where v2i and v2i+1 are time-domain complex Gaussian noise
T
symbols vector si si (0), . . . , si Nβ − 1 multiplexed vectors at the destination
node
with zero means and variances
T equal to σv2 σd2 + τσw2 . σd2 is the noise variance at the
with pilot symbols pi pi (0), . . . , pi (Nα − 1) . Tα and
destination node.
Tβ are permutation matrices used to insert the data with pilot
Channel characteristics are assumed to be invariant
symbols in one vector xi , and F is the FFT matrix defined by
over two consecutive OFDM symbols. hsd is L sd × 1
1 vector representing the channel impulse response (CIR)
F √ [f1 , f2 , . . . , f N ]H sd
sd H
N of source–destination link, and 0 f1L . . . f NL
sd
T is the FFT matrix applied to hsd with fnL
with fn 1, e j2π n/N , . . . , e j2π n(N −1)/N . sd T
To prevent inter-symbol interference (ISI) in multipath 1, e j2π n/N , . . . , e j2π n L −1 /N .
channel, the length of CP, L cp , is chosen to be We define the equivalent CFR of source–relay–destina-
equal or supe-
√ ∗
rior to the channel order L, with L max L sr , L sd , L r d , tion (S–R–D) links by h̃sr d τ h̃sr h̃r d 0̄h sr d , and the
where L sr , L sd , L r d denote the length of source–relay, equivalent channel impulse response (CIR) of (S–R–D) links
source–destination and relay–destination links respectively. by
The i-th received OFDM symbol at the relay node, after √
CP removing, is written as follows hsr d τxcorr (hsr , hr d ) (6)
H
∗
where
y2i F 0 X2i −X2i+1 00 hsd
∗
y2i+1 0 FH X2i+1 X2i 0 0̄ hsr d
S2i+t diag(s2i+t ); P2i+t diag(p2i+t ), and t ∈ {1, 2}
v2i
+
v2i+1 (5) FαH F H
Tα ; FβH H
F Tβ
123
Author's personal copy
M. Besseghier et al.
sd sd H
sd
sd H
0α fαL1 . . . fαLNα , 0β fβL1 . . . fβLN After Alamouti decoding of the received signal, we get
β
sr d H sr d
L sr d sr d H ∗
0̄α fα1 . . . fα Nα
L
, 0̄β fβ1 . . . fβLN
L Ỹβ, 2i Ỹβ,
β Yi 2i+1
∗
Ỹβ, 2i+1 −Ỹβ, 2i
Exploiting the orthogonality between Fα and Fβ , i.e.
Fα FβH 02Nα ×2Nβ , and the propriety Fα FαH I2Nα , and
where Ỹ diag ỹ .
by left multiplication of (8) by Fα , we get The transmitted data symbols s2i and s2i+1 can be recov-
ered after zero-forcing (ZF) channel equalization as follow
ỹα,i Pi α h + ṽα,i (9)
where where
ỹβ, 2i S2i −S∗2i+1 0β 0 hsd ṽβ, 2i σv2
+ , e2 (18)
ỹβ, 2i+1 S2i+1 S∗2i 0 0̄β hsr d ṽβ, 2i+1 2E p Nα Nb
123
Author's personal copy
Joint channel estimation and data detection for OFDM based cooperative system
It is important to know the CRLB in order to fairly assess CRLB(θ ) j−1 (θ ) (21)
the estimation performance of the proposed algorithm. The
CRLB describes the best achievable performance for unbi- where j(θ ) is (Q × Q) Fisher’s information matrix (FIM)
ased estimators. It represents the theoretical lower bound of defined by [21]
the MSE of the estimator.
H
We have derived analytical expression of the Cramer–Rao 2 ∂ ∂
lower bound (CRLB) of the channels estimator for OFDM j(θ) 2 ln F(y/θ ) ln F(y/θ ) (22)
σv ∂θ ∂θ
based cooperative system, which serves as a benchmark for
estimation performance.
For Nb transmitted blocks, we collect the received blocks The unknown parameters are computed as fellow
into matrices, and then Eq. (8) becomes
(a) For data computation
y FH H
α Pα + Fβ Sβ h + v (19)
∂ ∂
ln F(y/θ ) −j ln F(y/θ )
∂s R (n) ∂s I (n)
where
FH
β G 1,n β h (23)
T T
y y0T . . . yTNb −1 ; v v0T . . . vTNb −1
∂ ∂
⎡ ⎤ ⎡ ⎤ ln F(y/θ ) −j ∗ ln F(y/θ )
S0 . . . 0 P0 . . . 0 ∂s∗R (n) ∂s I (n)
⎢ .. . . .. ⎥ ⎢ . .. ⎥
S⎣ . . . ⎦; P ⎣ .. . . . . ⎦ FH
β G 2,n β h (24)
0 . . . S Nb −1 0 . . . P Nb −1
⎡ H ⎤ ⎡ ⎤
Fα . . . 0 FβH . . . 0 we define G1, n
Gn 0
, G2, n
0 Gn
, where Gn
⎢ . . . .. ⎥ ⎢ . . . ⎥ Gn 0 0 Gn
FH . . ⎦; F β ⎢ ⎥
H
α ⎣ .. ⎣ .. . . .. ⎦ is a diagonal matrix with the same size as Si that contains
0 . . . FαH 0 . . . FβH one at the n-th position and! zeros for the remain posi-
T T tions, and n ∈ 1 . . . Nβ . For Nb transmitted blocks,
α α, ⎡ ⎤
1 . . . α, Nb ; β β, 1 . . . β, Nb
T T T T
Gt, n . . . 0
⎢ ⎥
G t, n ⎣ ... . . . ... ⎦ is 2Nβ Nb × 2Nβ Nb matrix
where S and P are 2Nβ Nb × 2Nβ Nb and
0 . . . Gt, n
(2Nα Nb × 2Nα Nb ) data and pilot matrices respectively. F H α with t ∈ {1, 2}.
and F Hβ represents the IFFT matrices for data and pilot sym- (b) For channels derivation
sd
α and
β are 2Nα Nb × L + L
bols respectively. sr d
123
Author's personal copy
M. Besseghier et al.
where E s |s2i+t |2 is data symbols power. Based on previously MSE computation, the channels esti-
mation given by Eq. (11) can be decomposed as following
Isd 0 [22]
I
0 Isr d
I L sd jI L sd I L sr d jI L sr d ĥ h + h (30)
Isd ; Isr d
− jI L sd I L sd − jI L sr d I L sr d
where h is modeled as a zero-mean i.i.d complex Gaus-
H̃βsd j H̃βsd 0 0 sian vector with variance equal to e2 [22]. The previous ZF
Hβ
0 0 H̃βsr d j H̃βsr d equalizer coefficients become
with H̃βsd 0β hsd , H̃βsr d 0̄β hsr d . csd h̃ ∗sd + h̃ ∗sd ˆ −1
h , c sr d h̃ sr d + h̃ sr d ˆ −1
h
2 2
0β j0β 0 0 S̄2i −S̄∗2i+1
Eβ and Si ∗ . where ˆ h h̃ sd + h̃ sd + h̃ sr d + h̃ sr d . By dropping
0 0 0̄β j 0̄β S̄2i+1 S̄2i
the subscript k, Eq. (14) for the first and second symbol
S̄2i and S̄2i+1 are defined by becomes
ŝi ˆ −1
h (h si + ψi + ηi ) (31)
S̄2i (S2i + S2i+1 ); S̄2i+1 (S2i+1 − S2i ).
where
After computations, the CRLB for channels estimation is
given by
s2i
• Transmitted and estimated data symbols are si
s2i+1
σv2
CRLB(h) " Nb −1 H H ŝ2i
and ŝi respectively.
4E s N Nb I − i0 Eβ Si Hβ H†β Si Eβ ŝ
2i+1
(28) s2i −s∗2i+1 h̃ sd h̃ ∗sd + h̃ ∗sr d h̃ sr d
• ISI:ψi .
s2i+1 s∗2i h̃ sr d h̃ ∗sd − h̃ ∗sd h̃ sr d
where S i is 4Nβ × 4N β , Hβ is 2Nβ × 4Nβ and Eβ is ṽβ, 2i ṽ∗β, 2i+1 h̃ ∗sd + h̃ ∗sd
• Noise: ηi ∗ .
4Nβ × 2 L sd + L sr d matrices. Pilot and data symbols are ṽβ, 2i+1 −ṽβ, 2i h̃ sr d + h̃ sr d
used to derive the CRLB in (28). Based only on pilot symbols
the CRLB is reduced to The power of the incoming signal is calculated as
σv2 L sd + L sr d E si siH 22h ˆ −2
h Es (32)
CRLB(h) (29)
2E s Nα Nb
The power of the interference (ISI), generated by channels
which can be considered as a lower bound of (28). estimation errors, is computed as
E ψi ψiH ≈ 2e2 h ˆ −2h Es (33)
5 Outage probability and bit error
probability The noise power can be expressed by
We focus in this section on the SINR of OFDM based coop- E ηi ηiH 2−1
h σv
2
(34)
erative ZF receiver in presence of channels estimation errors.
Based on the approximation of SINR distribution, we develop The SINR is now defined by
a tight closed-form approximation for the outage probability
!
and the Bit Error Probability (BEP) of M-PSK and M-QAM E si siH
modulations. γ ! !
E ψi ψiH + E ηi ηiH
5.1 SINR derivation using the approximation ˆ h ≈ h + 2e2 , and after substitu-
tions, the SINR γ becomes
To assess the performance of the receiver, a definition of the
SINR is needed. Indeed, we focus in this subsection to the h E s
γ≈ (35)
derivation of a new expression of the SINR. e2 E s + σv2 + 2σv2 e2 −1
h
123
Author's personal copy
Joint channel estimation and data detection for OFDM based cooperative system
• For M-PSK
By substitution of Eq. (29) into (36), we obtain
2 $
max(1,M/4)
2E p Nα Nb E s h Pe (err /x) ≈
γ≈ (37) max log2 M, 2
σv2 E s + 2E p Nα Nb σv2
i1
% (2i − 1)π
Q 2γs xsin (43)
which can be written as M
x n/2−1 e−x/2σ
2
The BEP in Rayleigh fading channel is obtained by aver-
f (x) , x ≥ 0 and σ2 1/2 (40)
σ2 2n/2 (n/2) aging Eqs. (43) and (44) over the Chi square probability
density function (pdf) of x as follow
5.2 Outage probability #∞
BE P f (x)Pe (err /x)d x (45)
The outage probability represents an important performance 0
measure of communication systems [29]. It is defined as the
probability of SINR γ falls below a pre-defined threshold γx , Which gives the following expressions
and can be expressed as
• For M-PSK
γ#x /γs 2
B E PM−PSK ≈
Poutage (γx ) Pr(0 ≤ γs x ≤ γx ) f (x)d x (41) max log2 M, 2
0
$
max(1,M/4)
1
. 1 − μi 3 − μi2 (46)
For further clarification, Poutage is also defined as a 2
i1
cumulative distribution function (CDF) of the Chi square
)
distributed random variable x evaluated at γx /γs . Using inte- γs sin 2 ((2i−1)π /M)
where μi 1+γ
gration by parts on (41), the Poutage (γx ), for our proposed s sin ((2i−1)π /M)
2
• For M-QAM
cooperative system, can be expressed as √ √
log2 M 1−2−k M−1
2 $ $
γx −γx /γs B E PM−QAM ≈ √ √
Poutage (γx ) 1 − 1 + e (42) Mlog2 M k1
γs *
i0
+
i.2
√
k−1
i.2 k−1 1 1
(−1) M 2k−1 − √ + 1 − μi 3 − μi2
5.3 BEP M 2 2
(47)
Cooperative system uses distributed antennas on different '
relays to achieve close to the theoretical gains of MIMO. 3(2i−1)2 γs
where μi
2(M−1)+3(2i−1)2 γs
This will allow us to use the mathematical foundations of the
123
Author's personal copy
M. Besseghier et al.
10
Table 1 ETU channel fading model
True Channel
Estimated channel
Taps Doppler Excess tap delay Relative power
5
frequency (Hz) (ns) (dB)
1 300 0 − 1.0 0
-25
50 100 150 200 250 300 350 400 450 500
6 Simulation results k (Normalized time)
Simulations are used to compare the performances of the Fig. 2 Comparison between true and estimated channels gain
proposed ML channels estimator and data detector in terms
101
of MSE, CRLB, outage probability and BEP to better under- N =32 Simulation
stand the impact of channels estimation errors on system N =32 Analytic
100
performance and also verify our analytical derivations. The N =32 CRB
comparisons are calculated using 103 Monte Carlo (MC) N =64 Simulation
123
Author's personal copy
Joint channel estimation and data detection for OFDM based cooperative system
10-1 0.06
N =32 Simulation N =1
N =32 Analytic N =4
N =32 CRB 0.05 N = 16
N =64 Simulation N = 32
10-2
N =64 Analytic
N =64 CRB 0.04
Min CRB
Poutage
MSE
10-3 0.03
0.02
-4
10
0.01
10-5 0
10 20 30 40 50 60 70 80 90 100 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Nb x
Fig. 4 MSE of channels estimator versus the number of OFDM Blocks Fig. 6 Outage probability versus γx for different values of Nα
Nb
100
1
0.9 10-1
SNR [dB] = 0
SNR [dB] = 7
0.8 SNR [dB] = 14
SNR [dB] = 21 10-2
0.7
BEP M-PSK
10-3
0.6
Poutage
M=2 Simulation
0.5 10-4 Analytic
Exact CIR
0.4 M=4 Simulation
10-5 Analytic
Exact CIR
0.3
M=8 Simulation
Analytic
0.2 10-6 Exact CIR
0.1
10-7
0 5 10 15 20 25 30
0 Average SNR per bit [dB]
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
x
Fig. 7 B E P versus SNR for (M − PSK) modulations
Fig. 5 Outage probability versus γx for different values of average SNR
per bit
BEPs in LTE channel (ETU) versus different value of pilot
symbols Nα at fixed S N R 14 dB is illustrated in Fig. 9.
channel using (42). Here, the average SNR per bit is fixed at We remark an improvement in performances as the number
14 dB and we plot the outage probability for different values of pilot symbols, Nα , increases. When Nα exceeds the length
of Nα . It is seen that as Nα increases, the outage probability of the equivalent CIR of source–relay–destination links L sr d ,
decreases. This is due to the variance of estimation errors e2 the performances become closer to the CSI case.
which becomes large for a small value of Nα . For example,
for the normalized γx equals to 1.2 the outage probability
is about 20% when Nα 1, while for Nα 32 the outage 7 Conclusion
probability is less than 10.
Figures 7 and 8 illustrate BEPs versus SNR for M-PSK In this paper, we have proposed joint channels estimation
with M 2, 4, 8 and for M-QAM with M 4, 16, 64 respec- and data detection algorithms for OFDM based cooperative
tively. As expected, the analytic and numerical BEPs are the system using Alamouti STBC at the relay node. Based on
same and are closer to the channel state information (CSI) equispaced pilots, we develop ML channels estimator, and
case. then ZF equalizers are computed and applied to ameliorate
123
Author's personal copy
M. Besseghier et al.
123
Author's personal copy
Joint channel estimation and data detection for OFDM based cooperative system
20. Zhang, Z., Zhang, W., & Tellambura, C. (2009). Cooperative Mokhtar Besseghier was born
OFDM channel estimation in the presence of frequency offsets. in 1987 in Mascara, Algeria.
IEEE Transactions on Vehicular Technology, 58(7), 3447–3459. He received the Master degree
21. Tao, J., Wu, J., & Xiao, C. (2008). Channel estimation for OFDM in Telecommunications from
systems in the presence of carrier frequency offset and phase noise. MASCARA University in 2012,
In IEEE ICC communications. and the Ph.D. degree from Sidi
22. Jiang, F., Li, C., & Gong, Z. (2018). Accurate analytical BER per- Bel Abbes University, Algeria, in
formance for ZF receivers under imperfect channel in low SNR 2017. Currently, he was an assis-
region for large receiving antennas. IEEE Signal Processing Let- tant professor at the University
ters, 25(8), 1246–1250. of Mascara in the department of
23. Jacobs, L., & Moeneclaey, M. (2009). Effect of MMSE channel Electronics, and does his research
estimation on BER performance of orthogonal space–time block at the Telecommunication and
codes in rayleigh fading channels. IEEE Transactions on Commu- Digital Signal Processing Labora-
nications, 57(5), 1242–1245. tory in Sidi Bel Abbes University.
24. Ju, M. C., Song, H. K., & Kim, I. M. (2009). Exact BER analysis His current research interests
of distributed Alamouti’s code for cooperative diversity networks. include wireless communications and signal processing.
IEEE Transactions on Communications, 57(8), 2380–2390.
25. Wang, C., Au, E. K. S., Murch, R. D., Mow, W. H., Cheng, R.
S., & Lau, V. (2007). On the performance of the MIMO zero- Ahmed Bouzidi Djebbar received
forcing receiver in the presence of channel estimation error. IEEE engineering degree, Master
Transactions on Wireless Communications, 6(3), 805–810. degree and Ph.D. degree with
26. Han, S., Ahn, S., Oh, E., & Hong, D. (2009). Effect of channel honor from Electronic department
estimation error on BER performance in cooperative transmission. of Sidi Bel Abbes University,
IEEE Transactions on Vehicular Technology, 58(4), 2083–2088. Algeria in 1995, 1999 and 2006
27. Weinstein, S. B., & Ebert, P. M. (1971). Data transmission by respectively. He joined Sidi Bel
frequency division multiplexing using the discrete Fourier trans- Abbes University as assistant
form. IEEE Transactions on Communication Technology, COM-19, professor in 2000. From 2005
628–634. to 2006, he was scholarship at
28. Gore, D. A., Heath, R. W., & Paulraj, A. J. (2002). Transmit selec- OAE Department of University
tion in spatial multiplexing systems. IEEE Communication Letters, of Valenciennes, France and at
6(11), 491–493. the Signal and Image Processing
29. Li, Y., Cimini, Jr, L. J., & Himayat, N. (December 2008). Perfor- Department of ENST of Paris,
mance analysis of space–time block coding with co-channel MIMO France. His research interests are
interferers. In IEEE “GLOBECOM”. in lower complexity Blind and Semi-Blind algorithms for MIMO
30. Lu, J., Letaief, K. B., Chuang, J. C.-I., & Liou, M. L. (1999). M- multicarrier systems, performance analysis and multiuser detection.
PSK and M-QAM BER computation using signal-space concepts.
IEEE Transactions on Communications, 47(2), 181–184. Abdelhak Zouggaret was born
31. Liu, X., & Su, W. (2007). BER performance analysis of the opti- in Mascara, Algeria, in 1978.
mum ML receiver for decode-and-forward cooperative protocol. He received engineering degree,
In IEEE ICASSP communications. Master degree and Ph.D. degree
32. Kaviani, S., & Tellambura, C. (2006). Closed-form BER analysis in telecommunications and elec-
for antenna selection using orthogonal space–time block codes. tronics from Sidi Bel Abbes
IEEE Communications Letters, 10, 704–706. University, Algeria in 2001, 2004
33. Sriharsha, M. R., Dama, S., & Kuchi, K. (2017). A complete cell and 2016 respectively. In 2018
search and synchronization in LTE. EURASIP Journal on Wireless he joined Mascara University as
Communications and Networking, 2017, 101. an assistant professor of com-
34. Evolved Universal Terrestrial Radio Access (E-UTRA); Base Sta- munications engineering and
tion (BS) radio transmission and reception (3GPP TS 36.104 information theory. His current
version 11.2.0 Release 11). www.3GPP.org. research interests include wireless
communications signal and image
processing.
Publisher’s Note Springer Nature remains neutral with regard to juris-
dictional claims in published maps and institutional affiliations.
123
Author's personal copy
M. Besseghier et al.
123