You are on page 1of 6

18th international conference on Sciences and Techniques of Automatic control & STA'2017-PID4583-SIP

computer engineering - STA'2017, Monastir, Tunisia, December 21-23, 2017

Blind equalization of OFDM systems based an Optimal Bounding


Ellipsoid under different channels

A. Moussa∗† , M. Pouliquen∗ , M. Frikel∗ , S. Bedoui† , K. Abderrahim† , M. M’Saad∗


∗ Laboratoire d’Automatique de Caen - EA 7478

Normandie Univ, UNICAEN, ENSICAEN, LAC


14050 Caen Cedex, France
† Research Unit: Numerical Control of Industrial Processes National Engineering school of Gabes, Gabes
University
6029 Gabes, Tunisia
ali.moussa@ensicaen.fr

Abstract— In this paper, we propose a frequency recursive • The nature of OFDM structure: ZP-OFDM (Zero
blind equalization algorithm of data in orthogonal frequency Padding-OFDM)([8], [5]), CP-OFDM ([9], [12], [3],
division multiplexing (OFDM) system using a cyclic prefix. The [2], [24]) and GI-less OFDM (Guard Interval-less
proposed algorithm is based on a Set Membership identification
methods. More precisely, it is based an Optimal Bounding OFDM) ([6], [7]).
Ellipsoid (OBE) algorithm. In this document, we consider the • The structure of the channel and of the equalizer: SISO,
presence of bounded noise, of a single Input Single Output MIMO, SIMO, FIR, IIR.
(SISO) fast fading channels such as Rayleigh and Rice channels. What combines between different methods is the use of
The numerical simulation shows that the proposed algorithm
has good performance under different channels.
Gaussian noise.
Index Terms— Recursive blind equalization, OFDM, fast Some of these algorithms perform channel estimation to
fading channel. equalize, this strategy complicates the application of these
approaches in real time. Other algorithms require a large
I. I NTRODUCTION number of OFDM symbols to ensure rapid convergence.
So as to pass the limits of the previous methods, we pro-
OFDM is among the most interesting multicarrier mod- pose in this paper a frequency algorithm of blind equalization
ulation techniques in the wireless communication field for in the presence of bounded noise and for QAM modulation.
high data rate systems with limited bandwidth. Thanks to its In fact, the proposed approach is based on OBE algorithms
robustness and its resistance to impulsive noise and to the ([13], [21]) and it is the extension of the SBME (extension
phenomenon of multipath fading, OFDM has been chosen the Set Membership Blind Equalization) algorithm that has
in several applications, the most important: the digital video presented in [20], hence its name is eSBME. This method
broadcasting (DVB-T), the video system (DVB) and the has been described and studied for single-carrier modulation
digital terrestrial audio system (DAB) ([22], [10]). On the in [15], [17] and [16]. In this paper, we evaluate the proposed
other hand, this type of multicarrier modulation has been algorithm for multi-path channels, specifically Rayleigh and
implemented in different standards such as IEEE 802.11, Rice channels, on the other hand, we test this for a popular
IEEE 802.16 and IEEE 802.20. In order to perform consistent channel, which is AWGN channel.
data recovery and channel estimation, these standards require The remaining of this document is structured as follows:
training symbols and the insertion of pilot supports, this one section II presents system model, description of OFDM
reduces the bandwidth. To overcome this problem, another system and channel model. The proposed blind equalization
technique has been added in the communication chain called algorithm is provided in section III. Some simulation results
blind equalization. These algorithms have received a lot are reported in section IV and we draw our conclusion in
of research and developments due to their efficiency in section V.
retrieving data transmitted in high-bit-rate. In this context,
several approaches of blind equalization for OFDM system II. P ROBLEM STATEMENT
have been presented in the literature. A. Description of OFDM system
The main differences between the various approaches lie The baseband digital communication SISO system is
in the following aspects: shown in Fig. 1. This system consists of two main elements:
• The nature of equalization application field: frequency CP-OFDM system and equalizer. The input data {ak } is
domain ([11], [24], [1], [23]), time domain ([9], [12], modulated by single-carrier modulation to give symbols {sk }.
[3], [2], [8], [4]), frequency and time domain ([14], [18], The input symbol sequence {sk } is grouped in blocks {Sk },
[19]). each with N symbols, which can be written as

978-1-5386-1084-8/17/$31.00 ©2017 IEEE 599


XpCP
Temporal OFDM symbol
manner
xN −D,k
XkCP = TCP X p
Sp Xp xN −1,p
x0,p = TCP FNH S p
s0,p x0,p Add Parallel = [xN−D,p , ..., xN−1,p , x0,p , ..., xN−D,p , ..., xN−1,p ]T
Serial To
T
= xCP 0,p , ..., xD−1,p , xD,p , ..., xN,p , ..., xN+D−1,p
Cyclic CP CP CP CP
ak To Serial
Modulation IFFT
Parallel Prefix Converter (4)
xN −D,p
Converter
The previous signal data (X pCP ) are transmitted by a
xN −1,p channel (described in the next section).
sN −1,p xN −1,p
The cyclic-prefixed data sequence at the output of the chan-
C
h nel is defined by
a T
n YpCP = yCP CP CP CP CP
0,p , ..., yD−1,p , yD,p , ..., yN,p , ..., yN+D−1,p (5)
n
e
l Thereafter, the CP has been eliminated to give OFDM
symbols without a CP which can be written as
Frequency OFDM symbol bk
CP
y0,p Yp = [y0,p , y1,p , ..., yN−1,p ]T (6)
CP
yD−1,p The symbol block Yp is then demodulated and converted into
E frequency domain by FFT matrix FN , finally giving symbols
s0,p q z0,p y0,p Serial
Serial u Remove
prepared for the blind equalization process which can be
To
s0,p s1,p , ... To a
Cyclic
structured in following manner
l FFT Parallel
Parallel i Prefix Converter
Converter z Z p = FN Yp = [z0,p , z1,p , ..., zN−1,p ]T (7)
e CP
r yN +D−1,p
sN −1,p zN −1,p yN −1,p Before implementing our algorithm for the OFDM system,
Yp YpCP
we assume the following throughout the sequel
Fig. 1. Overall System Model 1) There exists a perfect synchronization of carriers and
symbols;
2) The elements of S p (p = 1, 2, ..., P) is a sequence of
independent identically distributed symbols, that are
assumed to be drawn from a constellation QAM;
Sk = [s0,p , s1,p , ..., sN−1,p ]T ; p = 1, 2..., P (1) 3) The length of the channel gain is shorter or equal than
the length of CP D;
4) The noise sequence is an unknown noise and it is
where (.)T is the transpose operator. assumed to be bounded i.e |bk | ≤ δb , where δb is an
We consider these elements to be in the frequency domain. upper bound;
Then each block may be treated by optional IFFT matrix
  5) The two sequences {sk } and {bk } are independent;
FNH ∈ RN×N , to finally give blocks in the time domain X p , 6) There exists an equalizer parameterized by its impulse
which is represented in the following manner: response coefficients {wi } such that if bk = 0, we get
the equalizer output
L
X p = FNH S p = [x0,p , x1,p , ..., xN−1,p ]T (2)
sk = ∑ wTi βk−i = φkT θ ∗ (8)
i=0
With where θ ∗ ∈ Cn×1 is the parameter vector with n = L + 1
⎡ ⎤ the number of parameters:
1 1 ··· 1 ⎛ ⎞ ⎛ ⎞
⎢ 1 e− j2π /N ··· e− j2π (N−1)/N ⎥ w0 βk
⎢ ⎥ ⎜ ⎟ ⎜ ⎟
FN = ⎢

.. .. ..
.
.. ⎥

(3) θ ∗ = ⎝ ... ⎠ and φk = ⎝ ... ⎠
. . .
1 e− j2π (N−1)/N ··· e− j2π (N−1)
2 /N wL βk−L
And we define βk (k = 0, 1, ..., (N − 1) ∗ P) in the
Next, D CP samples are inserted to each OFDM symbol. following manner:
CP
T insertion
can be represented by using a matrix TCP =
ICP , INT ∈ R(N+D)×N , which is a concatenation of the last [β0 , ..., β(N−1)∗P ] = [Z1T , Z2T , Z3T , ..., ZP−1
T , ZT ]
P
D rows of a identity matrix IN (that we denote as ICP ), and = [z0,1 , ..., zN−1,1 , ......, z0,P , ..., zN−1,P ]
the identity IN itself. This process is defined in the following (9)

600
In this paper we use the quantization function Q(.) deter- For the Rice Channel, the magnitude of the received
mined by   complex envelope rk is distributed by Rician distribution as
arg min   follows:
Q(y) =  z−y  (10) ⎧
z ∈ QAM ⎪ −r2 √
⎨ re(−K2 r ) e( 2σ 2 ) I0 ( r 22k ); r  0
σ σ
Q(y) corresponds to the most closer symbol in QAM con- p(r) = (15)


stellation. 0; r < 0
Los power signal s 2
B. Channel structure where Kr = Nlos power signals = 2σ 2 , Kr is the Rice factor,
s2 = m2I + m2Q is a non-centrality parameter and I0 (.) is the
The channel is the most interesting element in the com-
modified Bessel function of the 0 order.
munication chain because it is responsible for the transfer of
In the two preceding distributions, each path is suffering from
data from the transmitter to the receiver. One phenomenon
a Doppler Effects phenomenon, this phenomenon causes a
is widespread and dominant in wireless communication sys-
frequency shift which noted by fd .
tems that is fast fading. This occurs if the impulse response
The implementing of these two channels requires fixing two
of the channel changes rapidly in the duration of the symbol,
parameters: a carrier frequency fd and a simple period of the
it results in selective time loss due to Doppler spreading.
input data Ts .
There are two models that characterize fast fading phe-
nomenon: Rayleigth and Rice channels. III. E QUALIZATION ALGORITHMS
In this context, we use these channels in this paper. The In this part, we present how to design the equalizer that
Rayleigh model is appropriate in the position when there is is to say how to realize the estimation of the input sequence
no line-of-sight (NLOS) between the transmitter and receiver. {sk } knowing that the estimation of sk is denoted by sk . The
However, the Rice structure is convenient in the situation main purpose of the proposed algorithm is to decrease the
when there is line-of-sight (LOS). phase distortion in a channel, thus minimizing the channel
After passing by the Rayleigh and Rician channels which effect to improve transmission performance. The proposed
contain M paths, the received data is structured in the process is done by estimating the parameter vector θ ∗ .
following manner: Let consider vkf to be the effect of bk after the FFT operation
M
and to the equalizer input. vkf can be defined as follows
yk = ∑ αkj ηk−dkj + bk (11) N−1 −2π ∗k∗i
j=1 vkf = − ∑ bi e N (16)
i=0
where αkj
and dkj
are the gain factors and delay of specific
path j. We define the different elements ηk (k = 0, 1, ..., (N + From assumption 2), vkf is bounded and an upper bound is
D − 1) ∗ P) in the following manner: N−1
|vkf | ≤ δvf = ∑ δb (17)
i=0
[η0 , ..., η(N+D−1)∗P ] = [(X1CP )T , (X2CP )T , ..., (XPCP )T ]
We define hereafter vk is the impact of vkf at the output of
0,1 , ..., xN+D−1,1 , ......, x0,P , ...,
= [xCP CP CP
the equalizer, vk is given by
xCP
N+D−1,P ]
L
(12)
vk = − ∑ wTi vk−i
f
(18)
(11) can be written as i=0
From equation (17), vk is bounded and upper bounded is
yk = cos(2π fc )Ik − sin(2π fc )Qk + bk (13)
L

Where Ik and Qk are uncorrelated zero mean Gaussian |vk | ≤ δv = ∑ |wi |δvf
i=0
random variables with non-zero mean mI and mQ , and fc is
the frequency of the sinusoidal carrier signal. The estimation of the data is done in this paper as follows:
if the impact of the noise is bounded and lower than 1 (i.e.
lower than the half distance between two symbols in the
For the Rayleigh Channel, the magnitude of the received
constellation), then φkT θ ∗ is close to a point which belongs
complex envelope rk = |αk | is distributed by Rayleigh distri-
to the constellation. The membership set is defined by a circle
bution given in the following equation:
⎧ of radius δv < 1 around sk , consequently sk can be estimated
−r2

⎨ r22 e( 2σ 2 ) ; r  0 using Q(.). We have
σ  
p(r) = (14) Q φkT θ ∗ = φkT θ ∗ + vk (19)


0; r < 0 So it is clear that we have the following property:
If {bk } and {wi } satisfy δv < 1, then we have
with σ2 is the variance of the two parameters Ik and Qk .  
Q φkT θ ∗ = sk (20)

601
SNR=10 SNR=10

The purpose of the proposed method is to estimate the 2

1.5
2

1.5

parameter vector θ such that sk = φkT θ is in a neighborhood 1 1

(defined by a circle of radius less than δv ) of a point which 0.5 0.5

Imaginary

Imaginary
belongs to the constellation. Such condition is satisfied if θ
0 0

−0.5 −0.5

is computed in such a way that −1 −1

 
−1.5 −1.5

Q φkT θ = φkT θ + εk
−2 −2
(21) −2 −1 0
Real
1 2 −2 −1 0
Real
1 2

where Fig. 2. Simulation example Fig. 3. Simulation example


|εk | ≤ δ (22) 1: Constellation QAM of input 1: constellation QAM of output
equalizer for Rayleigh channel equalizer for Rayleigh channel
and SNR = 10dB and SNR = 10dB
with δv ≤ δ < 1.
SNR=20 SNR=20
2 2

The estimated parameter vector θk is updated at each 1.5 1.5

1 1

instant k in the following manner: 0.5 0.5

Imaginary

Imaginary
0 0

θk = θk−1 + Γk εk/k−1


−0.5 −0.5

(23) −1 −1

−1.5 −1.5

In the previous equation, Γk is the adaptation gain which is −2


−2 −1 0
Real
1 2
−2
−2 −1 0
Real
1 2

defined as follow:
⎧ Pk−1 φk σk
Fig. 4. Simulation example Fig. 5. Simulation example

⎨ Γk = λ +φ T Pk−1 φk σk 1: Constellation QAM of input 1: constellation QAM of output
k equalizer for Rayleigh channel equalizer for Rayleigh channel
(24)

⎩   and SNR = 20dB and SNR = 20dB
Pk = λ In − Γk φk Pk−1
1 T

In this expressions, there are two parameters: the first In the first two cases of
 the sigma expression, we calcu-
parameter is the forgetting factor set by the user to weight lated the error Q sk/k−1 − sk/k . After some mathematical
past data, this parameter is bounded i.e 0 < λ ≤ 1. The development, we obtained the following result
second parameter is σk .    
Q sk/k − sk/k  ≤ δ (26)
Before presenting the definition of σk , it is necessary In the latter case of the sigma expression, it is clear that the
to define respectively the a priori output of the equalizer, adaptation is frozen.
a priori error, a posteriori output of the equalizer and a
posteriori error: From (26), we note that the fundamental feature of the
⎧ proposed approach is: if φkT Pk−1 φk > 0, then the a posterior

⎪ sk/k−1 = φkT θk−1 error εk/k satisfies

⎪  

⎪ εk/k  ≤ δ < 1 (27)


⎨ εk/k−1 = Q( sk/k−1 ) − sk/k−1
With

⎪ δv ≤ δ < 1

⎪ sk/k = φkT θk (28)



⎪ We note that for each value of σk , the estimated parameter

εk/k = Q( sk/k ) − sk/k vector θk satisfies (21) and (22).
Next, we define σk as follows IV. SIMULATIONS RESULTS
⎧  ε   In this part, we study the feature and the performance
⎪ λ  k/k−1 

⎪ φk Pk−1 φk  δ  − 1 of the proposed blind method for CP-OFDM. We propose

T
⎪  ε  

⎪  k/k−1  a SISO OFDM scenario with one transmitted independent

⎪ if  δ >1

⎪ data stream, corresponding of one user. The user is equipped



⎪   with one transmit antenna. It transmits i.i.d. 4-QAM symbols

⎪ and φkT Pk−1 φk > 0

⎪ which are modulated by N = 1024 subcarriers and the length


⎨ of the CP is set to D = 128. The number of OFDM symbols
σk = 1  ε   (25)
⎪  k/k−1  is fixed to P = 3.

⎪ if   ≤ 1

⎪ δ {bk } was a complex white noise uniformly distributed with



⎪   |bk | ≤ δb , where δb has been adjusted to have a desired Signal



⎪ and φkT Pk−1 φk > 0 to Noise Ratio (SNR).



⎪ In order to implement our algorithm, we L = 7, λ equal to



⎪ 0 0.99 and δ = 0.99. Concerning the parameters relating to

else the channels, we have chosen the following values of the

602
Eye Diagram for In−Phase Signal Eye Diagram for In−Phase Signal
2 2
Eye Diagram for In−Phase Signal Eye Diagram for In−Phase Signal
2
2
1 1
1 1

Amplitude

Amplitude
Amplitude

Amplitude
0 0 0
0
−1
−1 −1
−1
−2

−2 −3
−2 −2
−0.5 0 0.5 −0.5 0 0.5 −0.5 0 0.5 −0.5 0 0.5
Time Time Time Time

Eye Diagram for Quadrature Signal Eye Diagram for Quadrature Signal Eye Diagram for Quadrature Signal Eye Diagram for Quadrature Signal
2 2 2
2

1 1 1 1
Amplitude

Amplitude

Amplitude

Amplitude
0
0
0 0
−1
−1
−2 −1 −1

−2 −3
−0.5 0 0.5 −0.5 0 0.5 −2 −2
Time Time −0.5 0 0.5 −0.5 0 0.5
Time Time

Fig. 6. Simulation example Fig. 7. Simulation example 1:


Fig. 13. Simulation example 2: Fig. 14. Simulation example 2:
1: eye pattern of input equalizer eye pattern of output equalizer
eye pattern of output equalizer with eye pattern of output equalizer with
with Rayleigh channel and SNR = with Rayleigh channel and SNR =
AWGN channel and SNR = 30dB Rayleigh channel and SNR = 30dB
10dB 10dB

Eye Diagram for In−Phase Signal


2

Eye Diagram for In−Phase Signal Eye Diagram for In−Phase Signal
2 2 1

Amplitude
1 1
0
Amplitude

Amplitude

0 0
−1

−1 −1
−2
−0.5 0 0.5
−2 −2 Time
−0.5 0 0.5 −0.5 0 0.5
Time Time Eye Diagram for Quadrature Signal
Eye Diagram for Quadrature Signal Eye Diagram for Quadrature Signal
2
2 2

1
1 1

Amplitude
Amplitude

Amplitude

0
0 0

−1
−1 −1

−2 −2 −2
−0.5 0 0.5 −0.5 0 0.5 −0.5 0 0.5
Time Time Time

Fig. 8. Simulation example Fig. 9. Simulation example 1: Fig. 15. Simulation example 2: eye pattern of output equalizer with Rice
1: eye pattern of input equalizer eye pattern of output equalizer channel and SNR = 30dB
with Rayleigh channel and SNR = with Rayleigh channel and SNR =
20dB 20dB
Rayleigh and Rician channels parameters: Ts = 10−6 , fd =
2
SNR=30
2
SNR=30 10−4 , a carrier frequency fc = 1800.10−6 and Kr = 6.6923.
1.5 1.5 The simulation results are obtained using the Monte Carlo
1 1
simulations of 100 independent tests.
0.5 0.5
Imaginary

Imaginary

0 0

−0.5 −0.5 Simulation example 1 : In this simulation, we examined


−1 −1
the impact of the noise level on the input and output of our
−1.5 −1.5

−2 −2
equalizer for a 4-QAM modulation under Rayleigh channel.
−2 −1 0 1 2 −2 −1 0 1 2
Real Real
Two values for the SNR have been considered which are
10dB and 20dB. Fig. 2, Fig. 3, Fig. 4 and Fig. 5 show the
Fig. 10. Simulation example 2: Fig. 11. Simulation example 2: con-
constellation of sk/k with eSBME for stellation QAM of sk/k with eSMBE constellation of sk/k for the eSMBE algorithm. It is noted
AWGN channel and SNR = 30dB for Rayleigh channel and SNR = that sk/k in a neighborhood of a point of the constellation.
30dB The eye patterns on the input and output of our equalizer
under Rayleigh channel are shown in Fig. 6, Fig. 7, Fig. 8
2
SNR=30 and Fig. 9. We remark the closing of the eyes at the input
1.5
of our equalizer and the opening of the eyes at the output of
1 our equalizer.
0.5
Imaginary

−0.5
Simulation example 2 : In this simulation, we tested
−1 the impact of the noise level on the output of our equalizer
−1.5
for 30dB under AWGN, Rayleigh and Rice. channels.
−2
−2 −1 0
Real
1 2
Constellations and the eye patterns at the equalizer output
for different channels are appearing in Fig. 10, Fig. 11, Fig.
Fig. 12. Simulation example 2: constellation of sk/k with eSBME for Rice 12, Fig. 13, Fig. 14 and Fig. 15.
channel and SNR = 30dB

603
After the two examples simulations, we can say that the [11] A. Gorokhov. Blind equalization in SIMO OFDM systems with
proposed algorithm provided a good performance for CP- frequency domain spreading. IEEE Transactions on Signal Processing,
48(12):3536–3549, 2000.
OFDM under AWGN, Rayleigh and Rice channels. [12] T. Han and X. Li. Minimum-output-energy method for blind equal-
ization of OFDM and systems with sufficient or insufficient cyclic
V. C ONCLUSION AND FURTHER WORKS prefix. Acoustics, Speech, and Signal Processing, 2003. Proceedings.
(ICASSP ’03). 2003 IEEE International Conference on, 4, 2003.
In this document, we introduce a recursive blind equal- [13] D. Joachim and J. R. Deller Jr. Multiweight optimization in optimal
ization algorithm, based an Optimal Bounding Ellipsoid in bounding ellipsoid algorithms. IEEE Transactions on Signal Process-
ing, 54(2):679–690, 2006.
the presence of bounded disturbance, for an OFDM system [14] R. K. Martin and C. R. Johnson Jr. Blind, Adaptive Per Tone
which in the attendance of a guard interval. The proposed E qualization for M ulticarrier R eceivers. In Pro c. Conf. on Information
equalization has a relatively low computational burden and Sciences and Systems, Princeton, NJ, 2002.
it is easily applied in real time. Simulation examples are [15] A. Moussa, M. Frikel M. Pouliquen, S. Bedoui, K. Abderrahim, and
M. M’Saad. Performance of a blind equalization algorithm for rayleigh
provided to illustrate the performance of the proposed al- and rician channels. 7th Conference on Sciences and Techniques of
gorithm under different channels (AWGN, Rayleigh and Automatic control and computer engineering, Sousse, 2016.
Rice). These simulations show that the proposed method [16] A. Moussa, M. Frikel M. Pouliquen, S. Bedoui, K. Abderrahim, and
M. M’Saad. Recursive blind equalization for the bounded noise
is robust to a fading channel (Rayleigh and Rice) in the case under different modulations. 25th Mediterranean Conference on
context of multicarrier modulation. However, the proposed Control and Automation, Malte, 2017.
technique has an implementation problem due mainly to [17] A. Moussa, M. Pouliquen, M. Frikel, S. Bedoui, K. Abderrahim, and
M. M’Saad. Comparative study of blind equalizers based on optimal
the high number of subcarriers, which increases the cost bounding ellipsoid algorithm under AWGN and fading channels. 14th
of implementing. In addition to that the application of the Multi-conference on Systems, Signals and Devices, Marrakech, 2017.
proposed algorithm is limited to the QAM modulation in this [18] B. H. B. Naeeni and H. Amindavar. PT-CMA and PT-MMA for
ofdm in wireless communication. First International Conference on
paper. Communications and Networking in China, 2006.
Our future activities will focus on: i) the adaptation of [19] B. H. B. Naeeni, H. Amindavar, and H. Bakhsh. Blind per tone
equalization of multilevel signals using support vector machines for
the blind equalization approach to arbitrary constellation, ii) OFDM in wireless communication. AEU-International Journal of
stability study of our method for OFDM system, iii) the Electronics and Communications, 64(2):186–190, 2010.
optimization of the implementation of the blind equalization [20] M. Pouliquen, M. Frikel, and M. Denoual. Recursive blind equal-
ization with an optimal bounding ellipsoid algorithm. in EUSIPCO,
to reduce the number of subcarriers 2014.
[21] M. Pouliquen, E. Pigeon, and O. Gehan. Output error identification
R EFERENCES for multi-input multi-output systems with bounded disturbances. 50th
IEEE Conference on Decision and Control-Orlando, 2011.
[1] T. Y. Al-Naffouri, A. A. Dahman, M. S. Sohail, W. Xu, and B. Hassibi.
[22] H. Sari, G. Karam, and I. Jeanclaude. Transmission techniques for
L ow- C omplexity Blind Equalization for OFDM S ystems With G eneral
digital terrestrial TV broadcasting. IEEE Communications Magazine,
C onstellations. IEEE Transactions on Signal Processing, 60(12):6395–
33, Feb. 1995.
6407, 2012.
[23] V. Savaux, F. Bader, and J. Palicot. Correcting CNA phase mismatch
[2] F. O. Alayyan, K. Abed-Meraim, and A. M. Zoubir. Blind equalization phenomena in frequency blind equalization for OFDM systems. Signal
and frequency offset estimation in OFDM systems exploiting guard Processing, 2016.
interval redundancy. Proceedings of the Eighth International Sympo- [24] L. Wei, C. Ming, S. Cheng, and H. Wang. A Complexity Reduced
sium on Signal Processing and Its Applications, 2005., 1, 2005. B lind Channel E qualization Scheme for OFDM Systems. IEEE
[3] F. O. Alayyan, K. Abed-Meraim, and A. M. Zoubir. Blind equalization 17th International Symposium on Personal, Indoor and Mobile Radio
in OFDM Systems Exploiting Guard Interval Redundancy. Conference Communications, 2006.
Record of the Thirty-Ninth Asilomar Conference onSignals, Systems
and Computers, 2005., 4, 2005.
[4] F. O. Alayyan, K. Abed-Meraim, and A. M. Zoubir. Blind MMSE chan-
nel identification and equalization algorithms for OFDM systems. 9th
International Symposium on Signal Processing and Its Applications,
2007.
[5] F. O. Alayyan, Y. H. Leung, and A. M. Zoubir. LSF-CMA Equalization
A lgorithm for OFDM System using Z ero- P adding T echnique. IEEE
International Conference on Signal Processing and Communications,
2007.
[6] F. O. Alayyan, Y. H. Leung, A. M. Zoubir, R. M. Shubair, and
O. Alketbi. Blind channel identification and equalization for GI-less
OFDM systems. 1st IFIP Wireless Days, 2008.
[7] F. O. Alayyan, R. M. Shubair, A. M. Zoubir, and Y. H. Leung.
Blind channel equalization in GI-less OFDM systems using property
restoral. 17th International Conference on Telecommunications, 2010.
[8] R. Boloix-Tortosa, F. J. Payan-Somet, and J. J. Murillo-Fuentes.
Reduced complexity blind equalization schemes for ZP-OFDM sys-
tems. IEEE 8th Workshop on Signal Processing Advances in Wireless
Communications, 2007.
[9] M. de Courville, P. Duhamel, P. Madec, and J. Palicot. Blind
equalization of OFDM systems based on the minimization of a
quadratic criterion. Communications, 1996. ICC ’96, Conference
Record, Converging Technologies for Tomorrow’s Applications. 1996
IEEE International Conference on, 3:1318–1322, 1996.
[10] B. L. Floch, M. Alard, and C. Berrou. Coded orthogonal frequency
division multiplex [TV broadcasting]. IEEE Proceedings, 83(6):982–
996, Jun. 1995.

604

You might also like