2016 30th International Conference on Advanced Information Networking and Applications Workshops

Adaptive Carrier Frequency Offset and Channel

Estimation for MIMO-OFDM Systems
Chang-Yi Yang Yu-Li Chen Hai-Yan Song
Department of Computer Science and Department of Computer Science and College of Information And
Information Engineering Information Engineering Communication Engineering
National Penghu University of Science Ming Chuan University Harbin Engineering University
and Technology Taoyuan, Taiwan Harbin, China
Penghu, Taiwan [email protected] [email protected]
[email protected]

Abstract—The issues of carrier frequency offset (CFO) and

channel estimation are very important for multi-input multi- Therefore, channel estimation is essential for design of
output orthogonal frequency division multiplexing (MIMO- MIMO-OFDM systems. Different research approaches have
OFDM) wireless communication systems. In this paper, the been adopted for channel estimation and tracking [6]-[11]. An
problems of CFO and channel estimation are considered. An
expectation-maximization (EM)-based channel estimation
adaptive estimation algorithm is proposed. It can estimate CFO
and channel parameters at the same time. Since the estimation method has been proposed for channel estimation to detect the
method is real-time, it can work at a fast time-varying transmitted data in [12]. However, the EM algorithm needs
environment. However, the model becomes a huge nonlinear numerous iterations to achieve optimal channel estimation
system if CFO and channel parameters are estimated results.
simultaneously. It is a heavy load for the wireless communication Although an extended Kalman filter with the pilot-
system, even the extended Kalman filter can solve the nonlinear
symbol-aided method to track a MIMO-OFDM channel has
estimation. A multiple model, which includes many subchannel
models, is constructed. The subchannel with CFO can be been proposed in [13], channel estimation with a pilot-symbol-
separately estimated. The accurate CFO is obtained by aided method may occupy the available bandwidth (valuable
integrating estimated CFOs from subchannel models with for the service provider) and may degrade the data throughput
weighting technology. The computational complexity can be [14], [15]. Other research works include the cyclic comb-type
reduced by this method. Finally, several simulations are given to training structure proposed in [16], in which all types of
prove the performance in comparison with other methods. training symbols are transmitted cyclically at each antenna and
estimated for the channel frequency response with weights
Index Terms—Multi-input multi-output orthogonal
frequency division multiplexing (MIMO-OFDM) communication obtained from the corresponding mean-square errors (MSEs)
system, carrier frequency offset (CFO), extended Kalman filter. at the receiver. An iterative maximum likelihood (ML) channel
estimation for signal detection under a large variation of
channel responses in the frequency domain was proposed in
I. INTRODUCTION [17]. In addition, a decision-directed channel estimation in the
With the success of the multi-input multi-output (MIMO) frequency domain using a Kalman-based filter was proposed
and the orthogonal frequency division multiplexing (OFDM) in [18]. All of the above researches only consider the channel
systems, the combination of both techniques, known as estimation problem. The CFO due to mismatch between
MIMO-OFDM, has gained considerable attention for transmitter and receiver oscillators is not taken into account in
broadband communications [1]-[4]. The advantages of the those algorithms.
MIMO-OFDM system include antenna diversity, high data Since the accuracy of the channel estimation result
rate, elimination of inter-symbol interference (ISI) effect and depends on the exact carrier frequency at the receiver, the idea
low complexity. of joint CFO and channel estimation has initially been
One of the disadvantages of MIMO-OFDM lies in its proposed for OFDM system [19], and then extended to
sensitivity of carrier frequency offset (CFO) due to carrier MIMO-OFDM systems [20]. A closed-form joint CFO and
frequency mismatches between transmitter and receiver channel estimation was proposed for point-to-point MIMO-
oscillators. The CFO induces a phase error that causes a loss OFDM systems [21] and then extended to coordinated MIMO-
of orthogonality among sub-carriers and, therefore, increases OFDM systems. However, these closed-form solutions are not
inter-carrier interference (ICI) and attenuates the desired real-time so that they are not suitable for fading channel
signal. These effects reduce the effective signal-to-noise ratio environment.
in OFDM reception such that the system performance is A pilot-aided algorithm has been developed in [22] for
degraded [5]. MIMI-OFDM systems operating in a fast time-varying
In order to approach the theoretical capacity and achieve environment. It performs the estimation using extended
maximum diversity gain, channel parameters are required. Kalman filter and basis expansion model (BEM) technology.

978-1-5090-2461-2/16 $31.00 © 2016 IEEE 949

DOI 10.1109/WAINA.2016.68
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on September 25,2023 at 18:01:58 UTC from IEEE Xplore. Restrictions apply.
 As the channel delay spread increases, the number of channel assumption that can be made is that the velocities of any
taps also increases, and a large number of BEM coefficients antenna pair between the transmitter and the receiver are all
have to be estimated. This requires more pilot symbols. equal in any time instant, i.e., a1j ,i (n)  a1 (n) , i, j . Since
The motivation of this paper is to develop a new adaptive
the velocity of the mobile station is time-varying and
algorithm capable of jointly estimating the CFO and channel
parameters. However, the model becomes a huge nonlinear unavailable, a1j ,i (n) is time-varying too.
system if CFO and channel parameters are estimated Meanwhile, it is assumed that all the transmitting antennas
simultaneously. A multiple model, which includes many at a BS share the same oscillator, as do the receiving antennas
subchannel models, is constructed. The subchannel with CFO at the MS. Therefore, a MS has one time-varying CFO,
can be separately estimated. The accurate CFO is obtained by f (n) . For a subcarrier frequency of fs Hz, the CFO is
integrating estimated CFOs from subchannel models with
normalized to the subcarrier frequency as v(n)  f (n) / f s .
weighting technology. The computational complexity can be
reduced by this method. The simulation results demonstrate As described in [22], v(n) can be defined as the following
the improvements of the proposed scheme over other methods. AR(1) model
v(n)  bv(n  1)  v (n) (3)
II. SYSTEM MODELS OF MIMO-OFDM SYSTEMS where b is typically chosen between 0.99 and 0.9999, and
A MIMO-OFDM system with Nt transmit antennas, Nr wv(n) is assumed to be zero-mean Gaussian. In order to
receive antennas, and Nc subcarriers is considered. As shown prevent the system performance degraded, the best way is to
in Fig. 1, the information bits are first modulated by an M-ary jointly estimate a1(n), v(n), and hl j ,i (n) .
quadratic amplitude modulator (QAM). The modulated M-
QAM symbols are encoded by an OSTBC encoder and fed
into an OFDM modulator. In the OSTBC scheme [1], the data III. ADAPTIVE ALGORITHM FOR JOINT CFO AND CHANNEL
stream to be transmitted is encoded in blocks, which are ESTIMATION
distributed among the spaced antennas and across time.
The transmitted signal shall be regenerated by an equalizer
from received signal. Therefore, the equalizer requires the
channel information to minimize the MSE between the
transmitted signal and the regenerated signal. Since the
channel information is unknown to the equalizer, the channel
estimation of the MIMO-OFDM system becomes an important
Fig. 1. Block diagram of the MIMO-OFDM transmitter issue. A decision-directed Kalman filter is proposed to
The time-domain channel impulse response from the jth estimate the channel information. The MMSE equalizer can be
transmit antenna to the ith receive antenna of the lth path implemented according to the estimation result. The block
sampling of the complex time-varying fading channel with diagram of the MIMO-OFDM receiver is shown in Fig. 2.
length L at the nth OFDM symbol is represented as hl j ,i (n) .
For the Rayleigh fading channel, an AR process can match its
special characteristics [13], [19], [23]. The second-order AR
model (AR(2)) is adopted here due to its simplicity and
accuracy. Therefore, the lth path of the time-domain channel
impulse response from the jth transmit antenna to the ith
receive antenna at the nth OFDM symbol can be modeled by
hl j ,i (n)  a1j ,i (n 1)hl j ,i (n 1)  a2j ,i hl j ,i (n  2)  lj ,i (n) (1)
for i  0, , Nr  1 , j  0, , Nt  1 , and l = 0, , Nc-1, Fig. 2. An MIMO-OFDM receiver with decision-directed channel tracking
where l j ,i (n) is a zero-mean complex white Gaussian Here, one receiver antenna (Nr=1) is considered in
process. Parameter a1j ,i (n) and a2j ,i are determined by Doppler designing the joint frequency offset and channel estimator,
since the same estimation procedure can be adopted on the
frequency shift due to the velocity change of the mobile
other receiver antennas. However, the estimates from different
station. These parameters are defined in [24] as receiver antennas can also be combined to achieve a more
a1j ,i (n)  2rd cos  2 f Dj ,i (n)T  , a2j ,i  rd2 (2) accurate estimate.
j ,i
Since all receiver antennas share the same receiver
where f (n) is the maximum Doppler frequency shift in the
D oscillator. The time-domain impulse response hl j ,i (n) can be
nth symbol period, and rd is the pole radius that corresponds to
simply expressed as hl j (n) . The training signal Si,k(n) means
the steepness of the peaks of the power spectrum. A possible

950

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on September 25,2023 at 18:01:58 UTC from IEEE Xplore. Restrictions apply.
 the signal from the ith subchannel of the kth transmit antenna. method to estimate them simultaneously. The extended
The received vector r (n) for all subcarriers is defined as Kalman filter is introduced here to solve the nonlinear
T estimation problem.
r (n)  r0 r1 rNc 1  C Nc . Therefore, the received From the dynamics of the state equations in (7) and
signal for each subchannel can be expressed as Kalman filter algorithm [25], we can obtain
ri (n)  y i (n)  wi (n), i  0, , Nc  1 (4) ˆ n  1| n  1))
X̂(n | n  1)  G(X( (10)
where X̂(n | n) = X̂(n | n  1) + Ka (n)e(n) (11)
2 iv ( n )
Nt 1 j Nt 1 where
y i ( n)  S
k 0
i,k ( n)e Nc
hik (n)  s
k 0
i,k (n)hik (n) e(n)  r (n)  D(X( ˆ n | n  1)) (12)
(5)
2 iv ( n ) and Ka(n) is the corresponding Kalman gain. Here, e(n)
j
si , k (n)  Si , k (n)e Nc
denotes the prediction error. Ka(n) is employed to minimize
the covariance E{e(n)e(n)T} of the prediction error. The
Let extended Kalman gain Ka(n) and the error covariance matrices
H
hi (n)   h0i (n) h1i (n) hNi c 1 (n)  P(n|n) of the smooth range prediction and estimation are
(6) obtained from the following equations:
H
h(n)   h H 0 (n) h H 1 (n) h H Nt 1 (n)  P  n | n  1  gP  n  1| n  1 g H   V2 I (13)
1
Joint CFO and channel estimation problem can be transformed Ka (n)  P  n | n  1 d dP  n | n  1 d  Q 
H H 2
(14)
into the following state-space equations.
P  n | n    I  Ka (n)d  P  n | n  1 (15)
X(n)  G(X(n 1))  V(n) (7) where
r (n)  D(X(n))  w(n) (8) b 0 0 0
0 0 
where G  X   1 0
 ( n) b (n) v (n) g=  (16)
 a ( n)     (n)  X 0 h(n) a1 I Nt Nc a2 I Nt Nc 
X(n)   1  , G (X(n))   a1 (n)  , V(n)  a 
 (9)  
 h(n)   a1 (n)h(n)  a2 h(n  1)  Vh (n)  0 0 I Nt N c 0 Nt Nc 
     
 h(n  1)   h(n)   0 
and
However, the joint estimation in (7) is a nonlinear problem
due to the coupling. It is not easy for any linear estimation
------------------------------------------------------------------------------------------------------------------------------------------------------------
0 0 s0,0 (n) 0 0 0 s00,11 (n) 0 0 s0,0 Nt 1 ( n) 0 0
 2 
 j y1 0 0 s1,0 (n) 0 0 0 s1,1 (n) 0 0 s0, Nt 1 ( n) 0 
D  X   Nc 
d=   (17)
X  
 2 ( N c  1) 
j yNc 1 0 0 0 0 sNc 11,00 (n) 0 0 sNc 1,1
1 1 ( n) 0 0 sNc 11, Nt 1 ( n) 
 Nc 

It is noted that r(n) C Nc , X(n) C 

2 Nt Nc  2
, Ka (n) C 
2 Nt Nc  2 Nc
, P(n | n) C 
2 Nt Nc  2 2 Nt Nc  2
C
2 Nt Nc  2 2 Nt Nc  2
, g , and
Nc  2 Nt Nc  2 N c  Nc
d C . Meanwhile, there is a matrix ( C ) to be inverted when K a (n) is calculated. The computational
complexity is very high.
-------------------------------------------------------------------------------------------------------------------------------- -----------------------
ri (n)  Di (Xi (n))  wi (n) (20)
IV. ADAPTIVE MULTIPLE SUBCHANNEL ESTIMATION where
 ( n) b (n) v (n)
 a ( n)      ( n) 
To reduce the computational complex, a multiple model,    a1 (n)   a 
X i ( n)  1
, G (X (n))  , V ( n)  (21)
 H i ( n)  i i  a1 (n) H i (n)  a2 H i (n  1)  i Vh (n) 
which includes many subchannel models, is proposed. For      i 
 H i (n  1)   H i ( n)   0 
subchannel i, we define the transmitted impulse response
vector Hi(n) as follows. Nt 1 j
2 iv ( n ) Nt 1

Nt 1 H Di (Xi (n)) = yi (n)  S i ,k (n)e Nc

hik (n)  s i ,k (n)hik (n) (22)
Hi (n)  h (n) h (n)
i
0 1
i h i (n)  (18) k 0 k 0

Associated with (4), the following state-space equation can be

By using the extended Kalman filter, the following equations
obtained for i  0, , Nc  1 .
can be obtained:
Xi (n)  Gi (Xi (n 1))  Vi (n) (19)

951

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on September 25,2023 at 18:01:58 UTC from IEEE Xplore. Restrictions apply.
 Note that ri  n  C1 , Xi (n) C 
2 Nt  2
, Ka (n) C 
2 Nt  21
ˆ (n  1| n  1))
X̂i (n | n  1)  Gi (X (23) ,
i
1 2 Nt  2
Pi (n | n ) C 
2 Nt  2 2 Nt  2
C
2 Nt  2 2 Nt  2
X̂i (n | n)  X̂i (n | n  1)  Kai (n)e(n) (24) , gi , d C .
Meanwhile, no matrix shall be inverted when K a (n) is
where
calculated. This algorithm can reduce the computational
ˆ (n | n  1)) complexity. Although there are Nc EKFs, it is still worthy.
ei (n)  ri (n)  di (Xi (25)
V. COMPUTER SIMULATION
The extended Kalman gain K ai (n) and the error covariance
Table 1 shows the power delay profile in the urban area
matrices Pi(n|n) of the smooth range prediction and estimation with the root-mean-square delay T  1s . Mobile velocities
are obtained from the following equations: from 5 to 200 km/hr are used to simulate different mobile
environments. The central frequency f c is 2 GHz in the
Pi  n | n  1  gi Pi  n  1| n  1 giH   2
Vi I (26) MIMO-OFDM system. The available bandwidth BW is 2.048
1
MHz which is divided into 512 subchannels. These
Kai (n)  Pi  n | n  1 diH di P  n | n  1 diH  Q2i  (27) correspond to a subcarrier symbol rate of 4 kHz and an
OFDM symbol duration of 312.5 s . The number of CPs
Pi  n | n    I  Kai (n)d  Pi  n | n  1 (28) N cp  128 is used to provide protection from ISI due to the
channel multipath delay spread. In the following simulations,
where
two transmitter antennas and two receiver antennas are used
b 0 0 0 ( Nt  2, N r  2 ) and the Alamouti code ( T  2 ) is adopted
0 0 
Gi  Xi  1 0 [2].
gi =  (29)
Xi 0 H i (n) a1 I Nt a2 I Nt 
  Table 1. Power Delay Profile in the Typical Urban Area
0 0 I Nt 0 Nt  Delay 0.0 0.2 0.5 1.6 2.3 5.0
Fractional 0.189 0.379 0.239 0.095 0.061 0.037
Di  Xi  2 i
di = j yi 0 si ,0 (n) si ,1 (n) si , Nt 1 (n) 0 0 0 Power
Xi  Nc 
(30)
By running the simulation for 50 trials and 1000 OFDM
There are Nc EKFs to estimate X̂i (n | n), i  1, , Nc . The block symbols per trial in this example, the performance of
the system is measured in term of the normalized mean square
covariance of E ei (n)eiT (n) of the smooth range prediction error (NMSE) of the channel estimation. The NMSE for the
error is expressed as mth subcarrier is defined as

E ei (n)eiT (n)  i (n)  di P  n | n  1 diH  Q2i (31)

NMSE( m) 

E H n (m)  Hˆ n (m)
2
 (36)
for i  0, , Nc  1 . Every X̂i (n | n) contains ν(n) and a1(n) .

E H n (m)
2

To get the accurate estimation result, we shall calculate the To simulate a velocity-varying mobile radio channel, we
weighting for each EKF. The likelihood function  i and the first assume that the velocity varies linearly in one second.
Following this, the initial and final velocities in one second
weighting wi for ith EKF, respectively, can be updated by are random variables uniformly distributed in the range from
5 km/hr ( f D  9.25 Hz) to 200 km/hr ( f D  370 Hz) in each
i (n)  N  ei (n); 0, i (n)  (32) trial. The compared methods are the algorithms which can
 i ( n) provide the channel estimation and prediction simultaneously.
wi (n) 
  ( n)
i
i
(33) The NMSEs, using different channel estimation methods, are
illustrated in the Fig. 3. The results of the 1st-order Kalman
where N  ei (n); 0, i (n)  denotes the Gaussian density filter [18] and the 2nd-order Kalman filter [25] were failures
function of ei(n) with zero mean and covariance i (n) . The because these two methods adopt fixed dynamics. The
estimated ν(n) and a1(n) can be obtained by proposed algorithm has best performance among testing
Nc 1
methods.
 w Xˆ (n) 1 0
T
vˆ(n)  i i 0 (34) VI. CONCLUSION AND FUTURE WORK
i
Nc 1 An adaptive carrier frequency offset and channel
 w Xˆ (n) 0 0
T
aˆ1 (n)  i i 1 0 (35) estimation for MIMO-OFDM systems has been proposed in
i
this paper. The channel model is described by a nonlinear
state-space dynamic equation. The states of the nonlinear

952

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on September 25,2023 at 18:01:58 UTC from IEEE Xplore. Restrictions apply.
 channel parameter system include the CFO, the time-varying [3] D. Gesbert, M. Shafi, D. Shiu, P. J. Smith, and A. Naguib, “From
theory to practice: An overview of MIMO space-time coded wireless
channel gains, and the dynamics of the channel. The CFO can systems,” IEEE J. Sel. Areas Commun., vol. 21, no. 3, pp. 281-302,
be described as an AR(1) model. However, the channel model April 2003.
becomes a huge nonlinear system. Its computational [4] A. J. Paulraj, D. A. Gore, R. U. Nabar, and H. Bolcskei, “An overview
complexity is very high. To reduce the computational of MIMO communications–a key to gigabit wireless,” Proc. IEEE, vol.
92, no. 2, pp. 198-218, February 2004.
complexity, a multiple model, which includes many
[5] R. N. A. J. Paulraj and D. Gore, Introduction to space-time wireless
subchannel model models, is constructed. The subchannel communications. Cambridge University Press, 2003.
state with CFO can be separately estimated. The accurate [6] H. Bocskei and A. J. Paulraj, Multiple-input multiple-output (MIMO)
CFO, then, is obtained by integrating estimated CFOs from wireless systems. Cambridge University Press, 2003.
subchannel models with weighting technology. The [7] G. J. Foschini and M. J. Gans, “On limits of wireless communications
simulation result indicates the proposed method can more in a fading environment using multiple antennas,” Wireless Personal
Commun., vol. 6, no. 3, pp. 311-355, 1998.
accurately estimate the channel state than existing methods.
[8] H. A. Suraweera and J. Armstrong, “Alamouti coded OFDM in
Several works are not finished yet . First, most of the channel rayleigh fast fading channels- receiver performance analysis,” IEEE
tracking algorithms assumes the transmitted signal sn(m) can Region 10 TENCON, pp. 1-5, Nov. 2005.
be obtained. This is only true in the training mode, but the [9] M. Uysal, N. Al-Dhahir, and C. N. Georghiades, “A space-time block-
coded OFDM scheme for unknown frequency selective fading
transmitted signal is unknown for receiver in the tracking channels,” IEEE Commun. Lett., vol. 5, no. 10, pp. 393-395, Oct. 2001.
mode. The proposed algorithm shall be modified for channel [10] B. Lu, X. Wang, and Y. G. Li, “Iterative receivers for space-time
estimation in the tracking mode. Second, after the channel block-coded OFDM systems in dispersive fading channels,” IEEE
parameters are estimated, they shall be used to design the Trans. Wireless Commun., vol. 1, no. 2, pp. 213-225, April 2002.
MMSE equalizer. A robust optimal MMSE equalizer wishes [11] R. S. Blum, Y. G. Li, J. Winters, and Q. Yan, “Improved space-time
coding for MIMO OFDM wireless communications,” IEEE Trans.
to be obtained from the information of the estimation Commun., vol. 49, no. 11, pp. 1873-1878, November 2001.
algorithm. These are our future works. [12] M. Enescu, T. Roman, and V. Koivunen, “Channel estimation and
tracking in spatially correlated MIMO OFDM systems,” In Proc. IEEE
Workshop Stat. Signal Process., pp. 347-350, Sep. 2003.
[13] C. Min, N. Chang, J. Cha, and J. Kang, “MIMO-OFDM downlink
channel prediction for IEEE802.16e systems using Kalman filter,” Proc.
IEEE WCNC’07, Kowloon, pp. 942-946, March 2007.
[14] F. Wan, W. P. Zhu, and M. N. S. Swamy, “A semi-blind channel
estimation approach for MIMO-OFDM systems,” IEEE Trans. Signal
Processing, vol. 56, no. 7, pp. 2821-2834, 2008.
[15] D. Angelosante, E. Biglieri, and M. Lops, “Sequential estimation of
multipath MIMO-OFDM channels,” IEEE Trans. Signal Process., vol.
57, no. 8, pp. 3167-3181, 2009.
[16] L. Li, H. Li, H. Yu, B. Yang, and H. Hu, “A new algorithm for MIMO
channel tracking based on Kalman filter,” IEEE Wireless
Communications and Networking Conference (WCNC ’07), pp. 164-168,
March 2007.
[17] T.-J. Ho and B.-S. Chen, “Tracking of dispersive DS-CDMA channels:
an AR-embedded modified interacting multiple-model approach,” IEEE
Trans. Wireless Commun., vol. 6, no. 1, pp. 166-174, January 2007.
[18] D. N. Kalofonos and M. Stojanovic and J. G. Proakis, “Performance of
Fig. 3. NMSEs of channel estimations by different methods. adaptive MC-CDMA detectors in rapidly fading Rayleigh channels”,
IEEE Trans. wireless Communications, vol. 2, no. 2, pp. 1375-1387,
Mar. 2003.
ACKNOWLEDGMENTS [19] T. Roman, M. Enescu, and V. Koivunen, “Joint time-domain tracking of
channel and frequency offset for OFDM systems,” in Proc. 4th IEEE
This work of C-Y Yang was partly supported by the Workshop SPAWC, Jun. 15-18, 2003, pp. 605-609.
Ministry of Science and Technology (MOST) in Taiwan [20] T. Roman, M. Enescu, and V. Koivunen, “Joint time-domain tracking of
channel and frequency offsets for MIMO OFDM systems,” Wireless
under grant MOST 105-2221-E-346-002. This work of H-Y Pers. Commun., vol. 31, no. 3/4, pp. 181-200, Dec. 2004.
Song was partly supported by the Natural Science Foundation [21] Y-J. Liang, G. Stuber, J-F. Chang, and D-N Yang, “A joint channel and
of Heilongjiang Province, China under grant QC2014C079. frequency offset estimator for the downlink of coordinated MIMO-
OFDM systems,” IEEE Trans. on Wireless Commun., vol. 11, no. 6, pp.
2254-2265, June 2012.
[22] E. Simon, L. Ros, H. Hijazi, J. Fang, D. Gaillot, and M. Berbineau,
REFERENCES “Joint carrier frequency offset and fast time-varying channel estimation
[1] V. Tarokh, H. Jafarkhani, and A. R. Calderbank, “Space-time block for MIMO-OFDM systems,” IEEE Trans. on Vehicular Technology, vol.
codes from orthogonal designs,” IEEE Trans. Inform.Theory, vol. 45, 60, no. 3, pp. 955-965, March 2011.
no. 5, pp. 1456-1467, July 1999. [23] W.-G. Song and J.-T. Lim, “Channel estimation and signal detection for
[2] S. M. Alamouti, “A simple transmit diversity technique for wireless MIMO-OFDM with time varying channels,” IEEE Communications
communication,” IEEE J. Sel. Areas Commun., vol. 16, no. 8, pp. 1451- Letters, vol. 10, no. 7, pp. 540-542, July 2006.
1458, Oct. 1998. [24] I. Barhumi, G. Leus, and M. Moonen, “Optimal training sequences for
channel estimation in MIMO-OFDM systems in mobile wireless

953

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on September 25,2023 at 18:01:58 UTC from IEEE Xplore. Restrictions apply.
 channels,” in Int. Zurich Symp. Broadband Communications, Feb.2002,
pp. 441-446, 2002.
[25] S. Haykin, Adaptive Filter Theory 4th Edition. Prentice-Hall, 2002.
[26] L. Lindbom, “Simplified Kalman estimation of fading mobile radio
channels: highperformance at LMS computational load,” Proc. ICASSP,
vol. 3, pp. 352-355, Apr. 1993.

954

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY BOMBAY. Downloaded on September 25,2023 at 18:01:58 UTC from IEEE Xplore. Restrictions apply.