INTERNATIONAL JOURNAL OF COMMUNICATION SYSTEMS

Int. J. Commun. Syst. (2013)

Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/dac.2510
SPECIAL ISSUE ARTICLE
Accurate estimation of Doppler shift in mobile communications
with high vehicle speed
J. Y. Hua
1,2,
*
,
, D. H. Yuan
1
, G. Li
1
and L. M. Meng
1
1
College of Information Engineering, Zhejiang University of Technology, Hangzhou, 310032, China
2
National Mobile Communication Research Laboratory, Southeast University, Nanjing 210096, China
SUMMARY
Representing the channel varying rate and the mobile speed of a mobile terminal directly, Doppler shift is
an important parameter in vehicular mobile communications and therefore is widely used in mobile target
detection and adaptive applications. Hence, this paper puts forward an accurate Doppler shift estimator in
mobile communications with high vehicle speeds, which can also be treated as a vehicular speed estimator
due to the well-known relation between the Doppler shift and the mobile speed. Specically, the proposed
estimator is based on the channel level crossing rate, and an iterative process is presented to achieve signal-
to-noise ratio (SNR) insensitive estimates in accordance with the level crossing rate estimation error analysis.
Moreover, we prove the convergency of the iterative Doppler shift estimator in theory. Computer simulations
conducted under a wide range of noise corruption clearly show that the proposed estimator substantially out-
performs several existing estimators in terms of accuracy and achieves a good SNR-insensitive performance
in a wide range of velocities and SNRs. Copyright 2013 John Wiley & Sons, Ltd.
Received 7 May 2012; Revised 29 October 2012; Accepted 5 January 2013
KEY WORDS: Doppler shift; iterative process; level crossing rate; high vehicle speed; mobile communi-
cations.
1. INTRODUCTION
Recently, with the development of mobile communications, signal reception in vehicular platform
has become a heated topic and attracts much attention [14], which together with the intelligent
transportation system (ITS), converges two important elds of electronics remote sensing and
mobile communication, and has been proposed to provide the transportation safety guarantee and
vehicular data communication [4]. Generally, vehicular communication includes vehicle-to-vehicle
and vehicle-to-infrastructure (V2I) communications, and this paper focuses on the signal processing
in V2I systems, such as the cellular mobile communication system.
Generally, the speed estimation can be done by the inductive loop detector (ILD) [5] or the
Doppler shift (
d
) [6]. However, as for V2I systems, the ILD requires additional cost of special
loop sensor systems, and therefore, the Doppler shift-based estimator is preferred, where the latter
exploits the fact that the fading rate of a channel depends on its maximum Doppler shift, and the
Doppler shift is related to the velocity of the mobile terminal (MT) [79]. This important rela-
tionship always comes into existence, irrespective of the background noise intensity. Moreover, the
rapid fading or fast mobile speed is a central issue in mobile communications [3, 7, 10, 11], which
usually causes negative inuence [1214], and as a branch of mobile communications, the vehicular
communication also suffers from the fast fading channel caused by high vehicle speeds [1, 3, 4].
*Correspondence to: Jingyu Hua, College of Information Engineering, Zhejiang University of Technology, Hangzhou,
310032, China.

E-mail: eehjy@163.com
Copyright 2013 John Wiley & Sons, Ltd.
J. Y. HUA ET AL.
According to the requirements of 3G and 4G systems, the vehicle speed may span from 0 to
500 km/h. Hence, an effective estimation of the Doppler shift or the MT velocity is of great impor-
tance in vehicular mobile communications and will produce a good power control performance [15],
a sound handoff performance [16], an effective dynamic channel parameter estimation [1720],
an outstanding adaptive transmission performance [21, 22], a good position estimation [23], and a
good target tracking ability [2]. These requirements conrm the necessity of an accurate Doppler
shift estimation in modern vehicular communication systems. To estimate the Doppler shift and
then the mobile speed, scientists had studied some methods in [16, 17, 2428] and their reference
therein. Unfortunately, all of the aforementioned Doppler shift estimators, except for that in [27],
did not explicitly give consideration to the noise effect in Doppler shift estimation; thus, they expe-
rienced signicant performance degradations as a result of the inuence of additive white Gaussian
noise (AWGN). In addition, the AWGN had been taken into account for the carrier frequency off-
set estimation [29]; however, the carrier frequency offset is very different from the Doppler shift
in the mechanism as well as the inuence [30] and therefore is beyond the scope of this paper.
Accordingly, we only focus on the Doppler shift estimation in our study.
In [27], Hua et al. proposed an iterative autocorrelation function (ACF) method to reduce the
effects of AWGN, but its intrinsic series approximation errors make it only effective for small
Doppler shift ranges. On the other hand, the ITS may require both high estimation accuracies and
large estimation ranges of Doppler shift; for example, the overspeed alarm is not operable in the
infrastructure center without accurate Doppler shift estimations. In fact, the crossing rate-based
Doppler shift estimator produces no series approximation errors and is robust to the propagation
environment [31]. Here, the level crossing rate (LCR) of a random process (RP) is dened as
the cross number per second when the RP envelope level down-crosses a certain threshold level.
Accordingly, this paper proposes a Doppler shift estimator capable of reducing noise inuence
effectively without reducing the estimation range, while conventional methods failed to resolve the
two problems at the same time. Specically, we rst give an analysis of the signal-to-noise ratio
(SNR)-insensitive conditions for the LCR-based method and then implement these conditions with
a simple iterative process. To our knowledge, it is the rst time that the iterative LCR method is
proposed to confront the two problems mentioned earlier. Moreover, we prove the convergency of
the proposed estimator by analytical derivation. We also verify our algorithm with Monte Carlo sim-
ulation, where accurate and SNR-insensitive estimates are obtained in a wide range of velocities and
SNRs, viz., within the error tolerance limit (5%). Because the concerned speed in this paper ranges
from 30 km/h ~ 480 km/h, it must be in the vehicular environment, and the conventional study
did not concern such a high-speed case as 480 km/h. In addition, the proposed estimator substan-
tially outperforms the traditional Doppler-based estimator [6] and maintains at least the comparable
performance as the ILD method, which makes the proposed Doppler shift estimator of signicant
novelty and suitable for real-world applications.
The remainder of this paper is organized as follows. Section 2 presents the signal model under
consideration together with the conventional LCR-based Doppler shift estimator. And then, the esti-
mation bias, the SNR-insensitive condition, and the iterative process are presented in Section 3.
Finally, numerical results are analyzed in Section 4, and conclusions are summarized in Section 5.
2. SIGNAL MODEL
In high-speed vehicular environments, the transmitted signal from the base station is corrupted by
the fading channel. Let us suppose that a band-limited pilot signal is transmitted over fading chan-
nels, and the distinguishable multi-path fading channels are wide-sense stationary and mutually
uncorrelated scattering processes. After synchronously matching the pilot signal, the expression of
channel estimates can be written according to [32]
c
l
(n) =c
l
(n) z(n) (1)
where c
l
(n), c
l
(n), and z(n) represent the channel estimates, the actual channels, and AWGN (with
variance o
2

), respectively. n is the discrete time index, and l denotes the path index. Here, the actual
Copyright 2013 John Wiley & Sons, Ltd. Int. J. Commun. Syst. (2013)
DOI: 10.1002/dac
ACCURATE ESTIMATION OF DOPPLER SHIFT
channels (c
l
(n)) are modeled as zero-mean complex Gaussian RPs with variance o
2
l
. For notational
brevity and without loss of generality, the strongest path is assumed to be the rst path, denoted
as c(n).
Generally, LCR of the envelope of a RP can be dened as the cross number per second when the
envelope level down-crosses the threshold level (
th
). Moreover, precise expressions of LCR over
Rayleigh channels have been derived in [33]:
N

LCR
=
s
b
2
b
0
e

2
(2)
b
n
=(2)
n
o
2
l

Z
f
d
f
d

n
p

n
d

n
d (2)
n
o
2

Z
1=2T
s
1=2T
s

n
d (3)
where ,
d
, and T
s
denote the ratio of the threshold level to the envelope root mean square level,
the actual Doppler shift, and the pilot symbol interval, respectively. Generally, we choose = 1.
Then, the simplied LCR expression and the LCR-based Doppler shift estimator (assuming zero
AWGN) can be derived as [17, 33]
N

LCR
=
_
2e

d
(4a)

d0
=
e
_
2

N
1
LCR
(4b)
In Equation (4b), is xed as 1. Note that different alphas will introduce different N

LCR
s and
therefore different expressions of Equation (4b), but the nal Doppler shift estimation will not be
affected by the choice of because must be known to the scientists in deriving the estimators.
Furthermore, we store 1 channel estimates of the lth path to estimate LCR

N
1
LCR

, so long as
the time length (1T
s
) is larger than the fading period. Although Equation (4b) does not exploit
any series approximation like the ACF method (i.e., its estimation range is much larger than that of
the latter), it is derived under the assumption of innite SNR. Unfortunately, AWGN is inevitable
in real-world scenarios and may cause some estimation biases. This kind of bias will be described
clearly in the next section.
3. ITERATIVE DOPPLER SHIFT ESTIMATOR
In [28], Hua and You proposed to use a decimator to rene the Doppler shift estimation, and in [27],
Hua et al. suggested to use adaptive autocorrelation lags to obtain the Doppler shift estimation.
These literature presented a rough discussion about the iterative Doppler shift estimator; however,
their estimation ranges are small.
In our investigation, we will further study the iterative estimation technique in LCR calculation.
First we will show a fairly high estimation bias in Equation (4b), followed by an effective iterative
method to remove this bias. The algorithm substantially utilizes the useful information hidden in
the estimation bias of Doppler shift and therefore outperforms conventional algorithms. Moreover,
the LCR calculation does not require series approximation, which effectively extends the Doppler
shift estimation range compared with the correlation-based method. These two advantages make up
of the novelties in our study.
3.1. Bias analysis of the estimator dened in Equation (4b)
According to Equations (2) and (4b),

N
LCR
is functional on AWGN, so is

d0
. Hence, the estimator
of Equation (4b) is unbiased only in noise-free scenarios. To account for this inuence, rst, we
Copyright 2013 John Wiley & Sons, Ltd. Int. J. Commun. Syst. (2013)
DOI: 10.1002/dac
J. Y. HUA ET AL.
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
0
5
10
15
20
25
0dB
5dB
10dB
15dB

Normalized Doppler shift (f

d
T
s
)
Figure 1. Estimation bias in noisy case.
assume that 1 is large enough

i.e.,

N
1
LCR
~N
1
LCR

; then, we dene the ratio of Doppler shift

estimation in noisy scenarios to that in noise-free scenarios as [33]
j =
e
p
2

N
1
LCR
(noisy)
e
p
2

N
1
LCR
(noise-free)
~
s
b
2
b
0

b
s2
b
s0
(5)
where b
s2
and b
s0
are calculated according to Equation (3) with o
2

= 0, while b
2
and b
0
are
computed with o
2

=0. After simple calculation and substitution, we have

j =
s
1
(1,2
d
T
s
)
2
6
6(;
s
1)
(6)
where ;
s
= o
2
l
,o
2

denotes the symbol signal to noise ratio. By Equation (6), we can investigate
the estimation bias caused by AWGN in theory. Accordingly, Figure 1 is presented through the
numerical computation of Equation (6).
From Figure 1, we clearly nd that the estimation bias is a function of SNR and
d
T
s
. It is
obvious that a higher SNR and a larger
d
T
s
lead to a smaller bias. These results cast light on the
improvement for the original LCR-based Doppler shift estimator.
3.2. SNR-insensitive conditions
From Equation (6) and Figure 1, when the actual Doppler shift (
d
) is given, different sample
intervals (different T
s
s) lead to different doppler shift estimation biases, when T
s

d
rises to a par-
ticular value, the j curves tend to be superposed, which means that the AWGN inuence tend to be
same for all the curves at this time. This is hidden and useful information because it supplies us with
additional information of
d
so long as more than one sampling rates are taken into consideration;
that is, if we can choose an appropriate sample interval, the estimation error will be trivial.
Generally, j must be larger than 1 as shown in Figure 1, and if the LCR-based Doppler shift
estimator approach the performance at noise-free scenarios, it must have j =1 and lead to a certain
sample interval. To nd this sample interval, we assume j = 1 and substitute it into Equation (6);
then, we have the following equation:
1 =j =
s
1
1,(2
d
T
s
)|
2
6
6(;
s
1)
(7)
Copyright 2013 John Wiley & Sons, Ltd. Int. J. Commun. Syst. (2013)
DOI: 10.1002/dac
ACCURATE ESTIMATION OF DOPPLER SHIFT
Solving Equation (7), we have the SNR-insensitive condition:
_
24 =
1

d
T
s
(8)
Because the sampling rate usually is invariant, a decimator analogous to [28] can be employed to
produce different sample intervals; then, Equation (8) should be rewritten as
_
24 =
1
M
d
T
s
(9)
where M denotes the decimating factor. In fact, the method in [28] can be considered as a specied
example of our study, and this paper further nds which decimating factor is optimal and presents
an effective iterative process to realize the optimal decimating factor.
3.3. Iterative process for rened estimation
To realize the optimal decimating factor, an iterative process is proposed and produces great per-
formance improvements compared with the conventional estimators. Note both this process and the
SNR insensitive condition are the main novelties of this paper.
After denoting

d
(i ) and M(i ) as the Doppler shift estimation and the computed decimating
factor at the i th iteration, we can illustrate the proposed iterative process as follows:
(1) At initialization, set a frequency difference threshold ^
th
and let {M(1) =1,

d
(0) =0, and
iteration counter 1
c
=1}.
(2) Estimate Doppler shift for current iteration and get

d
(1
c
) by Equation (4).
(3) Compute M(1
c
) =
j
1=
p
24
O
f
d
.I
c
/T
s
k
, where ]v denotes the oor function.
(4) If M(1
c
) = M(1
c
1) or [

d
(1
c
1)

d
(1
c
)[ < ^
th
, exit the iterative process; otherwise,
1
c
=1
c
1, go to step 2).
When the iterative process is nished, the nal Doppler shift estimation is expressed as

d
.
3.4. Convergency analysis
After dening j
n
as the bias ratio at the nth iteration, we can express the following recurve equation:
j
n
=
s
1
j
2
n1
1
1 ;
s
, n > 1 (10)
with the initial condition j
1
=
q
1
.1=2f
d
T
s
/
2
6
6.
s
C1/
> 1.
Solving Equation (10), the bias ratio at the N
I
th iteration can be rewritten as
j
2
n
1
j
2
n1
1
=
1
1 ;
s
=j
N
I
=
s
1
j
2
1
1
(1 ;
s
)
N
I
1
(11)
Looking at Equation (11), we can draw some benecial conclusions in the following:
v j
N
I
> 1 for any iteration number N
I
>1.
v For a given
d
T
s
, j
N
I
monotically decreases with N
I
increasing.
v If N
I
o, j
N
I
1.
v For a given j
N
I
, the iteration number varies with the actual
d
T
s
.
v For a given j
N
I
, different SNRs result in different iteration numbers.
v With the knowledge of SNR,
d
T
s
, and ^
th
, the iteration number can be calculated by
Equation (11) because j
N
I
=^
th
,
d
1.
Copyright 2013 John Wiley & Sons, Ltd. Int. J. Commun. Syst. (2013)
DOI: 10.1002/dac
J. Y. HUA ET AL.
So far, the convergency of the proposed iterative process is approved, and the improvement of its
performance is controlled by N
I
as shown in Equation (11). Hence, N
I
is an important parameter to
make a trade-off between complexity and performance, which is related to the ^
th
. In our research,
we found that ^
th
=10 Hz is an appropriate choice.
3.5. Further modication
Generally, the oor function exploited in the iterative process will cause some deviations from the
ideal estimation. Hence, an important task is to ensure the precision of

d
; here, we propose a simple
polynomial tting to overcome this problem. Given

d
s, we can derive the following process:
v For a certain SNR range (515 dB), set up the polynomial error expression by using Monte
Carlo simulations (1000 times in our study)
^
Doppler
=
3

3
d

2

2
d

1

0
(12)
where
3
~
0
is invariant for different SNRs; thus, a small bias will be remained after tting.
However, we nd in extensive simulations that this kind of performance loss will not cause
large bias.
v Calculate the improved estimation by

d
=(1 ^
Doppler
)

d
(13)
Using the aforementioned process, we can further improve the estimation for Doppler shift in
most scenarios. Note that both Equations (12) and (13) are determined by ofine simulations.
4. SIMULATIONS AND ANALYSIS
This section presents some simulation results in high-speed vehicular environments in terms of mean
square error (MSE) and accuracy. Then, we compare the performance of the proposed method with
those of the logarithmic envelope (LE) method [16], the ACF method [24], the iterative ACF method
[27], the LCR method [17], and the phase method [28].
The detailed simulation parameters are shown in Table I, where the total simulation duration is
1000 slots and the international telecommunications union (ITU) R. M.1225 Veh. B channel model
with six independent paths is used at the carrier of 2.11 GHz. Meanwhile, each slot consists of
1056 bits (bit rate 1.2288 Mbits/s). In each slot, ve pilot symbols, each of 32 bits, are time multi-
plexed with four data blocks and used for moving average channel estimation. Note that the M.1225
Veh. Bchannel model presented in ITUstandards is a widely used vehicular communication channel
for 3G/B3G systems; thus, our simulations reect the case in vehicular networks and therefore is
useful for V2I ITS systems.
Figure 2 illustrates the accuracy of some existing methods as well as the proposed method with
polynomial tting, where higher SNRs lead to smaller biases. This is consistent with the discussion
presented in Section 3. From Figure 2, we can clearly present that the LE method in [16] yields an
obvious bias and a distinguishable error oor in all scenarios. Furthermore, we demonstrate that the
proposed method yields similar performance as the iterative ACF method and outperforms the other
four methods. Additionally, the biases of the iterative-based method are small even if the SNR is
Table I. Simulation parameters.
Slot length 1056 bits Channel model ITU M.1225 Veh. B
Bit rate 1.2288 Mbit/s Simulation length 1000 slots
Pilot length 32 bits Path number 6
Carrier 2.11 GHz Data block length 224 bits
^
th
10 Hz Modulation QPSK
Copyright 2013 John Wiley & Sons, Ltd. Int. J. Commun. Syst. (2013)
DOI: 10.1002/dac
ACCURATE ESTIMATION OF DOPPLER SHIFT
50 100 150 200 250 300 350 400 450 500
0
100
200
300
400
500
600
700
Actual Doppler shift (Hz)
E
s
t
i
m
a
t
e
d

D
o
p
p
l
e
r

s
h
i
f
t

(
H
z
)
Ideal estimation
LE method
LCR method
ACF method
Phase method
Iterative ACF method
Proposed method with fitting
(a) SNR=0dB
50 100 150 200 250 300 350 400 450 500
0
100
200
300
400
500
600
700
Actual Doppler shift (Hz)
E
s
t
i
m
a
t
e
d

D
o
p
p
l
e
r

s
h
i
f
t

(
H
z
)
Ideal estimation
LE method
LCR method
ACF method
Phase method
Iterative ACF method
Proposed method with fitting
(b) SNR=5dB
50 100 150 200 250 300 350 400 450 500
0
100
200
300
400
500
600
700
Actual Doppler shift (Hz)
E
s
t
i
m
a
t
e
d

D
o
p
p
l
e
r

s
h
i
f
t

(
H
z
)
Ideal estimation
LE method
LCR method
ACF method
Phase method
Iterative ACF method
Proposed method with fitting
(c) SNR=10dB
50 100 150 200 250 300 350 400 450 500
0
100
200
300
400
500
600
Actual Doppler shift (Hz)
E
s
t
i
m
a
t
e
d

D
o
p
p
l
e
r

s
h
i
f
t

(
H
z
)
Ideal estimation
LE method
LCR method
ACF method
Phase method
Iterative ACF method
Proposed method with fitting
(d) SNR=15dB
Figure 2. Doppler shift estimation performance comparison II: the logarithmic envelope (LE) method [16],
the autocorrelation function (ACF) method [24], the level crossing rate (LCR) method [17], the phase method
[28], the iterative ACF method [27], and the proposed method.
0 dB, whereas the biases of the non-iterative methods are large in low SNR and at low speed. On the
other hand, although the proposed method does not outperforms the iterative method in Figure 2, in
real-world scenarios, the former is superior to the latter because of the larger estimation range of the
normalized Doppler shift.
To show the large estimation range of the proposed method, high-speed simulations are presented
in Figure 3, where the highest speed approaches 480 km/h (Doppler shift 938 Hz), and have not been
taken into consideration in conventional investigations. FromFigure 3, we explicitly see that the pro-
posed method produce precise estimates so long as SNR is larger than 5 dB, which is the working
SNR range for common communication systems. When SNR is 0 dB, the proposed method pro-
duces obvious but acceptable bias for high speeds. Moreover, Ki [5] presented that the ITS system
requires the estimation error less than 5%, and Jakus and Coe [6] produced the error about 3.3% at
90 km/h and 12.5% at 60 km/h. Obviously, if SNR > 5 dB, the proposed method performs better
than that in [6] and substantially fulls the requirements in [5]. In fact, at this time, the proposed
method produced the error about 2.73% at 90 km/h and 1.7% 60 km/h}, and the maximum error is
about 4.5%. Additionally, although not shown in Figure 3 for the sake of legible gure, we must
point out that compared with the proposed method, other ve methods, including the iterative ACF
method, will produce much larger biases at high speeds. Hence, only the proposed method is reliable
and suitable for Doppler shift estimation in high-speed scenarios.
Figure 4 compares the MSE of the tested methods averaged along speed dimension. Because the
conventional LCR method, the LE method, the ACF method, and the phase method cannot eliminate
Copyright 2013 John Wiley & Sons, Ltd. Int. J. Commun. Syst. (2013)
DOI: 10.1002/dac
J. Y. HUA ET AL.
0 100 200 300 400 500 600 700 800 900 1000
0
200
400
600
800
1000
1200
Actual Doppler shift (Hz)
E
s
t
i
m
a
t
e
d

D
o
p
p
l
e
r

s
h
i
f
t

(
H
z
)
Ideal estimation
SNR 0dB
SNR 5dB
SNR 10dB
SNR 15dB
Figure 3. The proposed estimator in high-speed scenarios. SNR, signal-to-noise ratio.
0 5 10 15
10
4
10
3
10
2
10
1
10
0
10
1
SNR (dB)
M
S
E
LE method
ACF method
Iterative ACF method
Phase method
LCR method
Proposed method without fitting
Proposed method with fitting
Figure 4. The mean square error (MSE) performance comparison of Doppler shift estimators. LE,
logarithmic envelope; ACF, autocorrelation function; LCR, level crossing rate; SNR, signal-to-noise ratio.
the effect of additive noise, they experience severe performance degradation. On the other hand, the
proposed method maintains low MSE and obtains at least one order of magnitude gain when SNR
falls into the range of 015 dB. Moreover, because the iterative estimation is larger and lesser than
the actual Doppler shift at an SNR of 0 dB and other higher SNRs, the tting operation causes a
little MSE increase at an SNR of 0 dB. Fortunately, at working SNR ranges, the proposed estimator
with polynomial tting can evidently observe performance improvements compared with the one
without tting. In addition, compared with the iterative ACF method, the proposed estimator with
polynomial tting produces similar and smoother MSEs at working SNR ranges while having larger
Doppler shift estimation ranges. Accordingly, we can conclude that the iterative technique plus poly-
nomial tting removes the AWGNinuence effectively and yields a great performance improvement
in a wide range of SNRs and velocities, which must be benecial for the ITS application requiring
accurate MT speed estimations.
Copyright 2013 John Wiley & Sons, Ltd. Int. J. Commun. Syst. (2013)
DOI: 10.1002/dac
ACCURATE ESTIMATION OF DOPPLER SHIFT
5. CONCLUSION
After investigating the noise corruptions, we propose a novel estimation algorithm for Doppler
shift in vehicular communications by applying the iterative technique and polynomial tting to
the conventional LCR method, where we study the SNR-insensitive condition and realize it through
an iterative process. With the help of the iteration hypostasis, we also analyze the convergency.
Both the simulation and the analysis show a good and consistent performance in a wide range
of velocities and SNRs, which will benet many communication applications, such as those in
3G and multi-input multi-output orthogonal frequency division multiplex (MIMO-OFDM) systems
[1719].
ACKNOWLEDGEMENTS
This paper is sponsored by the key project of Chinese ministry of education (grant no. 210087), Zhejiang
provincial NSF (grant no. LY12E07005), and the open research fund of National Mobile Communications
Research Laboratory, Southeast University (grant no. 2010D06).
REFERENCES
1. Al-Masud A, Mondal M, Ahmed KM. Vehicular communication system for vehicle safety using RFID. Proceedings
of the IEEE MICC09, Kuala Lumpur, Malaysia, 2009; 697702.
2. Bera R, Mondal D, Sil S, Dhar S, Sur S, Bhaskar D, Sarkar SK, Kandar D. Vehicular communication and safety in
realization of intelligent transport system. Proceedings of the CODEC09, Kolkata, India, 2009; 14.
3. Tan I, Tang W, Laberteaux K, Bahai A. Measurement and analysis of wireless channel impairments in DSRC
vehicular communications. Proceedings of the IEEE ICC08, Beijing, China, 2008; 48824888.
4. Kiokes G, Amditis A, Uzunoglu NK. Simulation-based performance analysis and improvement of orthogonal
frequency division multiplexing802.11p system for vehicular communications. IET Intelligent Transport Systems
2009; 3(4):429436.
5. Ki YK. Speed-measurement model utilising embedded triple-loop sensors. IET Intelligent Transport Systems 2011;
5(1):3137.
6. Jakus K, Coe DS. Speed measurement through analysis of the Doppler effect in vehicular noise. IEEE Transactions
on Vehicular Technology 1975; VT-24(3):3338.
7. Wang JZ, Zhu HL, Gomes NJ. Distributed antenna systems for mobile communications in high speed trains. IEEE
Journal on Selected Areas in Communications 2012; 30(4):675683.
8. Mousa A, Mahmoud H. Reducing ICI effect in OFDM system using low-complexity Kalman lter based on
comb-type pilots arrangement. International Journal of Communication Systems 2011; 24(1):5361.
9. Hafez H, Fahmy YA, Khairy MM. LTE and WiMAX: performance and complexity comparison for possible channel
estimation techniques. International Journal of Communication Systems Online rst, 2011. DOI: 10.1002/dac.1370.
10. Zhou YQ, Wang J, Sawahashi M. Downlink transmission of broadband OFCDM systemspart II: effect of Doppler
shift. IEEE Transactions on Communications 2006; 54(6):10971108.
11. Zhang ZY, Cheng P, Zhou XW, Qiu PL. System synchronization and channel estimation analysis for IEEE 802.16e
OFDMA downlink system. International Journal of Communication Systems 2009; 22(4):375398.
12. Karami E. Performance analysis of decision directed maximum likelihood MIMO channel tracking algorithm.
International Journal of Communication Systems Online rst, 2012. DOI: 10.1002/dac.2329.
13. Neelakantan PC, Babu AV. Computation of minimum transmit power for network connectivity in vehicular ad hoc
networks formed by vehicles with random communication range. International Journal of Communication Systems
Online rst, 2012. DOI: 10.1002/dac.2390.
14. Ozen A. A novel variable step size adjustment method based on channel output autocorrelation for the LMS training
algorithm. International Journal of Communication Systems 2011; 24(7):938949.
15. Monk AM, Miltein LB. Open-loop power control error in a land mobile satellite system. IEEE Journal on Selected
Areas in Communications 1995; 13(2):205212.
16. Sampath A, Holtzman J. Estimation of maximum Doppler frequency for handoff decisions. Proceedings of the IEEE
VTC93, Secaucus, New Jersey, 1993; 859862.
17. Ma Z, Yan Y, Zhao C, You X. An improved channel estimation algorithm based on estimating level crossing rate for
CDMA receiver. Chinese Journal of Electronics 2003; 12(2):235238.
18. Aboutorab N, Hardjawana W, Vucetic B. A new iterative Doppler-assisted channel estimation joint with parallel
ICI cancellation for high-mobility MIMO-OFDM systems. IEEE Transactions on Vehicular Technology 2012;
61(4):15771589.
19. Gao J, Zhu X, Wu Y. Kalman smoothing-based adaptive frequency-domain channel estimation for uplink
multiple-input multiple-output orthogonal frequency division multiple access systems. IET Communications 2011;
5(2):199208.
Copyright 2013 John Wiley & Sons, Ltd. Int. J. Commun. Syst. (2013)
DOI: 10.1002/dac
J. Y. HUA ET AL.
20. Hua JY, Meng LM, Xu ZJ, Li G. An adaptive signal-to-noise ratio estimator in mobile communication channels.
Digital Signal Processing 2010; 30(3):692698.
21. Heo J, Wang Y, Chang K. A novel two-step channel-prediction technique for supporting adaptive transmission in
OFDM/FDD system. IEEE Transactions on Vehicular Technology 2008; 57(1):188193.
22. He J, Xu C, Li L. Adaptive subcarrier bandwidth and power in OFDM-based cognitive radio systems for high
mobility applications. International Journal of Communication Systems Online rst, 2012. DOI: 10.1002/dac.2424.
23. Hammes U, Zoubir AM. Robust Mobile Terminal tracking in NLOS environments based on data association. IEEE
Transactions on Signal Processing 2010; 58(1):58725882.
24. Xiao C, Mann K, Olivier J. Mobile speed estimation for TDMA based hierarchical cellular systems. IEEE
Transactions on Vehicular Technology 2001; 50(4):981991.
25. Dogandzic A, Zhang B. Estimating Jakes Doppler power spectrum parameters using the Whittle approximation.
IEEE Transactions on Signal Processing 2005; 53(3):9871005.
26. Hua J, Sheng B, You X. The phase probability distribution of general Clarke model and its application in Doppler
shift estimation. IEEE Antennas and Wireless Propagation Letters 2005; 4(4):373377.
27. Hua J, Meng L, Li G, Wang D, You X. An accurate scheme for channel parameter estimation in mobile propagations.
IEICE Transactions on Electronics 2009; E92C(1):116120.
28. Hua J, You X. Ascheme for the Doppler shift estimation despite the power control in mobile communication Systems.
Proceedings of the IEEE VTC04 Spring, Milan, Italy, 2004; 284288.
29. van de Beek JJ, Sandell M, Borjesson PO. ML estimation of time and frequency offset in OFDM systems. IEEE
Transactions on Signal Processing 1997; 45(7):18001805.
30. Hua JY, Xu ZJ, Meng LM, Li G. A spectrum-efcient integer frequency offset estimator of mobile OFDM system
in double selective channels. International Journal of Communication Systems 2010; 23(8):10411056.
31. Tepedelenlioglu C, Giannakis GB. On velocity estimation and correlation properties of narrow-band mobile
communication channels. IEEE Transactions on Vehicular Technology 2001; 50(4):10391052.
32. Gao X, Jiang B, You X, Pan Z, Xue YS, Schulz E. Efcient channel estimation for MIMO single-carrier block
transmission with dual cyclic timeslot structure. IEEE Transactions on Communications 2007; 55(11):22102223.
33. Stuber GL. Principle of Mobil Communications. Kluwer Academic: Boston, MA, 2001.
AUTHORS BIOGRAPHIES
J. Y. Hua was born in Zhejiang province, China in 1978. He received his PhD degree of
Radio Engineering from Southeast university in 2006. Now he is an associate professor of
Zhejiang university of technology. His research interests lie in the area of channel parameter
estimation and multicarrier signal processing in mobile communication.
D. H. Yuan was born in Jiangsu province, China in 1987. She received his B.Sc degree of EE
from Huzhou normal university in 2010. Now she is pursuing her MSc degree in Zhejiang
university of technology. Her research interests lie in the area of parameter estimation and
wireless channel modeling.
Copyright 2013 John Wiley & Sons, Ltd. Int. J. Commun. Syst. (2013)
DOI: 10.1002/dac
ACCURATE ESTIMATION OF DOPPLER SHIFT
G. Li was born in Zhejiang province, China in 1983. He received his PhD degree of EE from
Nanjing University of science and technology in 2011. Now he is an assistant professor of
Zhejiang university of technology. His research interests lie in the area of signal propagation
and optical communication.
L. M. Meng was born in Zhejiang province, China in 1963. She received his PhD degree
of EE from Zhejiang university in 2003. Now she is a full professor of Zhejiang university
of technology. Her research interests lie in the area of WLAN and detection technique in
mobile communication.
Copyright 2013 John Wiley & Sons, Ltd. Int. J. Commun. Syst. (2013)
DOI: 10.1002/dac