Int. J. Electron. Commun.

(AEÜ) 63 (2009) 569 – 575

www.elsevier.de/aeue

A bandwidth-efficient method for cancellation of ICI in OFDM systems

Arvind Kumar∗ , Rajoo Pandey
Department of Electronics and Communication Engineering, National Institute of Technology (Deemed University), Kurukshetra 136119,
India

Received 18 September 2007; accepted 14 April 2008

Abstract
Orthogonal frequency division multiplexing (OFDM) is a very important modulation technique in wideband wireless com-
munication and multimedia communication systems. While, it can effectively deal with multipath delay spread produced by
frequency fading channels, its main drawback is the effect of frequency offset (FO) produced by the receiver local oscillator
or by motion-induced Doppler. The FO breaks the orthogonality among the subcarriers and hence causes intercarrier inter-
ference (ICI). In this paper, ICI caused by frequency drift is eliminated by equalizing the complex weighting coefficients
of interference. In most of the commonly used ICI cancellation schemes, bandwidth efficiency suffers because of the re-
quirement of redundancy in the transmission. In the proposed scheme, repetition of data symbols or transmission of training
sequence is not required. Thus, the bandwidth efficiency of normal OFDM system is maintained. The improved performance
of the present scheme is confirmed through extensive simulations.
䉷 2008 Elsevier GmbH. All rights reserved.

Keywords: OFDM; Intercarrier interference; Self-ICI cancellation; Interference coefficients equalization

1. Introduction In OFDM, the entire bandwidth is divided into N small

parts, and a block of N data symbols is modulated on N cor-
Orthogonal frequency division multiplexing (OFDM) is responding subcarriers, which are orthogonal to each other.
becoming the chosen multi-carrier modulation technique These sub-channels are transmitted in parallel, thereby in-
for wireless and multimedia communication systems. Mul- creasing the symbol duration and reducing the intersymbol
timedia wireless services require high-bit-rate transmission interference (ISI). In OFDM systems the effect of ISI can
over mobile radio channels [1,2]. OFDM can provide large be significantly reduced by cyclically extending the OFDM
data rates with sufficient robustness to radio channel im- symbol.
pairments. OFDM is a method that allows to transmit high Therefore, OFDM is a very attractive technique for the
data rates over extremely hostile channels at a comparable transmission of the high-bit-rate data. The major limita-
low complexity [3]. The OFDM has been used in wireless tion of the OFDM is that it is very sensitive to frequency
LAN standards such as American IEEE802.11a and Euro- errors caused by frequency differences between the local
pean equivalent HYPERLAN/2 and in multimedia wireless oscillators in the transmitter and receiver. The FO may
services such as Japanese Multimedia Mobile Access Com- also arise due to motion-induced Doppler. Since in an
munications. OFDM system a block of N data symbols is modulated
into N corresponding subcarriers, which are orthogonal to
∗ Corresponding author. Tel.: +91 1744233429. each other, the FO will destroy the orthogonality of the
E-mail addresses: [email protected] (A. Kumar), subcarriers and in turn induce the intercarrier interference
[email protected] (R. Pandey). and degrade the system performance significantly [4–7].
1434-8411/$ - see front matter 䉷 2008 Elsevier GmbH. All rights reserved.
doi:10.1016/j.aeue.2008.04.007
570 A. Kumar, R. Pandey / Int. J. Electron. Commun. (AEÜ) 63 (2009) 569 – 575

A number of methods, namely self-ICI cancellation schemes effect of noise as considered in [8], the demodulated signal
[8–11], discrete Fourier transform (DFT)-based ICI cancel- is given by
lation scheme [12], and channel estimation and frequency

N −1
domain equalization [13,14], have been developed to reduce
y(t) = e( j2 f t+0 ) xk,i q(t − kT /N ) (3)
this sensitivity to FO. The self-ICI cancellation scheme re-
k=0
quires repetition of input data which reduces the bandwidth
efficiency of the normal OFDM system. The DFT-based where q(t) is the combined impulse response of the channel,
scheme performs well only for a small FO. The channel es- transmitter, and receiver filters. 0 is the difference between
timation methods require pilot signals, which also limits the the phase of the receiver local oscillator and the carrier phase
throughput. at the start of the received symbol. As in [8], we assume
In the present work, a new bandwidth-efficient ICI can- that the q(t) satisfies the Nyquist criterion for samples taken
cellation scheme is presented. The proposed scheme cancels at intervals T /N . If y(t) is sampled at the optimum instant,
interference in OFDM systems by equalization of interfer- then the samples of y(t) are given by
ence coefficients without requiring any training data.
The paper is organized as follows. First we introduce yk,i = e( j 0 ) xk,i e( j2k /N ) (4)
the structure of the OFDM system in Section 2. In Section where  =  f T is the normalized FO.
3, the self-ICI cancellation schemes of [8,9] and the DFT- These samples are applied to the receiver DFT, after re-
based ICI cancellation scheme of [12] are briefly discussed moving the cyclic prefix. The output of receiver DFT is again
and analyzed. Then the proposed scheme is introduced in converted into serial form. The output of the DFT is given
Section 4. Section 5 presents the analysis of the performance by
of these schemes through simulation. The efficacy of all the
schemes considered in this study is judged on the basis of 
N −1
symbol error rate (SER) at different values of signal to noise Ym,i = yk,i e(− j2km/N ) (5)
ratio (SNR) and FO. Finally, the overall findings of the study k=0
are summarized in Section 6. Then from (4) and (5) it can be shown that
N −1
1 ( j 0 ) 
Ym,i = e X l,i
2. System model N
l=0

N −1
Fig. 1 shows the structure of the OFDM communica- × exp( j2k(l − m + )/N ) (6)
tion system. The N bits of high-speed input data signal
k=0
X 0 , X 1 , . . . , X N −1 modulate the N orthogonal subcarriers
to produce the output of the OFDM transmitter for the ith- After some simplification this can also be expressed as
transmitted symbol as −1

N
( j 0 )
−1 Ym,i = e cl−m X l,i (7)

N
( j2 f c t)
x(t) = e xk,i p(t − kT /N ) (1) l=0

k=0 where
N −1
where f c is the carrier frequency and p(t) is the impulse 1 
response of the low-pass filter in the transmitter and the cl−m = exp( j2k(l − m + )/N ) (8)
N
inverse discrete Fourier transform (IDFT) xk,i of X k,i , in the k=0
ith symbol, is given by
From (7), it can be seen that if  = 0 then Ym,i = e( j 0 ) X m,i
N −1 and each decoded value is simply the phase-rotated version
1  of the transmitted value. If   0, the ICI will occur and each
xk,i = X l,i exp( j2kl/N )
N output-decoded value will depend on all the input values.
l=0
for k = 0, 1, . . . , N − 1 (2) The decoded output Ym,i , therefore, consists of a wanted
component, which is due to X m,i but is subject to a change
If ts is the bit duration of the input, the OFDM symbol in the phase and amplitude given by
duration T will be equal to N t s . Thus the duration of the
N −1
OFDM symbol is increased by N times the duration of the 1 
c0 = exp( j2/N ) (9)
input data bit. This greatly reduces the effect of ISI [1]. N
k=0
To reduce the effect of ISI completely, the cyclic prefixing
of the OFDM symbol is done. At the receiver, the noisy Thus, the ICI depends on the N complex weighting coeffi-
signal received is mixed with a local oscillator signal having cients c0 , c1 , . . . , c N −1 . A plot of real and imaginary coeffi-
frequency  f above the correct frequency f c . Ignoring the cients of these weighting coefficients, for N = 16, is shown
 A. Kumar, R. Pandey / Int. J. Electron. Commun. (AEÜ) 63 (2009) 569 – 575 571

exp( j 2π f c t )

X 0, i x0, i
High speed
• •
data Serial to
•
N-point
•
Parallel to DAC ×
Parallel IDFT Serial and
converter • • Converter LPF
X N −1, i xN −1, i

h(t)

exp(− j 2π ( f c + Δf )t )
n(t)
+
yo , i
Y0, i
•
• N point
•
Serial to LPF ×
• DFT Parallel and BPF
• • Converter ADC
YN −1, i
yN −1, i

Fig. 1. OFDM system model.

1 3. ICI cancellation schemes

0.8 real
imaginary The ICI cancellation schemes, which do not require chan-
Weighting coefficients

0.6 nel estimation, are briefly revised here.

0.4
3.1. Self-ICI cancellation scheme
0.2

0 This scheme describes a method of reducing sensitivity to

frequency errors. In this scheme the input data are mapped
-0.2
onto adjacent pair of subcarriers rather than onto a single
-0.4 subcarrier, such that X o,i =−X 1,i , X 2,i =−X 3,i , . . . , X N −2 =
-15 -10 -5 0 5 10 15 −X N −1,i as proposed in [8]. The decoded value of the ith
Number of subcarriers
carrier is expressed as a function of weighting coefficients as
Fig. 2. Weighting coefficients for N = 16 and  f T = 0.2. Y0,i = exp( j0 ){(c0 − c1 )X 0,i + (c2 − c3 )X 2,i
+ · · · + (c N −2 − c N −1 )X N −2,i } (12)
in Fig. 2. It can be observed from this figure that the variation
in the weighting coefficients is smooth as the distance moves The ICI now depends on the difference between the adja-
from −15 to +15, except between −1 and 0 and between 0 cent weighting coefficients rather than on the coefficients
and 1. The main component of the symbol corresponds to themselves. As the difference between adjacent coefficients
the weighting coefficient at zero. is small, this results in a substantial reduction in ICI. In this
The received signal can also be written in the matrix form, process the component of ICI, which is constant between an
using (7) as adjacent pair of subcarriers, is cancelled out. This scheme,
termed as constant ICI cancellation, improves the perfor-
Y = CX (10) mance at any FO, because the ICI cancellation depends only
on the coefficients being a slowly varying function of offset.
where C is given by
To maximize the overall SNR, the values of output sym-
⎡ c c · · c N −1 ⎤
0 1 bols can be subtracted in pairs, resulting in further reduction
⎢ c−1 c0 · · c N −2 ⎥ of ICI:
⎢ ⎥
C = ⎢ c−2 c−1 c0 · c N −3 ⎥ (11)
⎣ · · · · · ⎦ Y0,i − Y1,i = exp( j0 ){(−c−1 +2c0 −c1 )X 0,i
c−N +1 c−N +2 · · c0 + · · · +(−c N −3 +2c N −2 −c N −1 )X N −2,i } (13)

where X = [X 0,i , . . . , X N −1,i ]T is the input data and In this process the remaining ICI depends on the factors of
Y = [Y0,i , . . . , Y N −1,i ] is the received data. the form {−c1 + 2c2 − c3 }. If the three weighting coeffi-
572 A. Kumar, R. Pandey / Int. J. Electron. Commun. (AEÜ) 63 (2009) 569 – 575

cients in each term were linear functions of offset, the ICI 1

would completely be zero. This is termed as the linear ICI real
0.8 imaginary
cancellation scheme [8].

Weighting coefficients
0.6
3.2. DFT-based ICI cancellation scheme 0.4

In this scheme, the input data symbols, after serial to par- 0.2
allel conversion, are divided into two groups, each of size
0
N /2. These groups are separately discrete Fourier trans-
formed; after this, the two groups are again combined and -0.2
then N point IDFT is taken. At the receiver side, after de-
modulation, the N point DFT of the received data symbols is -0.4
-15 -10 -5 0 5 10 15
taken and after this the symbols are divided into two groups Number of subcarriers
again and each inverse-discrete Fourier transformed, respec-
tively, to rebuild the transmitted symbols. It has been shown Fig. 4. Weighting coefficients of matrix C(1) for N = 16 and
in [12] that by doing this ICI is greatly reduced.  f T = 0.2.
For the DFT-based scheme the transmitted data can be
described in matrix form as
4. Proposed scheme
1
x = √ WHN BX (14)
N As discussed in the previous section, the ICI depends
on the weighting coefficients co , c1 , . . . , c N −1 . In order to
where WH N is the N × N IDFT matrix, B is the N × N design an equalizer for equalizing the weighting coefficients
matrix obtained after combining the two N /2 DFT blocks, that cause ICI, the matrix of these coefficients is obtained
X = [X 0,i , X 1,i , . . . , X N −1,i ]T , and H denotes the Hermitian from (11). The received signal at the receiver in the matrix
transpose. The data received are described in matrix form as form is given by
Y = C(1) X (15) Y = W N KW−1
N X

where C(1) = 1/N BH W N KWH j2/N , or Y = CX where W N is an N × N DFT matrix and W−1
N B and K = diag(1, e N
e j4 /N , ... , e j2(N −1)/N ). is an N × N IFDT matrix.
A plot of interfering coefficients of matrix C(1) is shown in When an equalizer is used, the output is given by
Figs. 3 and 4 for N =16 and at =0.1&0.2 respectively. Fig. Z = DY (16)
3 shows that the value of the coefficient corresponding to
the main symbol is significant compared to other interfering where D = C−1 is a circulant matrix, which denotes the
coefficients at small FOs, for   0.1. It can also be noted that equalizer filter matrix.
the coefficient corresponding to the main symbol decreases Commonly, the equalizer is designed by using training
and the interfering coefficients increase at  = 0.2. Thus the symbols. In the present scheme, however, an appropriate
scheme is effective only at small FOs. filter is chosen from a small set of predefined filters. This
is based on the observation that an equalizer designed for a
given value of  is effective for a variation ± around .
1
If the maximum expected value of normalized FO is max ,
real
then the given range of  can be divided into p segments,
imaginary
0.8 where p = max /2. Let the midpoint of these segments
Weighting coefficients

be denoted by (i) i = 1, 2, . . . , p. Now the equalizers D(i)

0.6 for each (i) i = 1, 2, . . . , p can be obtained from (16). The
appropriate equalizer from this set of predefined equalizers
0.4 D(i) , i =1, 2, . . . , p is selected according to the the following
criterion:
0.2
minimize{J (D(i) ) = E[|Z(i) − g(Z(i) )|2 ]} (17)
i
0
−15 −10 −5 0 5 10 15 with Z(i) =D(i) Y where g(·) is a nonlinear estimator of trans-
Number of subcarriers mitted symbols and E{·} denotes the expectation.

Fig. 3. Weighting coefficients of matrix C(1) for N = 16 and Theorem. If the objective function of (17) is to be used
 f T = 0.1. for selection of the equalizer, then the nonlinear estimator
 A. Kumar, R. Pandey / Int. J. Electron. Commun. (AEÜ) 63 (2009) 569 – 575 573

3 shown in Fig. 5. With this approximation the criterion for

selection of the equalizer, now becomes
2 Bayesian Estimator
Approximate Estimator minimize{J (D(i) ) = E[|Z(i) − Z(i)
s | ]}
2
(22)
i
1
(i)
where Zs denotes the decision on Z(i) .
g(z)

4.1. Complexity comparison

-1

The complexity of the proposed scheme and the conven-

-2
tional equalizer with a training sequence using the LMS al-
gorithm is shown in the following Table 1.
-3
-4 -3 -2 -1 0 1 2 3 4 Alternatively, in place of the cost function defined in (17),
z another cost function such as the CMA cost function [15]
can also be used. However, our simulation has shown that
Fig. 5. Plot of nonlinear Bayes estimator g(Z ) and its approxima- the cost function of (17) requires fewer samples to select
tion. the optimum equalizer in comparison to the CMA cost func-
tion. The plot of the proposed cost function and CMA cost
g(·) will depend upon the probability density functions of function is shown in Fig. 6.
transmitted symbols and convolution noise. In the proposed scheme the cost function is merely used
for the selection of the appropriate equalizer filter and is not
Proof. Using the Bussgang principle of blind deconvolution, minimized iteratively. It can be noted that the cost function
the equalizer must minimize an objective function of the is required to be differentiable when it is used for designing
type: J (n) = E[e2 (n)], as shown in [15], where the error is the equalizer using some higher-order statistics (HOS)-
(i) (i)
given by e(n) = (g(Z k ) − Z k ); k = 0, 1, . . . , N − 1, and based blind equalization scheme such as CMA. These
g(·) is an estimator of the transmitted symbol. The Bayes
estimate of transmitted symbol X k is given by Table 1. Comparison of complexity
∞ Operations Proposed Equalizer designed with
X̂ k = E[X k /Z k ] = X k f X (X k /Z k ) dX k (18) scheme a training sequence using
−∞
the LMS algorithm
for k = 0, 1, . . . , N − 1, where f X (X k /Z k ) denotes the con- Number of (N 2 + N − 1) p (N 2 + N − 1) per iteration
ditional probability density function (pdf) of X , given Z . complex additions
The observation Z k is expressed as Z k = 0 X k + k , where Number of (N 2 + N + 1) p (N 2 + N + 1) per iteration
k is the convolutional noise. In accordance with the statis- complex
tical model for the convolutional noise , X k , k are statis- multiplications
tically independent. With  modeled to have zero mean √ and Number of N –
variance 2 , the scaling factor 0 is given by 0 = 1 − 2 . nonlinear function
Using the Bayes’ rule, (18) becomes
∞
1
X̂ k = X k f Z (Z k / X k ) f X (X k ) dX k (19) 2.5
f Z (Z k ) −∞
Furthermore, from Z k = 0 X k + k , it follows that CMA
2
Proposed
f Z (Z k / X k ) = f  (Z k − 0 X k ) (20)
1.5
where f  (Z k − 0 X k ) is the pdf of convolutional noise.
|J|

By using (19) and (20) we get the estimate of the trans-

1
mitted symbol as
1 0.5
X̂ k = g(Z k ) =
f Z (Z k )
∞
× X k f  (Z k − 0 X k ) f X (X k ) dX k (21) 0
0 0.1 0.2 0.3 0.4 0.5
−∞
Normalized frequency offset
For the uniformly distributed symbols X, the plot of esti-
mator g(·) and its approximation used in the simulation is Fig. 6. CMA and proposed cost function.
574 A. Kumar, R. Pandey / Int. J. Electron. Commun. (AEÜ) 63 (2009) 569 – 575

schemes usually suffer from slow convergence and are not 100
suitable for the present case.
10-1

Symbol error rate

5. Simulation and results
10-2 stnd-ofdm

To evaluate the performance of various ICI cancellation self-ici-ofdm

schemes, considered in this paper, an OFDM system is sim- 10-3 DFT-ofdm
ulated to implement these schemes. In the proposed scheme equl-ofdm
the appropriate equalizer is selected from a set of four filters 10 -4

designed at (1) = 0.05, (2) = 0.1, (2) = 0.15 and (3) = 0.2. 0 5 10 15 20 25
SNR (dB)
For better accuracy a lager set of filters may be considered.
The performance of various schemes of ICI cancellation is Fig. 8. SER of standard, self-ICI cancellation, DFT-based ICI
judged on the basis of the SER. In our simulation of the cancellation, and equalized OFDM at  f T = 0.1.
OFDM system, QPSK modulation with 64 subcarriers is
considered. We assume the ideal reception of OFDM signals
in the self-ICI cancellation scheme, and DFT-based ICI can- 100
cellation scheme as used in [8] and [12], respectively. For
the computation of SER curves, the transmission of 10 000
10-1

Symbol error rate

symbols is considered.
The SER for various ICI cancellation schemes and stan-
dard OFDM is observed. The SER of standard OFDM, DFT- 10-2
based OFDM, self-ICI cancellation scheme, and the pro- stnd-ofdm
posed scheme is plotted in Fig. 7. This figure shows that -3 self-ici-ofdm
10
the SER of standard OFDM, self-ICI cancellation, and DFT- DFT-ofdm
based schemes is comparable at =0.05, while it is lower for equl-ofdm
the proposed scheme. As  increases the SER of equalized 10-4
0 5 10 15 20 25
OFDM, DFT-based, and the self-ICI cancellation scheme
SNR (dB)
improves, as shown in Fig. 8. The improvement in SER
of the proposed scheme is better than that of the standard Fig. 9. SER of standard, self-ICI cancellation, DFT-based ICI
OFDM, self-ICI cancellation scheme, and DFT-based ICI cancellation, and equalized OFDM at  f T = 0.15.
cancellation scheme. Figs. 9 and 10 show the SER for all
these schemes at a higher value of . From Figs. 9 and 10,
it is clear that the SER of the self-ICI cancellation scheme 100
is better than the DFT-based ICI cancellation scheme. The
main drawback of the self-ICI cancellation scheme is its
10-1
Symbol error rate

throughput, which is just half as compared to standard and

DFT-based ICI cancellation schemes. It can also be observed
that the SER of the DFT-based ICI cancellation scheme is 10-2
stnd-ofdm
self-ici-ofdm
10-3
100 DFT-ofdm
equl-ofdm
10-4
10-1 0 5 10 15 20 25
Symbol error rate

SNR (dB)

10-2
Fig. 10. SER of standard, self-ICI cancellation, DFT-based ICI
stnd-ofdm cancellation, and equalized OFDM at  f T = 0.2.
10-3 self-ici-ofdm
DFT-ofdm
equl-ofdm
better than the standard OFDM only for low values of .
10-4 As  increases the SER of the DFT-based ICI cancellation
0 5 10 15 20 25
SNR(dB) scheme increases and becomes comparable to the SER of
the standard OFDM.
Fig. 7. SER of standard, self-ICI cancellation, DFT-based ICI It is revealed by Figs. 7–10 that the present scheme
cancellation, and equalized OFDM  f T = 0.05. has better SER performance as compared to the standard
 A. Kumar, R. Pandey / Int. J. Electron. Commun. (AEÜ) 63 (2009) 569 – 575 575

OFDM, self-ICI cancellation and DFT-based ICI cancel- the consequences and solutions, In: IEEE 46th vehicular
lation schemes at all FOs. Also, the spectral efficiency of technology conference, Atlanta, GA, April 1996. p. 1564–8.
present scheme is maintained as that of the standard or the [8] Armstrong J. Analysis of new and existing methods of
normal OFDM system. reducing intercarrier interference due to carrier frequency
offset in OFDM. IEEE Trans Commun 1999;47:365–9.
[9] Yuping Zhao, Sven-Gustav Häggman. Intercarrier inter-
ference self-cancellation scheme for OFDM mobile commu-
6. Conclusions nication systems. IEEE Trans Commun 2001;49:1185–91.
[10] Wang Yonggang, Zhu Shibua, Li Ying. A method to improve
A new scheme based on equalization of interference co- the bandwidth efficiency of self-ICI cancellation in OFDM
efficients for cancellation of ICI in an OFDM system is systems. In: International conference on signal processing,
presented. The proposed scheme does not need any pilot (ICSP’02), vol. 2, August 2002, p. 1633–6.
symbols for equalization, maintaining the high spectral ef- [11] Sathananthan K, Tellambura C. New ICI reduction schemes
ficiency of a standard OFDM system. A comparison with for OFDM system. In: Proceedings of the 54th IEEE VTC’01,
other popular methods of cancellation of ICI shows that the Atlantic City, NJ, vol. 2, 2001. p. 834–8.
proposed scheme outperforms the other methods at all FOs. [12] Bing Han, Xiqi Gao, et al. A DFT-based ICI self-cancellation
The self-ICI cancellation scheme of [8] results in a low SER scheme for OFDM systems, communication technology
Proceedings (ICCT), vol. 2, April 2003. p. 1359–62.
as compared to the standard OFDM and DFT-based ICI can-
[13] Yücek T, Arslan H. ICI cancellation based channel estimation
cellation scheme, but its throughput is just half as compared
for OFDM systems. In: Proceedings of the IEEE radio and
to the DFT-based scheme of [12] and the proposed scheme. wireless conference, Boston, MA, August 2003. p. 111–4.
On the other hand, the DFT-based scheme yields a low SER [14] Yücek T, Arslan H. ICI cancellation based channel estimation
only at low values of FO, and is therefore better than the for OFDM systems. Letter IEICE Trans Commun 2005;E88-
standard OFDM at low values of FO. However, at a high B:2200–1.
value of FO the performance of the DFT-based ICI cancella- [15] Richard Johnson C. et al. Blind equalization using
tion scheme deteriorates over the standard OFDM. The SER the constant modulus criterion: a review. IEEE Proc
resulting from the proposed method is very low as com- 1998;86:1927–50.
pared to the standard OFDM, self-ICI cancellation scheme,
and DFT-based ICI cancellation scheme at any value of FO.
Arvind Kumar was born in India in
Also, in the present scheme repetition of data is not required, 1970 and received his B.Sc. (Maths),
and hence its spectral efficiency is maintained as that of the B.Tech. (Electronics and Telecommuni-
standard OFDM. cation Engineering), and M.E. (Control
and Instrumentation) degrees in 1992,
1996, and 1999 from Rohilkhand Uni-
References versity, Bareilly, Allahabad University,
Allahabad, and N.I.T. Allahabad, re-
[1] Van Nee R, Prasad R. OFDM for wireless multimedia spectively. Presently, he is a Lecturer in
communications. London: Artech House Publishers; 2000. the Department of of Electronics and
[2] Bahai ARS, Saltzberg BR. Multi-carrier digital Communication Engineering at N.I.T Kurukshetra. His research
communications: theory and applications of OFDM. New interests include signal processing and wireless communication
York: Kluwer Academic/Plenum; 1999. systems.
[3] Cimini LJ. Analysis and simulation of a digital mobile
channel using orthogonal frequency division multiplexing.
IEEE Trans Commun 1985;33:665–75. Rajoo Pandey was born in India in
[4] Pollet T, Van Bladel M, Moeneciaey M. BER sensitivity of 1968 and received his B.E., B.Tech.,
OFDM system to carrier frequency offset and Wiener phase and Ph.D. degrees in Electronics and
noise. IEEE Trans Commun 1993;43:191–3. Communication Engineering in 1989,
[5] Moose P. A Technique for orthogonal frequency division 1991, and 2001 from Government En-
multiplexing offset correction. IEEE Trans Commun gineering College, Jabalpur, R.E.C.
1994;42:2908–14. Kurukshetra, and Indian Institute of
[6] Russell M, Stüber GL. Interchannel interference analysis of Technology, Roorkee, India, respec-
OFDM in a mobile environment. In: Proceedings of IEEE tively. He is an Assistant Professor
vehicular technology conference, vol. 2, Chicago, IL, July in the Department of Electronics and
1995. p. 820–4. Communication Engineering at N.I.T Kurukshetra. His research
[7] Yuping Zhao, Sven-Gustav Haggman. Sensitivity to Doppler interests include signal processing, communication systems, and
shift and carrier frequency errors in OFDM systems: neural networks.