The 28th Wireless and Optical Communication Conference (WOCC 2019)

Universal Filtered OFDM Index Modulation System

with Enhanced ICI compression
Chaoyue Jin, Li Wei, Youyun Xu
National Engineering Research Center of Communications & Networking
Nanjing University of Posts and Telecommunications, Nanjing 210003, China
Email: {1017010316, weili , yyxu}@njupt.edu.cn

Abstract—Index modulation is a novel modulation technique The simulation results [9] indicate that the OFDM-IM is a
for OFDM system. In this letter, we propose a novel universal potential technique of ICI reduction by using less number of
filtered OFDM index modulation (UF-OFDM-IM) system to over- active subcarriers.
come the inter-carrier interference (ICI) problem in high-speed
environment. In the proposed scheme, the filtering operation is However, the OFDM-IM is still sensitive to the CFO [10],
used to reduce the out-of-subblock power leakage to minimize because the ICI will increase the received signal power of each
the ICI between the adjacent subblocks. The simulation results inactive subcarriers and decrease the power of each active sub-
indicate that the proposed scheme outperforms the OFDM-IM carriers. Then the above effects result in the error detection of
in the carrier frequency offset (CFO) compression.
the subcarriers index and the error demodulation of transmited
Index Terms—carrier frequency offset, bit error rate, index
modulation, universal filtered OFDM symbols. In [11], the ICI of the OFDM-IM caused by doppler
shift is cancelled by a two-path cancellation with symmetric
I. I NTRODUCTION mapping and mirror mapping. The filter bank multicarrier
(FBMC) system is combined with index modulation, which
Nowadays, China’s high-speed railway (HSR) has entered can provide the suppression of the out-of-band emission than
a new era. Because of the high-speed movement, one of the OFDM system [12].
most important challenges in HSR communications is the Inspired by the new waveforms of the FBMC system with
large Doppler shift, which will cause fast time-variation of index modulation, a filtering operation is adopted to decrease
fading channels. The fast time-variation brings the difficulty of the out of subblock leakage in this paper. It can reduce
estimating, tracking and predicting fading coefficients [1][2]. the intra-subblocks interference to make the system have
The Doppler shift created by the movement around the robustness against the ICI, which is different from using the
channel will result in carrier frequency offset (CFO). The selection of the filtering combinations from the filter bank to
OFDM signal is sensitive to frequency offset, even the small carry information in the paper [13]. Simulation results clearly
CFO makes serious ICI. Doppler estimation and compensation show that the proposed system obtains a much lower bit error
are difficult in high-speed mobile environment, e.g., real-time rate (BER) than that of OFDM-IM system.
Doppler estimation requires more pilot resources. Previous The rest of paper is organized as follows. Section II de-
studies have shown that the ICI induced by Doppler shift can scribes the OFDM-IM system model with the CFO. According
be reduced by decreasing the number of active subcarriers in to the theoretical results obtained in section II, a universal
OFDM systems. For any given CFO, the power of ICI is relat- filtered OFDM index modulation system is proposed in section
ed to the number of active subcarriers. If the active subcarriers III. The simulation results and analyses are given in section
are simply reduced, the spectral efficiency will be lost due to IV. Finally, section V concludes the paper.
nothing being transmitted on the inactivated subcarriers. In
order to reduce the number of the active subcarriers without II. S YSTEM M ODEL
reducing spectral efficiency, the index modulation for OFDM
(OFDM-IM) system has been proposed [3]. The same as the In the index modulation, each block contains N subcarriers
spatial modulation(SM) concept [4]-[6], OFDM-IM adopts the and b bits. Then, each block is split into subblocks, which
index of subcarriers to carry the information and to decrease contains D subcarriers and p bits. We assume that the number
the ICI with a few active subcarriers. of the subblocks is Z, which means b = p×Z and N = Z ×D.
Recently, there are many studies in this field [7]-[9]. The The modulation of OFDM-IM is not as similar as classical
OFDM-IM system based on discrete cosine transform (DCT) OFDM, the information consists of the indices of the subcar-
is proposed to enhance the spectral efficiency [7]. In paper riers and the constellation modulated symbols. The first p1 bits
[8], the simulation results have suggested that the Interleaved of the incoming p bits sequence are sent to the index selector.
subcarrier-index modulation improves the ability of resisting The selected indices are depended by the look-up table, which
ICI in high mobility environment. In order to analyze the sets the inactive subcarriers to zero. The active subcarriers
performance accurately, the ICI is properly divided into two carry the remaining p2 bits of this sequence mapped by the
parts: the intra-subblock ICI and the inter-subblocks ICI [9]. M-ary signal constellation. Obviously, p = p1 + p2 .

978-1-7281-0660-1/19/$31.00 ©2019 IEEE

 The 28th Wireless and Optical Communication Conference (WOCC 2019)

For each of the subblocks, the p1 information bits car- and m(r) is the set of the active subcarriers in the r-th
ried by active indices out of available indices are given subblock. There is a hypothesis in [9], Xi (z) and Xj (z)
by p1 = log2 C(D, v), where x means the floor function, represent the transmitted subblock and the erroneously de-
C(n, k) denotes the binomial coefficient. The remaining bits tected subblock respectively. There are two types of error
carried by v active subcarriers are mapped by M-ary signal detections from Xi (z) to Xj (z), which are different in one
constellation, we have p2 = v (log2 (M)).The vector of the active subcarrier or in one constellation symbol. We assume
modulation symbols which carries p2 bits is given by that α and β mean different subcarrier position of Xi (z) and
Xj (z), and different constellation symbol Sz (γ) and Ŝz (γ) in
Sz = [Sz (1), ..., Sz (i), ..., Sz (v)] (1) the γ-th active subcarrier.
Here, Sz (i) means the constellation symbol for z = 1, ..., Z Hence, the average bit error probability (ABEP) of the z-th
and i = 1, ..., v, i.e.. After the index selector, the subcarrier subblock is [14]
position in the indices is the constellation symbol and the re-
1 1  
D−1 D−1
maining subcarriers are zero which means inactive subcarriers.
Pb,z ≈ { P (Xi (z) → Xj (z)|α → β)[
TABLE I
p D α=0
β=0
 + 
Xi (z) → Xj + (z)2 ] + log2 M }+
LOOK - UP TABLE

Bits indices subblocks

2
log2 M 1 
D−1
[0 0] {1, 2} [Sz (1) Sz (2) 0 0] { P (Xi (z) → Xj (z)|Sz (γ) → Ŝz (γ))} (7)
2 D γ=0
[0 1] {2, 3} [0 Sz (1) Sz (2) 0]
[1 0] {3, 4} [0 0 Sz (1) Sz (2)] The P (Xi (z) → Xj (z)|α → β) and P (Xi (z) →
[1 1] {1, 4} [Sz (1) 0 0 Sz (2)] Xj (z)|Sz (γ) → Ŝz (γ)) are Q function, which is related to
the values from the formula (5)(6) [9].
The z-th subblock could be selected arbitrarily, so the ABEP
Example: Let us assume D = 4 and v = 2 and P1 = 2. of the whole OFDM-IM system can be expressed as
Every subblock has only 2log2 C(4,2) = 4 of C(4, 2) = 6
1 
Z−1
available indices. If the p1 bits are [1 0], the 3rd and 4th
P̄b = Pb,z (8)
subcarrier are selected. Then the remaining p2 bits are mapped Z z=o
by M-ary signal constellation to the two active carriers in
III. I NDEX M ODULATED F ILTERD -OFDM WITH ICI
sequential order.
The received signal of OFDM-IM in the frequency domain From the formula (7) we know the bad impact of ICI of the
can be expressed as inter-subblocks and the intra-subblocks on the BER. Then, a
subblock filter is taken into consideration to reduce the power

N −1
leakage outside the subblock. Fig. 1 demonstrates that the
R(k) = C(0)X(k)H(k) + C(τ − k)X(τ )H(τ ) + W(k)
subblock filter have a effect on the power leakage reduction
τ =0,τ =k
(2) [15]. A system of the UF-OFDM-IM is depicted in Fig. 2, and
where the transmitted OFDM-IM signal is its subblocks are processed by the subblock filter.
20

X = [X(0), X(1), ...X(k), ..., X(N − 1)] (3) 10

OFDM resource block with a subblock filter
OFDM resource block

and W(k) represents the frequency-domain Gaussain noise.

0

-10

The second term in (2) stands for the ICI caused by the CFO.
PSD (dBW/Hz)

-20

The C(k) denotes the ICI coefficient given by -30

sin π (εd + k) 1
C(k) = exp[jπ(1 − ) (εd + k)]
-40

(4)
N sin N
π
(εd + k) N -50

-60
-0.1 -0.05 0 0.05 0.1 0.15

Normalized frequency
here, εd is a normalized CFO.
Then, we consider the ICI from two parts [9]: the inter- Fig. 1. A subblock with a subblock filter and without a subblock filter.
subblocks ICI:

N −1 Since we focus on how to derive the transmitted signal,
ICI1,k = C(τ − k)X(τ )H(τ ) (5) the channel is assumed to be perfect. After index modulation,
τ =0,τ =k,τ ∈m(r)
/ every subblock is disposed by the N-point inverse FFT. For
an arbitrary subblock with the index i, the time domain signal
and the intra-subblocks ICI:
xi is converted by the frequency domain signal Xi [16]:

v−1
1  j2πkl
ICI2,k (r) = C(τ − k)X(τ )H(τ ) (6) xi (l) = Xi (k) exp{ } l = 0, .., N − 1 (9)
τ =0,τ =k,τ ∈m(r)
N N
k∈Bi
 The 28th Wireless and Optical Communication Conference (WOCC 2019)

where Bi contains the subcarrier indexes in the subblock i. The and Iz = {iz,1 , ..., iz,v } is the selected indices, where iz,η ∈
effect of filtering is equivalent to a discrete linear convolution [1, ..., D] for z = 1, ..., Z and η = 1, ..., v. Furthermore,
between the time domain signal xi and the filter f i : Rz (iz,η ) and Fz (iz,η ) represent the received frequency domain
signal and the corresponding filter frequency domain coeffi-

Z 
Z
cient on the η-th subcarrier of the z-th subblock, respectively.
y = yi = xi ∗ f i (10)
H(iz,η ) is the corresponding channel frequency domain coef-
i=1 i=1
ficient.
where f i = {fi (0), fi (1), ..., fi (LF − 1)} means tap coefficients
of the i-th subblock filter. LF is the length of the filter. P1 Index
N
bits Selector . Filter
Furthermore,
LF −1 we assume that the filter coefficients satisfy P
bits .
point
IFFT

2 P2 Mapper .

l=0 |f i (l)| =1. .

.
bits

When passing through a multipath Rayleigh fading channel, .

.
Add
the l-th sample of the symbol in this system is expressed as: Bit
. P1
bits
Index
Selector N
Filter
Cyclic
Prefix
Splitter P . point
bits IFFT
P2 .
Mapper
.
r(l) = h(l) ∗ y(l) + ω(l), l = 0, 1, ..., N − 1
bits
(11)
Index
The multipath Rayleigh fading channel can be repre- P
P1
bits Selector . N
point
IFFT
Filter

bits .
Channel
sented by the channel impulse response (CIR) coefficients P2
bits
Mapper .

{h(0), h(1), .., h(Lh − 1)}, and Lh is the length of the channel, Index modulation subblock
which is assumed to be shorter than the length of CP Lcp .
Since the length of the received signal is no longer N after . Time Remove
Domain Cyclic

the filtering operation, the length of the time domain signal .

.
Processing Prefix

Index Frequency 2N
is increased to 2N by the zeropadding operation, so that the Modulation
Demodulator
domain
Symbol
.
.
point
FFT
0
processing . 0
2N point DFT is used to obtain R(k) which is the frequency . . 0
. .
domain representation of r(l): . . 0
0


Z Fig. 2. UF-OFDM-IM system.
R(k) = H(k) X̃i (k)Fi (k) + W (k) k = 0, ..., 2N − 1
i=1
(12) IV. S IMULATION R ESULTS
where X̃i and Fi are 2N-point FFT of xi and f i . Considering In this section, the performance of the proposed UF-OFDM-
the separation of the signal from the interference part in (12), IM system and the OFDM-IM system are evaluated via
the X̃i can be derived by [16] comparing the BER in the case of the same CFO. Monte-
⎧ Carlo simulation is applied to estimate the BER.
⎪
⎪ Xi ( k2 ), if k is even We consider the simulations for D = 4, v = 2, N = 64,
⎨  sin( π (2m−k))
X̃i (k) = Xi (m) N sin(2 π (2m−k)) M = 2, and the symbols are chosen from a BPSK constellation
⎪
⎪ m∈Bi 2N
with unit symbol power. For the UF-OFDM-IM system in this
⎩
. exp{j π2 (2m − k)(1 − N1 )},
if k is odd paper, the Chebyshev windows is picked out. The sidelobe
(13) attenuation of the Chebyshev windows is 60dB and the center
where Xi is the frequency domain signal, which is in the i-th frequency is set at the middle of the every subblock.
subblock: For the simulation results of the Fig. 3 and Fig. 4, we
assume that each system is in such a scenario where the signal
0, if k ∈
/ Bi to noise ratio or the CFO is the same. Fig. 3 demonstrates the
Xi (k) = (14)
Xi (k), if k ∈ Bi BER performance of the proposed scheme and the OFDM-
Due to the 2N-point FFT detection, the original signal IM system in the same CFO. In this simulation, the CFOs
information is contained in the even subcarriers out of the are set as 0.1 and 0.05, respectively. As is shown in the
2N subcarriers. figure, the CFO causes energy leakage between subcarriers,
In order to correctly detect the pre-filtered data, a detection which will result in the destruction of orthogonality and make
based on subblocks is performed. The pre-filtered data consists the performance of the system worse. Therefore, the BER of
of the index of the subcarriers and the modulation symbols. UF-OFDM-IM system with filtering effect is better than that
Then, the ML detector algorithm is adopted to consider all of OFDM-IM system. When CFO increases, the interference
the possible subcarriers combinations and signal constellation between subcarriers will become more serious, and the filtering
points combinations by minimizing the following metrics: effect will become more obvious. In another point of view,
it can be observed clearly from the Fig. 4 that the proposed
v



2 scheme outperforms the OFDM-IM system under the influence
 
Rz (iz,η )

( I z , Sz ) = arg min

Fz (iz,η ) − H(iz,η )Sz (η)
(15)
of different CFOs. The UF-OFDM-IM system provides higher
Iz ,Sz
η=1 ICI robustness compared to the OFDM-IM system, due to
 The 28th Wireless and Optical Communication Conference (WOCC 2019)

the absence of the filtering operation which diminishes the subcarriers and their index combinations. After that, a the-
power leakage outside the subblock.When the CFO is small, oretical BER scenario is constructed from the ABEP, which
the influence of the CFO on the orthogonality of the system is will be used to consider the impact of the ICI on the BER.
slight, and the performances of the two systems are basically From ABEP, it can be found that the out of subblock power
no different. When the CFO is too large, the influence of the leakage has a significant impact on the BER. Therefore, the
CFO on the system is serious, resulting in the filtering effect power leakage outside the subblock could be decreased by
is no longer obvious. using filtering operation. Simulation results demonstrate that
our proposed scheme reduces the BER dramatically in the
OFDM-IM VS UF-OFDM-IM OFDM-IM systems compared with the previous works.
100
OFDM-IM CFO=0.05 ACKNOWLEDGMENT
UF-OFDM-IM CFO=0.05
10 -1
OFDM-IM CFO=0.1
This paper is supported by grants from the National Key
UF-OFDM-IM CFO=0.1 Research and Development Program of China under Contract
10-2 No. 2016YFE0200200.
CFO=0.1
R EFERENCES
10-3 CFO=0.05
[1] J. Wu and P. Fan., “A survey on high mobility wireless communica-
BER

tions:Challenges, opportunities and solutions,” IEEE Access., vol. 4, pp.

10-4 450-476, 2016.
[2] R. He, B. Ai, G. Wang et al., “High-Speed Railway Communications:
From GSM-R to LTE-R,” IEEE Vehicular Technology Magazine, vol. 11,
10-5 no. 3, pp. 49-58, Sep. 2016.
[3] E. Basar, ü. Aygölü, et al., “Orthogonal frequency division multiplexing
with index modulation,” IEEE Trans. Signal Process., vol. 61, no. 22, pp.
10-6 5536-5549, Nov. 2013.
[4] R. Mesleh, H. Haas, S. Sinanovic, C. W. Ahn, and S. Yun, “Spatial
modulation,” IEEE Trans. Veh. Technol, vol. 57, no. 4, pp. 2228-2241,
10-7
0 2 4 6 8 10 12 14 Jul. 2008.
SNR(dB) [5] J. Jeganathan, A. Ghrayeb, and L. Szczecinski, “Spatial modulation:
Optimal detection and performance analysis,” IEEE Commun. Lett., vol.
12, no. 8, pp. 545-547, Aug. 2008.
Fig. 3. Performance results of the OFDM-IM and UF-OFDM-IM for CFO [6] P. A. Martin, “Differential spatial modulation for APSK in time-varying
= 0.1 and CFO = 0.05. fading channels,” IEEE Commun. Lett., vol. 19, no. 7, pp. 1261-1264,
Jul. 2015.
[7] M. Chafii, J. P. Coon and D. A. Hedges, “DCT-OFDM With Index
Modulation,” IEEE Commun. Lett., vol. 21, no. 7, pp. 1489-1492, July
OFDM-IM VS UF-OFDM-IM 2017.
10-1 [8] Lixia Xiao, Bin Xu, Huirong Bai, Y. Xiao, X. Lei and S. Li, “Performance
evaluation in PAPR and ICI for ISIM-OFDM systems,” 2014 Interna-
tional Workshop on High Mobility Wireless Communications, 2014, pp.
10-2 84-88.
[9] Q. Ma, P. Yang, Y. Xiao, H. Bai, and S. Li, “Error probability analysis
of OFDM-IM with carrier frequency offset,” IEEE Commun. Lett., vol.
10-3 SNR=10 20,no. 12, pp. 2434-2437, Dec. 2016.
[10] S. Y. Yu, et al., “performance analysis of subcarrier index modulation-
OFDM in Doppler spread environments,” proc.int.conf.intell.syst., Mod-
BER

eling simulation , pp. 380-382, 2016.

10-4
[11] Y. Li, M. Zhang, X. Cheng, M. Wen and L. Yang, “Index modulated
SNR=12
OFDM with intercarrier interference cancellation,” 2016 IEEE Interna-
tional Conference on Communications (ICC), 2016, pp. 1-6.
10-5 [12] J. Zhang, M. Zhao, L. Zhang, J. Zhong and T. Yu, “Circular Convolution
OFDM-IM SNR=12
Filter Bank Multicarrier (FBMC) System with Index Modulation,” 2017
UF-OFDM-IM SNR=12 IEEE 86th Vehicular Technology Conference (VTC-Fall), 2017, pp. 1-5.
-6
10 OFDM-IM SNR=10 [13] S. Gokceli, E. Basar and G. K. Kurt, “Universal Filtered OFDM with
UF-OFDM-IM SNR=10 Filter Shift Keying - Invited Paper,” 2018 IEEE 87th Vehicular Technology
Conference (VTC Spring), 2018, pp. 1-5.
10-7 [14] Y. Ko, “A Tight Upper Bound on Bit Error Rate of Joint OFDM and
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 Multi-Carrier Index Keying,” IEEE Communications Letters, vol. 18, no.
normalized CFO 10, pp. 1763-1766, Oct. 2014.
[15] V. Vakilian, T. Wild, F. Schaich, S. ten Brink and J. Frigon, “Universal-
filtered multi-carrier technique for wireless systems beyond LTE,” 2013
Fig. 4. Performance results of the OFDM-IM and UF-OFDM-IM for SNR
IEEE Globecom Workshops (GC Wkshps), Atlanta, GA, 2013, pp. 223-
= 10 and SNR = 12.
228.
[16] Y. Li, B. Tian, K. Yi and Q. Yu, “A Novel Hybrid CFO Estimation
Scheme for UFMC-Based Systems,” IEEE Communications Letters, vol.
V. CONCLUSION 20, no. 12, pp. 2434-2437, Dec. 2016.

This work mainly considers the effect of CFO of the

OFDM-IM system, which transmits the information by the