Received January 8, 2017, accepted January 31, 2017, date of publication March 7, 2017, date of current version April

24, 2017.
Digital Object Identifier 10.1109/ACCESS.2017.2679135

Coding and Detection Schemes for Ambient

Backscatter Communication Systems
YANG LIU1 , GONGPU WANG2,3 , ZHONGZHAO DOU4 , AND ZHANGDUI ZHONG1,3
1 State Key Laboratory of Rail Traffic Control and Safety, Beijing Jiaotong University, Beijing 100044, China
2 Beijing Key Lab of Transportation Data Analysis and Mining, School of Computer and Information Technology, Beijing Jiaotong University, Beijing 100044,
China
3 Beijing Engineering Research Center of High-speed Railway Broadband Mobile Communications, Beijing 100044, China
4 Guangdong Planning and Designing Institute, Guangzhou, China, 510630

Corresponding author: Y. Liu (syangliu@bjtu.edu.cn)

This work was supported in part by the National Natural Science Foundation of China under Grant U1334202, Grant 61571037, and Grant
61402033 and in part by the ZTE Corporation.

ABSTRACT Ambient backscatter technology has attracted much attention recently, because it enables
battery-free devices, such as tags or sensors, to communicate through wireless energy harvesting and
radio backscattering. Existing studies about ambient backscatter assume that the tags have only two states:
backscatter or non-backscatter. Actually, some references have shown that the tags can readily realize three
states: positive and negative phase backscatter, and non-backscatter. In this paper, we propose a new coding
scheme for these tags with three states to improve the throughput of the ambient backscatter communication
system. We then design a maximum a posteriori (MAP) detector for the reader to extract binary information
from ternary coded signals. We also analyze the detection performance in terms of closed-form bit error
rate (BER) expressions. It is found that the proposed coding scheme can improve the throughput of an
ambient backscatter system, and there exists an error floor for the BER curve. Finally, simulation results are
provided to corroborate our theoretical studies.

INDEX TERMS Ambient backscatter, coding, maximum a posteriori (MAP) detection, bit error rate (BER),
error floor, system throughput.

I. INTRODUCTION performance versus communication distance through exper-

Radio Frequency Identification (RFID) technologies are
extensively used in Internet of Things [1]. One feature of
RFID is that the tag requires special signal source to pro-
vide power for communication [2]. To avoid special power
source, ambient backscatter communication systems is pro-
posed in [3] through harvesting the electromagnetic energy
from the environment.
The transmission process of ambient backscatter com-
munication systems could be divided into two phases [3]:
first, the battery-free devices such as tags or sensors transmit
0 or 1 signals through backscattering the ambient radio signal
or not; second, the reader separate the two kinds of signals
with certain signal processing skills. For example, the tag’s
data rate can be set lower than the ambient radio signal so
that the reader can obtain sufficient samples for each symbol
of the tag through averaging [3], [4].
Ambient backscatter, as a newborn communication tech-
nology, has attracted much attentions. Liu et al. [3]
first propose and implement a prototype of ambient
backscatter system, and investigate its bit error rate (BER)
iments. Wireless Fidelity (Wi-Fi) backscatter is designed to
enable the batter-free devices to communicate with off-the-
shelf Wi-Fi devices [5]. The Wi-Fi and ZigBee-compatible
signals used for backscattering can also be created by Blue-
tooth transmissions [6]. Furthermore, Parks et al. [7] intro-
duce the multi-antenna cancelation design for backscatter
devices and utilize a coding scheme to enlarge communica-
tion range. The signal detection and theoretical performance
analysis are provided in [4]. Reference [8] presents a novel
communication system utilizing full-duplex [9] to enable
high-throughput and long-range communications between
tags and a Wi-Fi access point.
Most existing studies assume that the tags can only
backscatter or not backscatter the surrounding signals. How-
ever, it has been pointed out in the references [2], [8],
and [10] that the tag can change its antenna impedance to
implement three operations on the environmental signals:
not reflecting, reflecting in the same phase and reflecting in
the opposite phase. That is, the tag can readily obtain three
states.

2169-3536
2017 IEEE. Translations and content mining are permitted for academic research only.
VOLUME 5, 2017 Personal use is also permitted, but republication/redistribution requires IEEE permission. 4947
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
 Y. Liu et al.: Coding and Detection Schemes for Ambient Backscatter Communication Systems
FIGURE 1. A communication system of RF-powered devices that utilizing
ambient backscatter.
Motivated by this, we study the situation where the tag
utilizes three states to transmit ternary information. Based
on this, we propose a new coding scheme and prove its
optimality. We also suggest a detection algorithm. In addi-
tion, we derive the closed-form expression of BER perfor-
mance and prove the existence of the error floor of the BER
curves. Finally, simulation results are provided to corroborate
our proposed studies. It is shown that the proposed coding
scheme and the corresponding detector can improve the sys-
tem throughput.
The rest of this paper is organized as follows. In Section II,
we introduce the coding scheme and theoretical model of our
ambient backscatter communication system. We propose a
detection algorithm and investigate its BER performance in
Section III. The simulation results are provided in Section IV.
Finally, Section V concludes this paper.
II. CODING SCHEME AND SYSTEM MODEL
Consider one ambient backscatter communication system
consisting of one reader and one tag, as shown in Fig. 1.
Unlike conventional RFID tags, the tag in Fig. 1 scavenges
other radio frequency (RF) signals such as radio television
signals to communicate with the reader [3].
A. CODING SCHEME
Suppose the RF transmit signal s(n) is an on-off keying mod-
ulated signal. The tag will receive the RF source signal and
will transmit its own binary signal B(n) to the reader through
positive and negative backscattering the signal s(n), and non-
backscattering. Specifically, the signal processing inside the
tag could be illustrated as follows:
1) If B(n) = 0, the tag changes its antenna impedance so
that little energy of s(n) can be reflected.
2) If B(n) = 1, the tag switches the antenna impedance to
scatter the signal without changing the phase.
3) If B(n) = −1, the tag adjusts its antenna impedance to
scatter the signal s(n) in the inverse phase.
Note that the tag has three states to represent its binary
information. Two ternary symbols can represent nine states,
while three binary symbols can only reflect eight states.
Therefore our main task is to find an optimal mapping from
4948
FIGURE 2. Signal constellation and coding scheme.
two ternary symbols to three binary symbols. Clearly there
exists one redundant code for the ternary-symbols-based
codebooks. Corresponding to the different redundant points,
we have nine coding schemes. Next, we investigate the coding
schemes in order to find the one that can minimize the BER.
Proposition 1: Among all the nine possible coding
schemes, the one without point (0, 0) has the minimum BER.
Proof: Coding theories show that the average
Euclidean distance can evaluate the BER performance
of different coding schemes [11], [12]. The constel-
lation figure of the two ternary-symbols-based coding
scheme is shown in Fig. 2. Owing to the symmetry
of Fig. 2, the nine points can be divided into three
sets: A = {(−L, −L), (L, −L), (L, L), (−L, L)}, B =
{(0, L), (0, −L), (−L, 0), (L, 0)} and C = {(0, 0)}, where L is
the unit distance between two adjacent constellation points.
When the coding schemes’ redundant points belong to the
same set, they have identical average Euclidean distance.
Consequently, we only need to choose one representative
redundant point from each set to calculate the average
Euclidean distance. For the set A, we choose only one
point out and calculate the average Euclidean distance of
the remaining eight points: DA ≈ 5.02L. Similarly, we
choose point (0, −L) and (0, 0) from set B and C respectively,
and obtain the corresponding average Euclidean distances:
DB ≈ 5.82L and DC ≈ 6.15L. Clearly, DA < DB < DC .
Thus the coding scheme without point (0, 0) has the minimum
BER.
Our following discussion of this paper is based on the
coding scheme shown in Fig. 2.
B. MATHEMATICAL EXPRESSIONS OF SYSTEM MODEL
Denote the channel between the RF source and the reader as h,
the channel between the RF source and the tag as f , and the
channel between the reader and the tag as g. The channels
f , g, and h are assumed as slow-fading and remain unchanged
during at least one consecutive interval of B(k). Furthermore,
VOLUME 5, 2017
Y. Liu et al.: Coding and Detection Schemes for Ambient Backscatter Communication Systems

we assume all these channels are known to the reader through

estimators provided in [13]–[15].
The signal received by the tag can be expressed as [10]
x(n) = fs(n). (1)
Moreover, the signal backscattered by the tag may be written
as
a(n) = ηB(n)x(n), B(n) ∈ {−1, 0, 1}, (2)
where η is the signal attenuation inside the tag. Note that
the data rate of the signal B(n) is much less than that of the
RF signal x(n), and B(n) will remain unchanged during N
symbols of x(n). That is, B(kN + j) will be the same for
j = 1, 2, · · · , N .
Therefore, the reader will receive the signal
y(n ) = hs(n) + ga(n) + wb (n) FIGURE 3. An example for the pdf Tk , k = −1, 0, 1.
= hs(n) + ηfgB(n)s(n) + wb (n), (3)
where wb (n) is the zero-mean additive white Gaussian noise √
1) If B(k) = 0, Tk is subjected to N( Php, N1 (Ph2 p2 +
(AWGN) term at the reader and its variance is Nwb . Next the Nwb )).
reader aims to recover the tag signals B(n) from the received √
2) If B(k) = 1, Tk is subjected to N( Pµp, N1 (Pµ2 p2 +
signals y(n). Nwb )).
Remark 1: Noting that the original binary bits "1" and "0" √
3) If B(k) = −1, Tk is subjected to N( Pνp, N1 (Pν 2 p2 +
are equiprobable. Let H0 , H1 and H−1 represent B(n) = 0, Nwb )).
B(n) = 1 and B(n) = −1, respectively. Therefore when the
coding scheme in Fig. 2 is adopted, we have Pr(H0 ) = 41 , One example of Tk ’s probability distribution under three
conditions is depicted in Fig. 3. It is shown that our statistic
Pr(H−1 ) = 38 and Pr(H1 ) = 38 , where Pr(Hi ) = Pr(B(n) = i).
Tk may have three probability density functions (PDFs):
III. DETECTION ALGORITHM AND PERFORMANCE Pr(Tk |Hi )
ANALYSIS √ !
N (Tk − Pνp)2

1
In this section, we introduce our proposed detection algorithm  exp − 2(Pν 2 p2 + N )



 r 
according to the ternary-symbol-based coding scheme shown
 1 2 2 wb
2π N (Pν p + Nwb



in Fig. 2, and analyze its BER performance. We derive the




closed-form BER expression and prove the existence of the

 i = −1, !

 √
error floor. Php)2


 1 N (Tk −

 exp −
2(Ph2 p2 + Nwb )
r  
= 1 2 p2 + N
A. DETECTOR DESIGN  2π N (Ph wb

Let us rewrite (3) as 

 i = 0,
√ 
 √ !
N (Tk − Pµp)2

√Phs(n) + wb (n), case H0 : B(k) = 0, 1
 

 exp − 2(Pµ2 p2 + N )


(4)

y(n) =
r 
Pµs(n) + wb (n), case H1 : B(k) = 1, 
 1 wb
√ 2
2π N (Pµ p + Nwb2
 

Pνs(n) + wb (n), case H−1 : B(k) = −1,





i = 1.
where P is the transmitting power of the RF source, µ , h +
(6)
ηfg and ν , h − ηfg denote the combined channel.
The statistic for detecting B(k) is chosen as Furthermore, we can derive the posteriori probability as
kN
1 Pr(Tk |Hi ) Pr(Hi )
.
X
Tk = y(n), 1 ≤ k ≤ K. (5) Pr(Hi |Tk ) = (7)
N Pr(Tk )
n=(k−1)N +1
In order to obtain the minimum error probability, we
Assume h, µ and ν are known to the reader, and s(n)
choose an maximum a posteriori (MAP) detector [4]. Let H b
is subjected to equiprobable
P binary distribution. When N is denote the detection result, and H̄i denote that the detection
large enough, the sum kN n=(k−1)N +1 s(n) is approximate to result is not Hi . The MAP decision rule is that, decide H
b =
the normal distribution N(Np, Np2 ), where p is the probability Hi , if
that s(n) = 1, i.e., p = 21 . Therefore, we can obtain the
following results: Pr(Tk |Hi ) Pr(Hi ) > Pr(Tk |Hl ) Pr(Hl ), ∀i 6 = l. (8)

VOLUME 5, 2017 4949

 Y. Liu et al.: Coding and Detection Schemes for Ambient Backscatter Communication Systems

To find the threshold, we should find suitable Tk s that sat- 2) In the case of fg < 0,
isfy the following two unary quadric equations respectively 
H−1 , if Tk > ξ1 ,

Pr(T−1 |H−1 ) Pr(H−1 ) = Pr(T0 |H0 ) Pr(H0 ), (9) H = H0 ,
b if ξ2 < Tk < ξ1 , (16)
Pr(T1 |H1 ) Pr(H1 ) = Pr(T0 |H0 ) Pr(H0 ). (10) 

H1 , if Tk < ξ2 .
The solutions of (9) is derived as (11), shown at the bottom
of this page, where B. BER PERFORMANCE ANALYSIS
The error probability of our detector (15) and (16) is given
H = Ph2 p2 + Nwb , by
V = Pν 2 p2 + Nwb ,
Pe = Pr(H0 ) Pr(H̄0 |H0 ) + Pr(H1 ) Pr(H̄1 |H1 )
κV = Vh − H ν.
+ Pr(H−1 ) Pr(H̄−1 |H−1 ), (17)
Owing to the relationship of h, µ and ν, the PDF curve of
Tk under H0 is always in the middle of the three curves in where
Fig. 3. Therefore we choose one solution in (11) nearer the  √ 
expectation of Tk under H0 . The threshold based on (9) is set ξ1 − Pνp
,
Pr (H̄−1 |H−1 ) = Q  q (18)
as 1 2 2
( √ √ + Nwb )
N (Pν p
ζ1,a , if |ζ1,a − Php| < |ζ1,b − Php|, √
ξ1 =
 
√ √ (12) ξ − Php
ζ1,b , if |ζ1,b − Php| ≤ |ζ1,a − Php|. Pr (H̄0 |H0 ) = 1 − Q  q
1 
1 2 2
Similarly, the other threshold can be obtained from the N (Ph p + Nwb )
solutions of (10), as (13), shown at the bottom of this page, √ !
ξ2 − Php
where +Q 1 , (19)
2 2
N (Ph p + Nwb )
U = Pµ2 p2 + Nwb ,  √ 
ξ2 − Pµp
κU = Uh − H µ. Pr (H̄1 |H1 ) = 1 − Q  q , (20)
1 2 p2 + N )
N (Pµ wb
And the threshold according to (10) is chosen as
( √ √ where Q(·) denotes the Gaussian Q-function [16] and is given
ζ2,a , if |ζ2,a − Php| < |ζ1,b − Php|,
ξ2 = √ √ (14) as
ζ2,b , if |ζ2,b − Php| ≤ |ζ1,a − Php|. Z ∞
1 t2
Finally the related detection rule can be summarized as Q(x) = √ e− 2 dt. (21)
follows 2π x
1) In the case of fg > 0, Define γ = NPwb as the transmit SNR. Next we inves-
 tigate the asymptotic error probability performance when
H−1 , if Tk < ξ1 ,

γ → ∞, i.e., lim Pe . For the simplicity of our discus-
H = H0 ,
b if ξ1 < Tk < ξ2 , (15) γ →∞
sion, we only consider the situation when ξ1 = ζ1,a and
H1 , if Tk > ξ2 .


ξ2 = ζ2,a .

 r h i
κ κ 2 − (v2 − h2 ) PN p2 (h2 − v2 ) + VH
ln 3H



 V + V wb N 2V
ζ


 1,a
 = √
r P(v2 − h2 )p (11)
h i
κV − κV2 − (v2 − h2 ) PNwb p2 (h2 − v2 ) + VH
ln 3H


N 2V


ζ1,b =


 √
P(v2 − h2 )p
 r h i
κ κ 2 − (u2 − h2 ) PN p2 (h2 − u2 ) + UH
ln 3H



 U + U wb N 2U
ζ2,a =

 √
P(u2 − h2 )p

r h i (13)
κU − κU2 − (u2 − h2 ) PNwb p2 (h2 − u2 ) + UH 3H



 N ln 2U
ζ2,b =


 √
P(u2 − h2 )p

4950 VOLUME 5, 2017

Y. Liu et al.: Coding and Detection Schemes for Ambient Backscatter Communication Systems

Theorem 1: The BER (17) has an error floor in high SNR

region, and the error floor is given by
√  !
5 3 ν N h
lim Pe = + Q + G(ν) − 1
γ →∞ 8 8 |ν| h + ν
√  !
1 h N ν
− Q + G(ν) − 1
4 |h| h + ν
√  !
1 h N µ
+ Q + G(µ) − 1
4 |h| h + µ
√  !
3 µ N h
− Q + G(µ) − 1 , (22)
8 |µ| h + µ
where G(x) is defined as
q
3h2
|xh| (x − h)2 − (x 2 − h2 ) ln 2x 2 FIGURE 4. Error of probability for the reader with different N.
G(x) = 2 2
. (23)
x(x − h )
Proof: Since Pe is a sum of several separate parts, we
can obatin
lim Pe = Pr(H0 ) lim Pr(H̄0 |H0 )
γ →∞ γ →∞
+ Pr(H1 ) lim Pr(H̄1 |H1 ) + Pr(H−1 ) lim Pr(H̄−1 |H−1 ).
γ →∞ γ →∞
(24)
For the first item of (24), we find
 √ 
ξ1 − Pνp
lim Pr(H̄−1 |H−1 ) = lim Q  q 
γ →∞ γ →∞ 1 2 2
N (Pν p + Nwb )
 √ 
ξ1 − Pνp
= Q  lim q 
γ →∞ 1 2 2
N (Pν p + Nwb )
ξ1
 
√ − νp FIGURE 5. Error of probability for the reader with different SNRs.
P
= Q  lim q 
γ →∞
Nν p
1 2 2

ξ1 ambient backscatter system utilizing existing coding scheme

− νp
 
lim √
γ →∞ P and detector [4] is also provided for comparison. To examine
= Q q . (25) the proposed scheme, each experiment performs 105 Monte
Nν p
1 2 2
Carlo trials. Here we set η = 0.7 and Pr(s(n) = 0) =
And it can be readily checked that Pr(s(n) = 1) = 0.5.
q Since the normal approximation accuracy depends on the
1 − ν+h
|νh||ν−h| 3h 2
ξ1 h+ ν(ν−h) ν−h ln 2ν 2 size of sample number N , we first investigate the influence
lim √ = νp · . (26) of N on BER Pe under different SNR scenarios in Fig. 4 . It
γ →∞ P ν+h
is shown that in all SNR scenarios, the BER performance is
Substituting (26) into (25) we can write the first item of (24) improved with increasing N , especially in low SNR. It can
as (27) provided on the top of the next page. Similarly, the also been found in Fig. 4 that the condition N ≥ 60 in the
other items in (24) can be given by (28) and (29) in the SNR region greater than 5dB, or the condition N ≥ 120
following page. in low SNR region is required to guarantee reliable perfor-
Finally, we can obtain (22) through substituting (27), (28) mance.
and (29), shown at the top of the next page, into (24), which The BER performance of the reader versus SNR is depicted
is irrelevant to the SNR γ . in Fig. 5, where the sample number N is set as 100. It is shown
IV. SIMULATION RESULTS that our theoretical BER results (17) well agree with the sim-
In this section, we illustrate the proposed detection algorithm ulation results. It can be also found that there exists an error
and theoretical BER performance analysis through Monte floor in high SNR region, which verify our derivation (22)
Carlo simulations. The uplink throughput performance of the in Section III.

VOLUME 5, 2017 4951

 Y. Liu et al.: Coding and Detection Schemes for Ambient Backscatter Communication Systems

 √  q 
|νh| 3h 2
ν N  h(ν − h) + ν (ν − h)2 − (ν 2 − h2 ) ln 2ν 2
lim Pr(H̄−1 |H−1 ) = Q  − 1 (27)
γ →∞ |ν| ν 2 − h2

 √  q 
|νh| 3h2
h N ν(ν − h) + ν (ν − h)2 − (ν 2 − h2 ) ln 2ν 2
lim Pr(H̄0 |H0 ) = 1 − Q   − 1
γ →∞ |h| ν 2 − h2
 √  q
3h2

h N µ(µ − h) + |µh|
µ (µ − h)2 − (µ2 − h2 ) ln 2µ 2
+Q  − 1 (28)
|h| µ2 − h2
 √  q
2 − (µ2 − h2 ) ln 3h2

µ N h(µ − h) + |µh|
µ (µ − h) 2µ 2
lim Pr(H̄1 |H1 ) = 1 − Q   − 1 (29)
γ →∞ |µ| µ2 − h2

V. CONCLUSION
Ambient backscatter is a newborn technology that has many
open problems. This paper focused on the coding and detec-
tion of the uplink ambient backscatter systems where the
tag has three states. We proposed a new coding scheme
and designed the corresponding detector. The closed-form
expression of detection BER was obtained, and its error floor
was also proved and derived. It was shown that the throughput
was improved with our new coding scheme. Finally, sim-
ulation results were provided to corroborate our theoretical
results.

FIGURE 6. Throughput of different coding schemes and detectors.
The throughput performance of the uplink ambient
backscatter system based on different coding schemes and
detectors are plotted in Fig. 6. The traditional coding scheme,
the corresponding detector and its BER P0e are given in
[4, Eq. (46)], and its throughput is calculated as follows
(1 − P0e )
Rold = log2 (1 + γ ).
N
For our proposed scheme, the throughput is given by
(1 − Pe ) log2 3
Rnew = log2 (1 + γ ),
N USA, 2014, pp. 1–12.
where the BER Pe is provided in (17). Fig. 6 plots both
throughput curves for Rold and Rnew . It can be seen from
Fig. 6 that the new coding scheme can significantly improve
the system throughput. Besides, when the traditional coding
scheme is applied, both the traditional detector and our pro-
posed detector are suitable and can achieve similar perfor-
mance, which indicates that the proposed detector has similar
detection accuracy with the traditional one.
REFERENCES
[1] L. Xie, Y. Yin, A. V. Vasilakos, and S. Lu, ‘‘Managing RFID Data:
Challenges, opportunities and solutions,’’ IEEE Commun. Surveys Tut.,
vol. 16, no. 3, pp. 1294–1311, Mar. 2014.
[2] C. Boyer and S. Roy, ‘‘Backscatter communication and RFID: Coding,
energy and MIMO analysis,’’ IEEE Trans. Commun., vol. 62, no. 3,
pp. 770–785, Mar. 2014.
[3] V. Liu, A. Parks, V. Talla, S. Gollakota, D. Wetherall, and J. R. Smith,
‘‘Ambient backscatter: Wireless communication out of thin air,’’ in Proc.
ACM SIGCOMM, Hong Kong, 2013, pp. 1–13.
[4] G. Wang, F. Gao, R. Fan, and C. Tellambura, ‘‘Ambient backscatter com-
munication systems: Detection and performance analysis,’’ IEEE Trans.
Commun., vol. 64, no. 11, pp. 1–10, Aug. 2016.
[5] B. Kellogg, A. Parks, S. Gollakota, J. R. Smith, and D. Wetherall, ‘‘Wi-Fi
backscatter: Internet connectivity for RF-powered devices,’’ in Proc. ACM
SIGCOMM, Chicago, IL, USA, 2014, pp. 1–12.
(30) [6] V. Iyer, V. Talla, B. Kellogg, S. Gollakota, and J. R. Smith, ‘‘Inter-
technology backscatter: Towards Internet connectivity for implanted
devices,’’ in Proc. ACM SIGCOMM, Florianópolis, Brazil, 2016,
pp. 1–14.
[7] A. Parks, A. Liu, S. Gollakota, and J. R. Smith, ‘‘Turbocharging ambient
(31) backscatter communications,’’ in Proc. ACM SIGCOMM, Chicago, IL,
[8] D. Bharadia, K. R. Joshi, M. Kotaru, and S. Katti, ‘‘BackFi: High through-
put WiFi backscatter,’’ in Proc. ACM SIGCOMM, London, U.K., 2015,
pp. 1–14.
[9] Z. Zhang, K. Long, V. A. Vasilakos, and L. Hanzo, ‘‘Full-duplex wireless
communications: Challenges, solutions and future research directions,’’
Proc. IEEE, vol. 104, no. 7, pp. 1369–1409, Jul. 2016.
[10] J. Kimionis, A. Bletsas, and J. N. Sahalos, ‘‘Increased range bistatic scatter
radio,’’ IEEE Trans. Commun., vol. 62, no. 3, pp. 1091–1104, Mar. 2014.
[11] J. G. Proakis and M. Salehi, Digital Communications, 5th ed. New York,
NY, USA: McGraw-Hill, 2008.

4952 VOLUME 5, 2017

Y. Liu et al.: Coding and Detection Schemes for Ambient Backscatter Communication Systems
[12] F. Xu, Y. Q. Jin, and A. Moreira, ‘‘A preliminary study on SAR advanced
information retrieval and scene reconstruction,’’ IEEE Geosci. Remote
Sens. Lett., vol. 13, no. 10, pp. 1443–1447, Oct. 2016.
[13] F. Gao, T. Cui, and A. Nallanathan, ‘‘On channel estimation and optimal
training design for amplify and forward relay network,’’ IEEE Trans.
Wireless Commun., vol. 7, no. 5, pp. 1907–1916, May 2008.
[14] F. Gao, R. Zhang, and Y.-C. Liang, ‘‘Optimal channel estimation and train-
ing design for two-way relay networks,’’ IEEE Trans. Commun., vol. 57,
no. 10, pp. 3024–3033, Oct. 2009.
[15] C. Xing, S. Ma, Z. Fei, Y. C. Wu, and H. V. Poor, ‘‘A general robust linear
transceiver design for multi-hop amplify-and-forward MIMO relaying
systems,’’ IEEE Trans. Signal Process., vol. 61, no. 5, pp. 1196–1209,
Mar. 2013.
[16] I. M. Ryzhik, A. Jeffrey, and D. Zwillinger, Table of Integrals, Series and
Products. San Diego, CA, USA: Academic, 2007.
YANG LIU received the B.E. degree from the School of Electronic
and Information Engineering, Beijing Jiaotong University, Beijing, China,
in 2011, where he is currently pursuing the Ph.D. degree with the State
Key Laboratory of Rail Traffic Control and Safety. His current research
interests are in the field of detection theory, spectrum-sensing algorithm and
performance analysis for cognitive radio networks, and ambient backscatter
communication systems.
GONGPU WANG received the B.Eng. degree in communication engi-
neering from Anhui University, Hefei, China, in 2001, and the M.Sc.
degree from Beijing University of Posts and Telecommunications (BUPT),
China, in 2004, and the Ph.D. degree from the University of Alberta,
Edmonton, Canada, in 2011. From 2004 to 2007, he was an Assistant
Professor with the School of Network Education, BUPT. He has since joined
the School of Computer and Information Technology, Beijing Jiaotong Uni-
versity, China, where he is currently an Associate Professor. His research
interests include Internet of Things, wireless communication theories, and
signal processing technologies.
VOLUME 5, 2017
ZHONGZHAO DOU received the bachelor’s degree in communication
engineering and the M.Sc. degree from Beijing University of Posts and
Telecommunications, Beijing, China, in 2001 and 2004, respectively. He is a
Ph.D. candidate of South China University of Technology. He is a Wireless
Chief Engineer of the Guangdong Planning and Designing Institute and he
is currently the Vice President of the Digital and Economy Commission of
International Chamber of Commerce China. His research interests include
SON of LTE and wireless communication theory.
ZHANGDUI ZHONG received the B.E. and M.S. degrees from Beijing
Jiaotong University, Beijing, China, in 1983 and 1988, respectively. He
is currently a Professor and an Advisor of Ph.D. candidates with Beijing
Jiaotong University, where he is also a Director of the School of Computer
and Information Technology, and a Chief Scientist with the State Key
Laboratory of Rail Traffic Control and Safety. He is also a Director of
the Innovative Research Team, Ministry of Education, Beijing, and a Chief
Scientist of the Ministry of Railways, Beijing. His interests include wireless
communications for railways, control theory and techniques for railways, and
GSM-R systems. His research has been widely used in railway engineering,
such as Qinghai-Xizang railway, Datong-Qinhuangdao Heavy Haul railway,
and many high-speed railway lines in China. He is an Executive Council
Member of the Radio Association of China, Beijing, and a Deputy Director
of the Radio Association, Beijing.
He holds five invention patents, and has authored or co-authored seven
books, and more than 200 scientific research papers in his research area.
He received the MaoYiSheng Scientific Award of China, the ZhanTianYou
Railway Honorary Award of China, and the Top 10 Science/Technology
Achievements Award of Chinese Universities.
4953