858 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 60, NO.

3, MARCH 2011
Blind Spectrum Sensing for OFDM-Based
Cognitive Radio Systems
Simin Bokharaiee, Student Member, IEEE, Ha H. Nguyen, Senior Member, IEEE, and Ed Shwedyk
AbstractGiven the ever-growing demand for radio spectrum,
cognitive radio has recently emerged as an attractive wireless tech-
nology. Since orthogonal frequency-division multiplexing (OFDM)
is one of the major wideband transmission techniques, detection
of OFDM signals in low-signal-to-noise-ratio scenario is an impor-
tant research problem. In this paper, it is shown that cyclic pre-
x (CP) correlation coefcient (CPCC)-based spectrum sensing,
which was previously introduced as a simple and computationally
efcient spectrum-sensing method for OFDM signals, is a special
case of the constrained generalized likelihood ratio test (GLRT) in
the absence of multipath. As such, the performance of this algo-
rithm degrades in a multipath scenario, where OFDM is usually
implemented. Furthermore, by considering multipath correlation
in the GLRT algorithm and employing the inherent structure of
OFDM signals, a simple and low-complexity algorithm called the
multipath-based constrained-GLRT (MP-based C-GLRT) algo-
rithm is obtained. The MP-based C-GLRT algorithm is shown
to outperform the CPCC-based algorithm in a rich multipath
environment. Further performance improvement can be achieved
by simply combining both the CPCC- and MP-based C-GLRT
algorithms. A simple GLRT-based detection algorithm is also
developed for unsynchronized OFDM signals, whose performance
is only slightly degraded when compared with the synchronized
detection in a rich multipath environment.
Index TermsCognitive radio (CR), generalized likelihood ra-
tio test (GLRT), likelihood ratio test (LRT), orthogonal frequency-
division multiplexing (OFDM), spectrum sensing.
I. INTRODUCTION
W
ITH THE recent rapid growth in wireless communica-
tion applications and systems, the problem of spectrum
utilization has become more critical than ever before. As an
emerging solution, cognitive radio (CR) technology has been
introduced to take advantage of the licensed spectrum when
it is not in use by its primary users. The temporary use of
idle licensed frequency bands by unlicensed users provides an
efcient exploitation of limited spectrum bands. However, it is
Manuscript received February 8, 2010; revised July 6, 2010 and October 27,
2010; accepted November 29, 2010. Date of publication January 6, 2011; date
of current version March 21, 2011. This work was supported by Discovery
Grants from Natural Sciences and Engineering Council of Canada (NSERC).
The review of this paper was coordinated by Dr. Y.-C. Liang.
S. Bokharaiee is with the Department of Electrical and Computer Engi-
neering, University of Manitoba, Winnipeg, MB R3T 5V6, Canada (e-mail:
simin@ee.umanitoba.ca).
H. H. Nguyen is with the Department of Electrical and Computer Engineer-
ing, University of Saskatchewan, Saskatoon, SK S7N 5A9, Canada (e-mail:
ha.nguyen@usask.ca).
E. Shwedyk, retired, was with the Department of Electrical and Computer
Engineering, University of Manitoba, Winnipeg, MB R3M 2G7, Canada.
Color versions of one or more of the gures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identier 10.1109/TVT.2010.2103095
crucial for the CR system to continually monitor the spectrum
to prevent any interference with the primary users.
Orthogonal frequency-division multiplexing (OFDM), which
has been known to be one of the most effective multicarrier
techniques for broadband wireless communications due to its
inherent capability to combat multipath fading and avoid in-
tersymbol interference (ISI), has attracted signicant attention
in the development of CR technology [1], [2]. In an OFDM-
based system, CR spectrum sensing can be performed in either
the time or the frequency domain. In the frequency domain,
a subcarrierwise detection can be performed [3], [4], which
becomes important when the available bandwidth is divided
into subbands, which are allocated to different primary users.
In [3], a frequency-domain spectrum-sensing method has been
proposed, which optimizes sensing parameters over each sub-
carrier, whereas the algorithm in [4] is proposed to detect avail-
able portions of the spectrum. Although some OFDM-based
technologies, such as digital audio broadcasting or IEEE 802.16
wireless metropolitan area networks, employ data multiplexing
for different primary users, single-transmission OFDM systems
are also common in practice. Among them are digital video
broadcasting, IEEE 802.11 wireless local area networks, and
IEEE 802.16e mobile WiMAX. For such wireless systems,
the CR receiver needs to detect the primary OFDM signals to
monitor the licensed bandwidth.
The likelihood ratio test (LRT) is known to be the optimal
algorithm for spectrum sensing [5] if the exact knowledge
of channel state information and noise variance can be made
available at the receiver. Unfortunately, such knowledge is
generally difcult to obtain at the CR receiver. On the other
hand, the energy detection algorithm does not require channel
information and performs well if the noise variance is known.
However, even a slight uncertainty, i.e., around 0.5 dB in noise
variance, causes its performance to severely drop below the
performance with known noise variance [6]. It should also
be pointed out that the energy detector (ED) is the optimal
spectrum-sensing method for independent and identically dis-
tributed (i.i.d.) primary signals, and its performance is degraded
if the received primary signals are correlated [7]. Therefore,
it is of great importance to develop efcient spectrum-sensing
algorithms for the detection of signals, which are not necessar-
ily i.i.d. To alleviate the requirement on precise knowledge of
channel and noise variance, which are generally unavailable at
the CR receiver, a suitable detection algorithm is expected to
extract and use some particular features of the received primary
signals. In fact, a number of techniques have been proposed and
evaluated; among them are the statistical covariance algorithm
0018-9545/$26.00 2011 IEEE
BOKHARAIEE et al.: BLIND SPECTRUM SENSING FOR OFDM-BASED CR SYSTEMS 859
[8], eigenvalue-based algorithm [9], [10], and cyclostationary
algorithm [11], [12].
As OFDM is an effective technique to combat multipath
fading, it is expected that OFDM is employed in rich scat-
tering environments. With advanced compression techniques,
the information signals (in discrete-time complex baseband
representation) can be considered as i.i.d. random variables.
However, when these signals are passed through a multipath
fading channel, the received signals at the receiver are corre-
lated versions of the transmitted signals, which typically have
different statistics from the background noise. On the other
hand, a very important feature of OFDM transmission is the
use of cyclic prex (CP), which results in nonzero correlation
of the received primary signal samples at certain delays. In [11],
a method based on the cyclic feature of OFDM blocks in the
time domain has been proposed. Although the algorithm shows
good performance, the signal correlation induced by multipath
propagation is not exploited in such an algorithm.
The aim of a detection algorithm is to decide between the two
hypotheses of whether the primary signal is present or absent. In
the case when some parameters (e.g., channel state information
and noise variance) are not known, the hypothesis is called a
composite hypothesis [13]. For a composite hypothesis, one
approach is to perform estimations of the unknown parameters
(typically, the maximum-likelihood estimates). The estimated
parameters are then used in the LRT as if they are the correct
values. This results in the so-called generalized LRT (GLRT)
[5], [13]. GLRT has been widely employed in many hypothesis-
testing problems, e.g., [12], [14], and [15], including spectrum-
sensing applications [16], [17].
In this paper, it is rst shown that the GLRT algorithm can
exploit both multipath and cyclic correlations to yield a novel
blind spectrum-sensing algorithm. It is then veried that the
cyclic-prex correlation coefcient (CPCC)-based detection al-
gorithm is a special case of the constrained GLRT algorithm in
the absence of multipath fading channel. It is further shown that,
when multipath fading is present, which is the case for OFDM
applications, performance of the detection based on the CPCC
degrades. Furthermore, by exploiting the known structure of
the OFDM channel matrix in a constrained GLRT algorithm,
a detection algorithm that is solely based on the multipath
correlation coefcients (MPCCs) is obtained. By combining
the CPCC- and MPCC-based algorithms, an even more reliable
spectrum-sensing method can be realized.
The spectrum-sensing algorithms developed in the rst part
of this paper (Sections IV and V) assume that perfect syn-
chronization can be obtained at the CR receiver. As such,
the detection performance of sensing algorithms in this case
serves as an upper bound for situations where synchronization
has to be actually performed at the CR receiver. In [18], a
blind synchronization algorithm was proposed for maximum-
likelihood estimations of time and carrier frequency offsets
(CFOs) of OFDM signals. While such a blind synchronization
algorithm can be conveniently incorporated in our spectrum-
sensing framework, there are two main drawbacks: 1) Using
a synchronization algorithm adds more complexity and may
cause delay to the sensing task. 2) In the low-signal-to-noise-
ratio (SNR) region where the CR is operating, the synchro-
nization is far from perfect; hence, the sensing performance is
signicantly degraded when compared with the case of perfect
synchronization. These drawbacks motivate us to develop a
simple GLRT-based algorithm in Section VI that does not
require timing synchronization to be established between the
primary and secondary users.
The rest of the paper is organized as follows: Section II
reviews the system model and describes the spectrum-sensing
problem. Section III presents a general framework of the GLRT
detection scheme and shows how to exploit CP and multipath
correlation features. In Sections IV and V, two constrained
GLRT algorithms are presented, where the data and cyclic parts
of the OFDM signal are separately considered to exploit the
structure of the covariance matrix in enhancing the performance
of the GLRT algorithm. It is shown in both cases that the con-
strained GLRT algorithms lead to detection algorithms, which
solely depend on the sample correlation coefcients. A com-
bined detection scheme is also proposed. The spectrum-sensing
algorithm for unsynchronized OFDM signals is proposed in
Section VI. Simulation results are presented in Section VII, and
Section VIII concludes the paper.
II. SYSTEM MODEL
The OFDM signal model considered in our work is the same
as that in [11] and [19], which assumes that the primary OFDM
system employs L subcarriers and the CR and primary users
can be perfectly synchronized. The case of no timing synchro-
nization is separately discussed and treated in Section VI. Let
S
n,k

L1
k=0
, with E[S
n,k
[
2
=
2
S
, be the complex symbols
to be transmitted in the nth OFDM block. Then, the baseband
OFDM modulated signal can be expressed as
s
n
(m) =
1

L
L1

k=0
S
n,k
e
j2mk
L
, m = 0, . . . , L 1. (1)
For a large number of subcarriers L [i.e., the size of discrete
Fourier transform (DFT)/inverse DFT (IDFT)], s
n
(m) can be
approximately modeled as a zero-mean circularly symmet-
ric complex Gaussian random variable of variance
2
S
, i.e.,
s
n
(m) (^(0,
2
S
).
Represent the length-(L +L
p
) vector of the nth transmitted
OFDM block as
s
n
=

s
n
(L 1) . . . s
n
(0) s
n
(L 1) . . . s
n
(L L
p
)
. .
Cyclic Prex

(2)
where L
p
denotes the number of samples in the guard interval,
i.e., the length of the CP. The corresponding received signal and
noise vectors are denoted by
x
n
=[x
n
(L1) x
n
(L2) . . . x
n
(0)x
n
(1) . . . x
n
(L
p
)]

(3)
v
n
=[v
n
(L1) v
n
(L2) . . . v
n
(0)v
n
(1) . . . v
n
(L
p
)]

(4)
where the noise samples v
n
(l)s are i.i.d. (^(0,
2
v
) random
variables.
860 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 60, NO. 3, MARCH 2011
The primary user signal is received through a wireless mul-
tipath fading channel whose discrete-time baseband model is
represented by channel lter taps h
i
, i = 1, . . . , L
c
, where L
c
denotes the number of multipath components. It is also assumed
that the fading process remains static during the interval of
spectrum sensing. This implies that the channel lter taps can
be treated as unknown constants during the period of spectrum
sensing. The relationship of x
n
, s
n
, and v
n
can be expressed in
matrix form as follows:
x
n
= hs
n
+v
n
(5)
where s
n
= [s

n
, s
n1
(L 1), . . . , s
n1
(L L
c
+ 1)]

, and h
is (L +L
p
) (L +L
p
+L
c
1) Toeplitz matrix constructed
from the channel lter taps as
h =

h
1
h
L
c
0 0
0 h
1
h
L
c
0 0
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. 0 h
1
h
L
c
0 0
.
.
.
.
.
.
.
.
.
.
.
.
0
0 0 h
1
h
L
c

. (6)
Note that the last L
p
samples of x
n
is the ISI part.
Since x
n
(l)s are linear combinations of zero-mean complex
Gaussian random variables, they are also zero-mean complex
Gaussian random variables. Based on (5), the variance of x
n
(l)
is
2
x
=
2
S

L
c
i=1
[h
i
[
2
+
2
v
when the primary users signal is
present; otherwise,
2
x
=
2
v
. It is also of interest to dene the
SNR in the presence of the primary users signal as SNR =

2
S

L
c
i=1
[h
i
[
2
/
2
v
.
Two binary hypotheses H
0
and H
1
are dened in spectrum
sensing, in which H
0
denotes the idle state of the primary
user and H
1
represents the active state of the primary user. To
classify the observations into H
0
or H
1
, a test statistics T is
formulated, and a general test decision is given as follows:
_
Decide H
0
, if T
Decide H
1
, if T >
(7)
where is some threshold value. Two probabilities of interest
are given as follows: 1) the probability of detection P
d
, which
is the probability that the primary user is correctly detected in
its active mode, and 2) the probability of false alarm P
f
, which
represents the probability of a false detection of the primary
user when it is in the idle state. Mathematically
P
f
= Pr T > [H
0
(8)
P
d
= Pr T > [H
1
. (9)
III. GENERALIZED LIKELIHOOD RATIO TEST
As mentioned before, spectrum sensing based on GLRT has
been presented in [17], in which different tests are obtained
under different parameter assumptions, i.e., unknown noise
variance and/or signal covariance matrix. In the sequel, the
GLRT is reviewed in its general form, and it will be employed
for the detection of OFDM signals in the succeeding sections.
Let y
n
(^(0, R
y
) denote the length-M column vector of
the nth received signal block, and let z
n
(^(0,
2
z
I) denote
the length-M column vector containing the noise samples.
In the scenario in which the noise variance
2
z
and signal
covariance matrix R
y
are unknown, the GLRT is given as
follows [17]:
L
G
(y) =
f
y[H
1
, R
y
(y[H
1
,

R
y
)
f
y[H
0
,
2
z
(y[H
0
,
2
z
)
H
1

H
0
(10)
in which y = [y
1
, . . . , y
N
] is a collection of N received blocks.
In the preceding test,

R
y
and
2
z
are the maximum-likelihood
estimates of R
y
and
2
z
under hypotheses H
1
and H
0
,
respectively.
The maximum-likelihood estimate of
2
z
can be obtained as

2
z
= max

2
z
_
ln f
y[H
0
,
2
z
_
y[H
0
,
2
z
__
(11)
where
f
y[H
0
,
2
z
_
y[H
0
,
2
z
_
=
N

n=1
1
(
2
z
)
M
exp
_

2
z
|y
n
|
2
_
(12)
and | | denotes the vector Euclidean norm. It follows
that [17]

2
z
=
1
NM
tr(yy
H
) =
1
M
tr(

R
y
). (13)
where

R
y
= 1/Nyy
H
denotes the sample covariance matrix,
and H represents the Hermitian transpose.
On the other hand, the maximum-likelihood estimate of R
y
can be obtained as

R
y
= max
R
y
o
R
y
ln f
y[H
1
,R
y
(y[H
1
, R
y
) (14)
where o
R
y
species the set of R
y
having certain structures,
and
f
y[H
1
,R
y
(y[H
1
, R
y
)=
N

n=1
1

M
det(R
y
)
exp
_
y
H
n
R
1
y
y
n
_
.
(15)
To obtain a more explicit expression of the test in (10),
rst rewrite f
y[H
0
,
2
z
(y[H
0
,
2
z
) and f
y[H
1
,R
y
(y[H
1
, R
y
) as
follows:
f
y[H
0
,
2
z
_
y[H
0
,
2
z
_
=
_
1
(
2
z
)
M
_
N
exp
_

2
z
N

n=1
|y
n
|
2
_
=
_
1
(
2
z
)
M
_
N
exp
_

2
z
tr(yy
H
)
_
=
_
1

2
z
exp
_

2
z

2
z
__
NM
(16)
BOKHARAIEE et al.: BLIND SPECTRUM SENSING FOR OFDM-BASED CR SYSTEMS 861
f
y[H
1
,R
y
(y[H
1
, R
y
)
=
_
1

M
det(R
y
)
_
N N

n=1
exp
_
y
H
n
R
1
y
y
n
_
=
_
1

M
det(R
y
)
_
N
exp
_
tr
_
y
H
R
1
y
y
__
=
_
1

M
det(R
y
)
_
N
exp
_
tr
_
R
1
y
yy
H
__
yy
H
=N

R
y
=
_
1

M
det(R
y
)
_
N
exp
_
Ntr
_
R
1
y

R
y
__
=
_
1
() det
1
M
(R
y
)
exp
_

1
M
tr
_
R
1
y

R
y
_
_
_
NM
. (17)
After substituting

R
y
and
2
z
into (16) and (17), we have
f
y[H
0
,
2
z
_
y[H
0
,
2
z
_
=
_
1

2
z
exp(1)
_
NM
=
_
1

M
tr(

R
y
)
exp(1)
_
NM
(18)
f
y[H
1
, R
y
(y[H
1
,

R
y
)
=
_
1
det
1
M
(

R
y
)
exp
_

1
M
tr
_

R
1
y

R
y
_
_
_
NM
. (19)
Finally, substituting (18) and (19) into (10) yields the follow-
ing GLRT algorithm:
T
G
(y) = L
G
(y)
1
MN
=
1
M
tr(

R
y
)
det
1
M
(

R
y
)
(

R
y
)
H
1

H
0
(20)
where
(

R
y
) =
exp
_

1
M
tr
_

R
1
y

R
y
__
exp(1)
(21)
=
1/MN
is a xed threshold value selected to meet a require-
ment on the probability of false alarm, and it is independent of
the received primary signal characteristics.
It is important to point out that the form in (20) is still very
general and encompasses a large class of GLRT. What it says
is that the test itself and, hence, its performance depend on
how one estimates the covariance matrix of the received signal
block under H
1
. According to [20], under the mild condition
that both R
y
and its variation (R
y
) have the same structure
(e.g., R
y
does not have any constant entries), the structured
maximum-likelihood estimate of R
y
satises tr(

R
1
y

R
y
) =
M. This implies that (

R
y
) = 1, and the test becomes
T
G
(y) =
1
M
tr(

R
y
)
det
1
M
(

R
y
)
=
1
MN
tr(yy
H
)
det
1
M
(

R
y
)
H
1

H
0
. (22)
Since the covariance matrices involved in our study meet the
aforementioned structural condition, the test in (22) shall be
employed for spectrum sensing throughout the paper.
On the other hand, for the special case when there is no con-
straint on the structure of the covariance matrix, the maximum-
likelihood estimate of R
y
is

R
y
=

R
y
, and the test is the same
as that developed in [17] and has the following form:

T
G
(y) =
1
M
tr(

R
y
)
det
1
M
(

R
y
)
H
1

H
0
. (23)
The preceding unconstrained GLRT (U-GLRT) is described
for a general observation y = [y
1
, . . . , y
N
], where y
n

(^(0, R
y
), n = 1, . . . , N. When applied to the OFDM-based
CR system in Section II, one can simply replace y
n
by x
n
in
(5). Furthermore, it is also possible to use only a portion of
the complete observation y in the GLRT algorithm. Although
this appears to be counterproductive, the key advantage is that,
by concentrating on a certain part of the observation, one can
exploit structural properties of the covariance matrix to improve
its estimation in the GLRT. This is presented in detail in the
succeeding sections for an OFDM-based CR system.
IV. GENERALIZED LIKELIHOOD RATIO TEST BASED ON
CYCLIC PREFIX CORRELATION
One way to exploit a strong structural correlation property of
the observation is to use only the head and tail of each received
OFDM block in the GLRT algorithm. To this end, dene
x
n
= [x
n
(L 1) . . . x
n
(L L
p
), x
n
(0) . . . x
n
(L
p
)]

(24)
as the vector containing the 2L
p
samples, i.e., the L
p
samples
in the rst part and the L
p
samples in the last part, of the
nth OFDM block. The corresponding transmitted signal and
additive white Gaussian noise (AWGN) vectors are dened as
s
n
= [s
n
(L 1) . . . s
n
(L (L
p
+L
c
1)) s
n
(L 1) . . .
s
n
(L L
p
)s
n1
(L 1) . . . s
n1
(L L
c
+ 1)]

v
n
= [v
n
(L 1) . . . v
n
(L L
p
), v
n
(1) . . . v
n
(L
p
)]

.
Then, one has
x
n
=

h s
n
+ v
n
(25)
where

his the following 2L
p
2(L
p
+L
c
1) block diagonal
channel matrix:

h=
_

h
A
0
0

h
A
_
,

h
A
=

h
1
. . . . . . h
L
c
0 . . . 0
0 h
1
. . . . . . h
L
c
.
.
.
0
.
.
.
.
.
.
0
0 . . . 0 h
1
. . . h
L
c

L
p
(L
p
+L
c
1).
(26)
Let x = [ x
1
, . . . , x
N
] and s = [ s
1
, . . . , s
N
]. The covariance
matrix R
x
= E x
n
x
H
n
under H
1
can be shown to be
R
x
=

hR
s

h
H
+
2
v
I
L
p
=
_

h
A

h
H
A

h
B

h
H
B

h
B

h
H
B

h
A

h
H
A
_

2
S
+
2
v
I
L
p
(27)
862 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 60, NO. 3, MARCH 2011
where

h
B
is the rst L
p
L
p
block of matrix

h
A
, and I
L
p
denotes the identity matrix of size L
p
.
A CP correlation coefcient (CPCC)-based spectrum-
sensing algorithm was, in fact, proposed in [11] with the focus
on AWGN channels. Next, it is shown that this CPCC-based
sensing algorithm is exactly the constrained version of the
GLRT algorithm based on observation x and in the absence of
multipath environment. It will be shown in Section VII that the
constrained GLRT algorithm provides a substantial improve-
ment over the U-GLRT algorithm when only the observation x
is used. Furthermore, since the CPCC-based algorithm is only
equivalent to the constrained GLRT (C-GLRT) algorithm when
there is no multipath, it shall also be explicitly shown that its
performance is degraded in a multipath environment.
A. No-Multipath Propagation
In the absence of the multipath propagation effect, one has
L
c
= 1 and the covariance matrix R
x
in (27) has the following
simpler form:
R
x
=
2
x
_
I
L
p
I
L
p
I
L
p
I
L
p
_
(28)
where = Ex
n
(k)x

n
(k L)/Ex
n
(k)x

n
(k), k = L
1, . . . , L L
p
is the correlation coefcient among the corre-
sponding L
p
samples in the head and tail of the OFDM block.
It is simple to show that this correlation coefcient is given as
=
_
0, H
0

2
S
[h
1
[
2

2
S
[h
1
[
2
+
2
v
=
SNR
1+SNR
, H
1
.
(29)
Given the structure of the covariance matrix in (28), estimat-
ing R
x
is equivalent to estimating and
2
x
. Their maximum-
likelihood estimates are given as follows:
_
,
2
x
_
= max
,
2
x
ln f
x[H
1
,,
2
x
_
x[H
1
, ,
2
x
_
. (30)
Substituting y = x, M = 2L
p
, and R
y
= R
x
in (15) gives
ln f
x[H
1
,R
x
( x[H
1
, R
x
)
= NLln(2) N ln (det(R
x
))
N

n=1
x
H
n
R
1
x
x
n
. (31)
Using the identities det(rA
MM
) = (r
M
) det(A
MM
)
and det
_
A B
C D
_
= det(ABD
1
C), one has det(R
x
) =
(
2
x
)
2L
p
(1
2
)
L
p
. In addition, R
1
x
= (1/
2
x
(1
2
))
_
I
L
p
I
L
p
I
L
p
I
L
p
_
. Substituting these expressions in (31)
evaluates to
ln f
x[H
1
,,
2
x
_
x[H
1
, ,
2
x
_
= NLln(2) N ln
_
(
2
x
)
2L
p
(1
2
)
L
p
_
=
1

2
x
(1
2
)
(g
1
( x) g
2
( x)) (32)
where g
1
( x) =

N
n=1
( x
H
1,n
x
1,n
+ x
H
2,n
x
2,n
) and g
2
( x) =

N
n=1
( x
H
1,n
x
2,n
+ x
H
2,n
x
1,n
), and x
1,n
and x
2,n
are the vectors
containing the rst and last L
p
components of x
n
, respectively.
Let a = 1/
2
x
. Then, (31) can be rewritten as
f(, a) = NLln(2) + 2NL
p
ln(a) NL
p
ln(1
2
)

a
(1
2
)
(g
1
( x) g
2
( x)) . (33)
The rst derivatives of (33) with respect to and a can be
obtained as
f(, a)
a
=
2NL
p
a

1
(1
2
)
(g
1
( x)g
2
( x)) (34)
f(, a)

=
2NL
p

1
2

2a
(1
2
)
2
(g
1
( x)g
2
( x))+
ag
2
( x)
1
2
.
(35)
By simultaneously solving f(, a)/a = 0 and f(, a)/
= 0, one obtains

2
x
=
1
a
=
g
1
( x)
2NL
p
=

N
n=1
_
x
H
1,n
x
1,n
+ x
H
2,n
x
2,n
_
2NL
p
(36)
=
g
1
( x)
g
2
( x)
=

N
n=1
( x
H
1,n
x
2,n
+ x
H
2,n
x
1,n
)

N
n=1
_
x
H
1,n
x
1,n
+ x
H
2,n
x
2,n
_ (37)
which are exactly the sample variance and sample correlation
coefcient, respectively.
Next, (22) can be employed to obtain the test statistics by
setting M = 2L
p
and R
y
=

R
x
. With the aid of (28), the
test is
T
G
( x) =
1
2NL
p
tr( x x
H
)
det
1
2L
p
(

R
x
)
1
_
1
2
H
1

H
0

H
1

H
0

2
1

.
(38)
As can be seen, the test statistics in (38) simply compares
the cyclic correlation coefcient with a threshold. It is therefore
identical to the detection algorithm proposed in [11].
B. Multipath Channel Propagation
In this part, the asymptotic behavior of under H
1
is
analyzed to illustrate the performance degradation of the test in
(38) in the multipath scenario. First, observe that, in an AWGN
channel, we have the following limit:
lim
N
=
SNR
1 + SNR
. (39)
The sample covariance matrix

R
x
can be decomposed into
four L
p
L
p
block matrices as follows:

R
x
=
_

R
x,11

R
x,12

R
x,21

R
x,22
_
(40)
where

R
x,21
=

R
H
x,12
. Thus, (37) can be expressed as
=
tr(

R
x,12
+

R
x,21
)
tr(

R
x,11
+

R
x,22
)
. (41)
BOKHARAIEE et al.: BLIND SPECTRUM SENSING FOR OFDM-BASED CR SYSTEMS 863
Employing (27), the denominator of (41) has the following
asymptotic behavior under H
1
:
T
h
= lim
N
tr(

R
x,11
+

R
x,22
)=2
_
tr
_

h
A

h
H
A
_

2
S
+L
p

2
v
_
=2L
p
_
L
c

i=1
[h
i
[
2

2
S
+
2
v
_
=2L
p

2
v
(1 + SNR). (42)
For the nominator of (41), we have
^
h
= lim
N
tr(

R
x,12
+

R
x,21
) = 2tr
_

h
B

h
H
B
_

2
S
=2
2
S
L
p

j=1
j

i=1
[h
i
[
2
< 2
2
S
L
p

j=1
L
c

i=1
[h
i
[
2
= 2L
p

2
S
L
c

i=1
[h
i
[
2
= 2L
c

2
v
SNR.
(43)
It then follows that the CP correlation coefcient in the
multipath channel
h
is equal to
H
1
:
h
= lim
N
=
^
h
T
h
=

L
p
j=1

j
i=1
[h
i
[
2

2
S
L
p
_

L
c
i=1
[h
i
[
2

2
S
+
2
v
_
<
SNR
1 + SNR
= (44)
where the subscript h indicates the dependence on channel
realization h = [h
1
, . . . , h
L
c
]
T
. Note that, for a given P
f
, the
threshold in (38) is xed and does not depend on the corre-
lation coefcient. Therefore, (39) and (44) can be compared
to observe the degradation of the detection performance of
CPCC-based detection algorithm in the presence of multipath
propagation. To underestand the role of L
p
and L
c
in the
detection performance, one simple way is to obtain the expected
value of
h
with respect to all channel realizations with a xed
SNR. In Appendix A, it is shown that
H
1
: =E
h
=
SNR
1 + SNR
_
2L
p
L
c
+ 1
2L
p
_
=
_
2L
p
L
c
+ 1
2L
p
_
. (45)
From (45), it is clearly seen that, for a xed L
p
and SNR, the
average CP correlation coefcient decreases with an increase in
the number of multipath delay components.
V. GENERALIZED LIKELIHOOD RATIO TEST ALGORITHM
BASED ON MULTIPATH CORRELATION
As shown in the previous section, the CPCC-based detec-
tion algorithm suffers a performance degradation in a multi-
path channel. On the one hand, this is expected, because the
CPCC-based algorithm only uses observation in the head and
tail of an OFDM block to exploit the correlation structure,
which results from the use of the CP. On the other hand, multi-
path also introduces strong correlation to the received OFDM
samples, which could also be exploited in the constrained
GLRT algorithm. This is precisely the motivation and objective
of this section. The developed algorithm shall use the portion of
the received OFDM symbol that does not include the ISI part.
In this way, the known structure of the observation can be taken
into account to improve the estimation of the signal covariance
matrix. Furthermore, a simplied test statistics is derived as a
function of the received signal correlation coefcients.
A. Constrained GLRT Algorithm
The portion of the received OFDM block without the ISI
part and the corresponding transmitted signal and noise vectors
can be represented as x
n
= [x
n
(L 1), . . . , x
n
(0)]

, s
n
=
[s
n
(L 1), . . . , s
n
(0)]

, and v
n
= [v
n
(L 1), . . . , v
n
(0)]

,
respectively. They are related according to
x
n
=

hs
n
+ v
n
(46)
where

h is the L L circulant channel matrix, whose rst row
is [ h
1
h
L
c
1
h
L
c
0 0 ].
Let R
x
[r
k,j
]
k,j=1,...,L
= E x
n
x
H
n
be the signal covari-
ance matrix. Using (46), under H
1
, we have
R
x
=

h

h
H

2
S
+
2
v
I. (47)
The matrix R
x
has the following properties:
1) R
x
is Hermitian: r
k,j
= r

j,k
.
2) All the diagonal elements are equal: r
k,k
=

L
c
i=1
[h
i
[
2

2
S
+
2
v
, k = 0, . . . , L 1
3) Since

h

h
H
is circularly symmetric, R
x
is also circu-
larly symmetric. This last property is specic to OFDM
transmission.
From the preceding properties, one has the following
proposition.
Proposition: Let L
p
denote the length of the CP. To obtain an
estimate of the covariance matrix R
x
, it is sufcient to estimate
L
p
values.
Proof: Since L
c
L
p
, the circulant channel matrix

h has
at most L
p
nonzero values in each row. Therefore,

h

h
H
has at
most 2L
p
1 nonzero values in each row that also appear in all
the other rows of

h

h
H
due its circularity. Since

h

h
H
is also a
Hermitian matrix, there are L
p
1 conjugate pairs in each row,
excluding the diagonal element. Therefore, only L
p
values are
needed to completely dene R
x
.
The rst row of the covariance matrix can be expressed as
[r
1,0
r
1,1
. . . r
1,L1
]
=
_

0

1
. . .
L
p
1
0 . . . 0

L
p
1
. . .

1
_
. (48)
The consecutive rows are obtained through a right circular
shift of the previous row. To employ the GLRT algorithm, the
vector = [
0
,
1
, . . . ,
L
p
1
] can be estimated based on the
criterion in (14), which is equivalent to
= max

ln f
x[H
1
,
( x[H
1
, ). (49)
864 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 60, NO. 3, MARCH 2011
Solving the preceding problem is quite challenging since the
term R
1
x
cannot be easily differentiated with respect to .
Instead, we consider an equivalent optimization problem as
described here.
Since R
x
is a circulant matrix, all the vectors w
k
=
(1/

L)[1, e
j2k/L
, . . . , e
j2k(L1)/L
]

, k = 0, . . . , L 1,
are its eigenvectors with the corresponding eigenvalues, i.e.,

k
=
L1

i=0
r
1,i
e

j2ki
L
=
0
+
L
p
1

m=1
Re
_

m
e

j2km
L
_
. (50)
Let W= [w
0
, . . . , w
L1
] denote the matrix of eigenvectors
(which is also an IDFT matrix), and let = diag(
k
) denote
the diagonal matrix of the eigenvalues. Then, R
x
= W
H
W
is the eigenvalue decomposition of R
x
, and we have
W
H
R
x
W= W
H
WW
H
W= . (51)
Let X
n
= W
H
x
n
= [X
0,n
, . . . , X
L1,n
]

be the DFT of
x
n
. It is obvious that
R
X
=E
_
X
n
X
H
n
_
= E
_
W
H
x
n
x
H
n
W
_
=W
H
E
_
x
n
x
H
n
_
W= . (52)
The preceding means that
k
s represent the average energy
per each subcarrier, and hence, they are positive. From (50), it
is observed that

m
=
1
L
L1

k=0

k
e
j2km
L
, m = 0, . . . , L
p
1. (53)
Therefore, if we can nd the maximum-likelihood estimate
of
k
s by solving

k
= max

k
ln f
x[H
1
,R
x
( x[H
1
, R
x
) (54)
then the maximum-likelihood estimate of
m
s can be ob-
tained as

m
=
1
L
L1

k=0

k
e
j2km
L
, m = 0, . . . , L
p
1. (55)
To express ln f
x[H
1
,R
x
( x[H
1
, R
x
) as a function of
k
s, rst,
substitute y = x and M = L in (15) to yield
ln f
x[H
1
,R
x
( x[H
1
, R
x
)
= NLln(2) N ln det(R
x
)
N

n=1
x
H
n
R
1
x
x
n
. (56)
With the aid of (52), it is observed that
x
H
n
R
1
x
x
n
= x
H
n
W
1
W
H
x
n
= X
H
n

1
X
n
=
L1

k=0
[X
n,k
[
2

k
(57)
det(R
x
) =
L1

k=0

k
. (58)
Using (57) and (58), (56) can be equivalently expressed as
(
0
, . . . ,
L1
) = NLln(2) N ln
L1

k=0

n=1
L1

k=0
[X
n,k
[
2

k
. (59)
For convenience, dene
k
= (1/
k
) > 0, and rewrite
(59) as
(
0
, . . . ,
L1
) = NLln(2) +N ln
L1

k=0

n=1
L1

k=0

k
[X
n,k
[
2
. (60)
It is simple to verify that
(
0
, . . . ,
L1
)

k
=
N

n=1
[X
n,k
[
2
(61)

2
(
0
, . . . ,
L1
)

2
k
=
N

2
k
< 0. (62)
Therefore, (
0
, . . . ,
L1
) is a concave function of

0
, . . . ,
L1
. Using (61), we have

k
=
1

k
=
1
N
N

n=1
[X
n,k
[
2
. (63)
Employing (63) in (55) gives

m
=
1
NL
L1

k=0
N

n=1
[X
n,k
[
2
e
j2km
L
=
1
NL
L1

k=0
N

n=1
X

n,k
e
j2km
L
1

L
L1

i=0
x
n
(i)e

j2ki
L
=
1
NL
N

n=1
L1

i=0
x
n
(i)
1

L
L1

k=0
X

n,k
e

j2(im)k
L
=
1
NL
N

n=1
L1

i=0
x
n
(i)
_
1

L
L1

k=0
X
n,k
e
j2(im)k
L
_

=
1
NL
N

n=1
L1

i=0
x
n
(i)x

n
([i m] mod L)
m = 0, . . . , L
p
1 (64)
which represent the sample correlation values up to the maxi-
mum delay of L
p
1 samples.
Next, the covariance matrix estimate

R
x
can be constructed
from the sample correlation values obtained in (64). The result-
ing test statistics can then be established by exploiting (22) with
y = x,

R
y
=

R
x
, and M = L. In the time domain, one has
T
G
( x) =
1
LN
tr( x x
H
)
det
1
L
(

R
x
)
H
1

H
0
. (65)
BOKHARAIEE et al.: BLIND SPECTRUM SENSING FOR OFDM-BASED CR SYSTEMS 865
B. MPCC-Based Test
The spectrum-sensing framework elaborated so far in this
section makes use of the correlation property of the primary
signal to identify it fromthe background noise. It is of interest to
establish an approximated but simpler test that can still capture
the multipath correlation of the primary signal. Some recent
works [21], [22] have also intuitively developed test statistics as
functions of the received signal in the time domain by exploit-
ing the multipath correlation property. Appendix B shows that,
by making appropriate approximation in the low-SNR region,
the test in (65) can also be simplied as a function of the sample
correlation coefcients. In particular, the simplied test is

T ( x) =
L
p
1

m=1
[
m
[
2
(66)
where
m
represents the sample correlation coefcient corre-
sponding to a delay of m samples, which is given as

m
=

N
n=1

L1
i=0
x
n
(i)x

n
((i m) mod L)

N
n=1

L1
i=0
[x
n
(i)[
2
. (67)
It is demonstrated in Section VII that the performance of the
preceding simplied test is very close to that of the constrained
GLRT in Section V.
C. Combination of CPCC- and MPCC-Based
Detection Algorithms
As discussed at the end of Section III, the full multipath and
cyclic correlations can be jointly considered in one covariance
matrix. However, the success of the constrained GLRT algo-
rithms introduced in Sections IV and V with a nite sample
size and at low-SNR value lies in the structural constraints
of their covariance matrices. A simple but effective approach
to combine multipath and cyclic correlations is to decide H
1
whenever one of the two constrained GLRT algorithms detects
the presence of the primary user. The combined test always
yields the best performance between the two detection algo-
rithms in each channel realization. It should be noted that the
threshold values have to be selected in such a way that the
overall probability of false alarm meets the required constraint.
VI. DETECTION ALGORITHM FOR UNSYNCHRONIZED
ORTHOGONAL FREQUENCY DIVISION
MULTIPLEXING SIGNALS
As pointed out before, the spectrum-sensing algorithms
presented in Sections IV and V require symbol timing synchro-
nization between the secondary and primary users. In the ab-
sence of symbol timing synchronization, the cyclic correlation
is taken into account by considering samples located within the
lags of L[11]. The consequence of this approach is a decrease
in the correlation coefcient to = (L
p
/L +L
p
) [11], which
causes a drop in the performance of the detection algorithm.
At the secondary users receiver, the received signal samples
are divided into blocks of L samples each. This means that
the corresponding samples in adjacent blocks are correlated
due to the CP [11]. Compared with the received signal model
Fig. 1. Timing relation between transmitter and receiver in the unsynchro-
nized case.
in the synchronized case [see, e.g., (5)], the receiver in the
unsynchronized case does not know exactly when an OFDM
block will start. As such, the timing index in this section is
with reference to the time the secondary users receiver starts
to collect the receive signal samples. In general, the timing
origin at the secondary users receiver can be lead or lag over
the timing origin at the transmitter by samples, as shown in
Fig. 1. For N transmitted OFDM blocks, the number of sample
blocks processed by the receiver is

N = N(L +L
p
)/L|.
To develop an efcient GLRT-based detection algorithm for
unsynchronized OFDM signals, consider only the last portion
of

L = L L
p
+ 1 samples in each block of L samples (see
Fig. 1), which can be represented as
x
n
= [ x
n
(L 1), . . . , x
n
(L
p
1)]

=

hs
n
+ v
n
n = 1, . . . ,

N. (68)
In the preceding expression,

h is the

L L Toeplitz
channel matrix with the rst row [h
1
, . . . , h
L
c
, 0, . . . , 0] and
rst column [h
1
, 0, . . . , 0]

. The corresponding length-L

transmitted signal vector is denoted by s
n
= [ s
n
(L 1),
. . . , s
n
(0)]

, n = 1, . . . ,

N. It is important to emphasize
that, due to unsynchronization, s
n
does not necessarily
align with the transmitted OFDM blocks, and it generally
contains data symbols from two consecutive transmitted
OFDM blocks. It also has to be noted that the reason for
employing the signal model in (68) in our analysis is to
efciently exploit the correlation among the transmitted
signals due to the presence of CP. Based on our model, all the
corresponding samples in the neighboring transmitted vectors,
i.e., s
n
and s
n1
, are correlated with the average correlation
coefcient = E s
n
(i) s

n1
(i)/E[ s
n
(i)[
2
= L
p
/L +L
p
.
Consequently, the corresponding samples in the neighboring
received vectors, i.e., x
n
and x
n1
, are correlated with the
correlation coefcient = E x
n
(i) x

n1
(i)/E[ x
n
(i)[
2
=
(
2
S

L
c
i=1
[h
i
[
2
/
2
S

L
c
i=1
[h
i
[
2
+
2
v
) = (SNR/SNR + 1).
To take the correlation between the neighboring vectors x
n
into account, dene the length-2

L vectors x
n
= [ x

n
, x

n1
]

,
n = 2, . . . ,

N. The covariance matrix of x
n
is expressed and
approximated as
R
x
=E
_
x
n
x
H
n
_
=

h
2
S
_
I
L
I
L
I
L
I
L
_

h
H
+
2
v
I
2L

=
_
R
x
R
x
R
x
R
x
_
. (69)
866 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 60, NO. 3, MARCH 2011
The matrix R
x
is an

L

L Hermitian Toeplitz matrix, and
it can be described by its rst row [
0
, . . . ,
L
p
1
, 0, . . . , 0]
and its rst column [

0
,

1
, . . . ,

L
p
1
, 0, . . . 0]

. Let x =
[ x
2
, . . . , x

N
] and x = [ x
1
, . . . , x

N
]. To facilitate matrix ma-
nipulations and parameter estimations, the correlation among
adjacent x
n
s is ignored. This allows one to apply the general
GLRT test in (23) to the observation x by substituting y = x,
M = 2

L, and N =

N 1. The resulting test is

T
G
( x)

=
1
2

L(

N1)
tr( x x
H
)
det
1
2

L
(

R
x
)
=
1

L

N
tr( x x
H
)
_
(1

2
) det
1/

L
(

R
x
)
H
1

H
0

(70)
where

R
x
,

R
x
, and

are estimates of R
x
, R
x
, and , respec-
tively. Based on (70), both R
x
and need to be estimated.
Since obtaining the ML estimate of R
x
appears to be very
cumbersome, the entries of R
x
shall be estimated by the sample
correlation coefcients as follows:

m
=
1

N

L

n=1
L

i=m
x
n
(i) x

n
(im)
+

N1

n=1
m1

i=1
x
n
(i) x

n1
(L+im)

, m=0, . . . , L
p
1. (71)
To obtain the ML estimate of , the log-likelihood function
of x is approximately expressed as
ln f
x[H
1
,R
x
( x[H
1
, R
x
)

= (

N 1)

L ln(2)
(

N 1)

L ln (det(R
x
))

n=1
x
H
n
R
1
x
x
n
. (72)
It can be easily veried that
R
1
x
=
1
1
2
_
R
1
x
R
1
x
R
1
x
R
1
x
_
. (73)
Using the preceding expression and det(R
x
) = (1

2
)

L
det
2
(R
x
), (72) can be equivalently expressed as

( , R
x
)=(

N1)

Lln(2)(

N1)

Lln(1
2
)
2(

N1) ln(R
x
)
1
1
2

N1

n=2
_
x
H
n
R
1
x
x
n
x
H
n1
R
1
x
x
n
x
H
n
R
1
x
x
n1
+ x
H
n1
R
1
x
x
n1

. (74)
Solving

( , R
x
)/ = 0 gives
2 (

N1)

L
(1
2
)

2
(1
2
)
2

n=2
_
x
H
n
R
1
x
x
n
x
H
n1
R
1
x
x
n
x
H
n
R
1
x
x
n1
+ x
H
n1
R
1
x
x
n1

+
1
(1
2
)

n=2
_
x
H
n1
R
1
x
x
n
+ x
H
n
R
1
x
x
n1

=0. (75)
To simplify (75), we can make use of the following approxi-
mation under the assumption of sufciently large

N:
1
2(

N 1)

n=2
x
H
n
R
1
x
x
n
+ x
H
n1
R
1
x
x
n1

=
1
2(

N 1)

L
tr

R
1
x

n=2
x
n
x
H
n
+

n=2
x
n1
x
H
n1

=
1

L
tr
_
R
1
x

R
x
_

= 1 (76)
where

R
x
=(1/2(

N 1))(

N
n=2
x
n
x
H
n
+

N
n=2
x
n1
x
H
n1
)
represents the sample covariance matrix of xs. Using (76), (75)
can be simplied to
+
2
(1
2
)

1+

2(

N1)

n=2
_
x
H
n1
R
1
x
x
n
+ x
H
n
R
1
x
x
n1

+
1
2(

N1)

n=2
_
x
H
n1
R
1
x
x
n
+ x
H
n
R
1
x
x
n1

=0. (77)
It can be shown that the following solution satises (77):

=
1
2(

N 1)

n=2
_

x
H
n1
R
1
x

x
n
+

x
H
n
R
1
x

x
n1
_
. (78)
In summary, after nding

R
x
based on (71),

can be
obtained from (78). The results are then used in (70) to realize
the test.
VII. SIMULATION RESULTS
The simulation parameters are chosen similarly to those
in [11]. In particular, the primary users OFDM system has
L = 32 subcarriers and transmits i.i.d. 16-QAM symbols with
normalized unit power. The detection period is taken to be
equal to N = 100 OFDM blocks, and the results are averaged
over 1000 random realizations of a Rayleigh multipath fading
channel. Except for Fig. 9, the channel coefcients are i.i.d.
complex Gaussian random variables. The case of correlated
channels is considered for Fig. 9. Note that, for an OFDM
system having a bandwidth of 5 MHz, 32 subcarriers, and a
CP length of 8 (similar to [11]), the sensing time is roughly
((32 + 8)/5 10
6
) 100 = 8 10
4
s or 0.8 ms. The per-
formance of different spectrum-sensing algorithms is evaluated
and compared via the probability of detection P
d
for a constant
false alarm rate of P
f
= 0.05.
First, Fig. 2 compares the detection performance of the
energy detector (ED) and three spectrum-sensing algorithms
developed and analyzed in this paper under perfect synchro-
nization assumption, i.e., CPCC-based algorithm (Section V-A
and [11]), multipath correlation-based constrained GLRT
(MP-based C-GLRT algorithm, Section V-A), and the simpler
BOKHARAIEE et al.: BLIND SPECTRUM SENSING FOR OFDM-BASED CR SYSTEMS 867
Fig. 2. Performance comparison of C-GLRT and U-GLRT spectrum-sensing
algorithms (L
c
= L
p
= 8).
Fig. 3. Performance comparison of C-GLRT and U-GLRT spectrum-sensing
algorithms (L
c
= L
p
= 8).
MPCC-based algorithm (Section V-B). For this particular g-
ure, L
p
= L
c
= 8 is used. As pointed out before, the ED algo-
rithm requires a precise knowledge of the noise variance, and a
small noise uncertainty, e.g., 0.5 and 1.0 dB, causes huge per-
formance degradation, as can be seen from the gure. In con-
trast, the three other algorithms are completely blind, and their
performances are impressive. Note that the simplied MPCC-
based algorithm performs closely to the MP-based C-GLRT
algorithm, and both of them clearly outperform the CPCC-
based algorithm. This superior performance is expected since,
with a large number of channel taps (L
c
= 8), it would be more
benecial to exploit multipath correlation than CP correlation.
Next, Fig. 3 shows performance improvement of the con-
strained GLRT algorithms over their unconstrained counter-
parts, both with multipath correlation and CP correlation. For
this gure, we also set L
p
= L
c
= 8. Recall that the U-GLRT
algorithm is basically (23), except that it uses only CP por-
tions of the observations in the CP-based algorithm or ISI-free
Fig. 4. Effect of L
p
and L
c
on the performance of spectrum-sensing
algorithms.
portions of the observations in the multipath-based algorithm.
In both cases, the improvements in detection performance are
very large.
Fig. 4 shows the detection performance of both MP- and
CPCC-based C-GLRT algorithms under different values of L
p
and L
c
. Observe that the performance of MP-based C-GLRT
changes very little between L
p
= L
c
= 4 and L
p
= L
c
= 8.
This can be explained as follows: While a bigger value of L
c
is desirable in terms of having stronger correlation property,
there is a larger number of quantities to estimate with the
same size of observations. These opposing effects appear to
cancel out in the scenarios considered in Fig. 4. In contrast,
the performance of CPCC-based algorithm can be signicantly
improved by increasing the length of the CP (comparing the
settings of L
p
= L
c
= 4 with those of L
p
= 8, L
c
= 4). Such
a performance improvement is obviously expected but comes
at the expense of additional resources since, as far as ISI
avoidance is concerned, it is desirable to use the minimum CP
length of L
p
= L
c
= 4. Furthermore, performance degradation
of the CPCC-based C-GLRT algorithm in an environment with
higher channel taps (rich multipath environment) can also be
observed by comparing the curves with L
p
= 8, L
c
= 4 and
L
p
= L
c
= 8. Such an observation agrees with the analysis in
(45) in Section IV-B.
Fig. 5 shows the convergence behavior of the probability of
detection of the proposed MP-based C-GLRT algorithm as a
function of N. As can be seen from the gure, for every SNR
value, the probability of detection can be made arbitrarily close
to 1 by allowing sufcient observations.
Fig. 6 shows the performance of the combined algorithm
(discussed in Section V-C) for the case L
c
= L
p
= 8. It is
seen that the combined sensing algorithm outperforms both the
MP- and CPCC-based C-GLRT algorithms. Such a result is
as expected since the combined algorithm takes into account
all the observations (both ISI-free and CP portions of received
blocks) and, at the same time, benets from the covariance
structures exploited in MP- and CPCC-based algorithms.
868 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 60, NO. 3, MARCH 2011
Fig. 5. Performance of MP-based C-GLRT algorithm with respect to obser-
vation size N.
Fig. 6. Performance of the combined algorithm (L
p
= L
c
= 8).
Fig. 7 shows the performance of the MP-based C-GLRT and
CPCC-based algorithms when the synchronization algorithm in
[18] is performed rst over the sensing interval. The normalized
residual CFO is set to 0.5, which is the worst case. Note that the
types of sensing algorithm are distinguished by different line
styles, whereas the combinations of imperfect/perfect timing
synchronization (indicated as I-Syn and P-Syn) and CFO
values are identied by different markers. The CFO introduces
a phase shift to the time-domain samples of an OFDM signal
[18]. Nevertheless, the circularity of the received signal co-
variance matrix is preserved in this case, and the MP-based
C-GLRT algorithm is not affected by a CFO. However, the
CP correlation coefcient becomes a complex value when
frequency offset is present. The effect of CFO on the CPCC-
based spectrum-sensing algorithm can be compensated by con-
sidering the magnitude of the sample correlation coefcient
in the CPCC-based algorithm [18]. As shown in Fig. 7, the
performance of sensing algorithms remains unchanged in the
presence of CFO. On the other hand, it is seen that the imperfect
Fig. 7. Performance of the spectrum-sensing algorithms in the presence
of residual time and CFOs (L
c
= L
p
= 8). Note that the types of sensing
algorithm are distinguished by different line styles, whereas the combina-
tions of imperfect/perfect timing synchronization (indicated as I-Synch and
P-Synch) and CFO values are identied by different markers.
Fig. 8. Performance comparison of spectrum-sensing algorithms in synchro-
nized, imperfectly synchronized, and unsynchronized transmission scenarios.
symbol timing causes performance degradation in both algo-
rithms. For the MP-based C-GLRT in the presence of timing
offset, the degradation is due to the fact that the ISI part
cannot be perfectly removed; therefore, the covariance matrix
of the received signal is not truly circulant. The time offset
also degrades the performance of the CPCC-based algorithm
since the identical head and tail of the OFDM signal cannot be
perfectly retrieved at the receiver.
Fig. 8 compares the performance of the proposed sensing
algorithms for perfectly synchronized, imperfectly synchro-
nized, and unsynchronized OFDM signals, which are labeled
in the gure as P-Syn, I-Syn, and Unsyn, respectively. In
particular, the combined algorithm in Section V-C is compared
against the algorithm in (70) for two different numbers of
channel taps, i.e., L
c
= 8 and L
c
= 1, as well as when the
CP length is set to L
p
= 8. Observe the opposite performance
BOKHARAIEE et al.: BLIND SPECTRUM SENSING FOR OFDM-BASED CR SYSTEMS 869
Fig. 9. Performance comparison of MP-based C-GLRT in a time-varying
Rayleigh fading channel (L
c
= L
p
= 8).
behaviors when the number of channel taps reduces from L
c
=
8 to L
c
= 1; the performance gets better for the synchronized
and imperfectly synchronized cases, whereas it gets worse for
the unsynchronized case. The latter behavior is, in fact, ex-
pected since performance degradation of the proposed detection
algorithm for unsynchronized OFDM signals is mainly due to
the reduction in the cyclic correlation coefcient. Because the
CP correlation strongly determines the performance of the de-
tection algorithm in a fading environment with a fewer channel
taps, the unsynchronized detection algorithm performs quite
poorly in such an environment. In contrast, in a rich multipath
environment, for example, with L
c
= 8, there is only a small
performance loss when detecting unsynchronized OFDM sig-
nals. In addition, in such an environment, the imperfectly syn-
chronized algorithm performs worse than the unsynchronized
algorithm. Considering the added complexity due to performing
synchronization, the algorithm developed for unsynchronized
OFDM signals in Section VI is a better candidate for spectrum
sensing in a rich multipath environment.
Finally, Fig. 9 shows the performance of MP-based C-GLRT
algorithmover time-varying Rayleigh fading channels when the
Doppler frequency f
d
is set to different values. Typical Doppler
shifts correspond to the mobile velocities of about 360 km/h.
If the system is operated in frequency bands of 24 GHz
(e.g., IEEE 802.11, IEEE 802.16, and IEEE 802.20), the
Doppler shifts are about 5200 Hz. It can be seen that, for the
typical Doppler frequencies, the multipath correlation is still
very benecial, and the sensing performance is not substantially
degraded, even for fast changes in channel taps.
VIII. CONCLUSION
In this paper, a spectrum-sensing method for OFDM-based
cognitive radio systems has been developed based on the
GLRT framework. The key feature in our development is to
explicitly take into account the structure (constraint) of the
covariance matrix of the underlying OFDM signal so that the
ML estimations of unknown parameters are improved, which
leads to robust and efcient spectrum-sensing tests. In partic-
ular, it has been shown that the CPCC-based test, which was
recently proposed in [11], can be obtained as a constrained
GLRT for an AWGN channel. It has also been shown that
the performance of CPCC-based test degrades in a multipath
environment. Moreover, by exploiting the multipath correlation
in the GLRT framework, an efcient test has been obtained,
which can be sequentially updated with any new reception
of OFDM symbol. A simplied MPCC-based test has also
been presented. Simulation results verify that both the CPCC-
and MP-based C-GLRT algorithms greatly outperform energy
detection in an environment with noise uncertainty. The MP-
based C-GLRT algorithm performs better than the CPCC-based
algorithm in a rich multipath environment. Furthermore, a
simple algorithm that combines both the CPCC- and MP-based
C-GLRT algorithms is suggested, which can further improve
the detection performance in a multipath environment.
While our studies have mainly focused on the detection of
synchronized OFDM signals, the developed algorithms can be
applied together with the synchronization algorithmin [18], and
they experience only a small performance loss due to imperfect
synchronization. Lastly, a simple GRLT-based algorithm has
also been proposed for the detection of unsynchronized OFDM
signals from the primary user. Simulation results demonstrated
satisfactory performance of such an algorithm in a rich multi-
path fading environment.
APPENDIX A
PROOF OF (45)
For a xed SNR and an i.i.d. channel with E[h
i
[
2
=
2
h
,
we have
SNR = E
_

2
S

2
v
L
c

i=1
[h
i
[
2
_
=
L
c

2
S

2
h

2
v
. (79)
First, rewrite ^
h
as
^
h
= 2
2
S

L
c
1

j=1
j

i=1
[h
i
[
2
+
L
p

j=L
c
j

i=1
[h
i
[
2

. (80)
It can be easily seen that
E^
h
=2
2
S

2
h
_
L
c
(L
c
1)
2
+ (L
p
L
c
)L
c
_
=2
2
v
SNR
_
2L
p
L
c
+ 1
2
_
(81)
ET
h
=T
h
= 2L
p

2
v
[SNR + 1]. (82)
Therefore
H
1
: = E
h
=
E^
h

T
h
=
SNR
1 + SNR
_
2L
p
L
c
+ 1
2L
p
_
.
(83)
870 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 60, NO. 3, MARCH 2011
APPENDIX B
DEVELOPING THE MULTIPATH CORRELATION
COEFFICIENT TEST
First, observe from (65) that
_
det(

R
x
)
_ 1
L
=e
ln
_
det( R
x
)
1
L
_
(84)
ln
_
det(

R
x
)
_ 1
L
=
1
L
ln
_
det(

R
x
)
_
=
1
L
L

k=1
ln(

k
) (85)
where

k
s are the eigenvalues of

R
x
. From (50), we have

k
=
0
+
L
p
1

m=1
Re
_

m
e

j2km
L
_
. (86)
Hence
ln(

k
)=ln

0
+
L
p
1

m=1
Re
_

m
e
j2km
L
_

=ln

1+

L
p
1
m=1
Re
_

m
e
j2km
L
_

0

=ln(
0
)+ln

1+

L
p
1
m=1
Re
_

m
e
j2km
L
_

0

. (87)
Since (

L
p
1
m=1
Re(
m
e
j2km/L
)/
0
) 1 for low SNR val-
ues, the second term of (87) can be expanded by keeping the
rst two terms of the Taylor series. That is
ln

1+

L
p
1
m=1
Re
_

m
e
j2km
L
_

0

L
p
1
m=1
Re
_

m
e
j2km
L
_

0

1
2

L
p
1
m=1
Re
_

m
e
j2km
L
_

0

2
.
(88)
Therefore
1
L
L

k=1
ln(

k
)

= ln
0
+
1
L
L

k=1

L
p
1
m=1
Re
_

m
e

j2km
L
_

0

1
2

L
p
1
m=1
Re
_

m
e

j2km
L
_

0

. (89)
The rst term inside the square brackets of (89) can be
simplied as follows:

L
p
1
m=1

L
k=1
Re
_

m
e

j2km
L
_

0
=

L
p
1
m=1
Re
_

m

L
k=1
e

j2km
L
_

0
= 0. (90)
The second term can be rewritten as (91), shown at the
bottom of the page.
Therefore, (89) can be approximated by
1
L
L

k=1
ln(

k
)

= ln
0

1
4
L
p
1

m=1
[
m
[
2

2
0
. (92)
Consequently
_
det(

R
x
)
_ 1
L

=
0
e

1
4

L
p
1
m=1

2
0
. (93)
L

k=1

L
p
1
m=1
Re
_

m
e

j2km
L
_

0

2
=
L

k=1

L
p
1
m=1
1
2
_

m
e

j2km
L
+

m
e
j2km
L
_

0

2
=
L

k=1

L
p
1
m=1

L
p
1
q=1
1
4
_

m

q
e

j2(k+q)m
L
+

q
e
j2(k+q)m
L
)
_

2
0
+
L

k=1

L
p
1
m=1

L
p
1
q=1
1
2
Re(
m

q
e

j2(kq)m
L
)

2
0
=

L
p
1
m=1

L
p
1
q=1
1
4

L
k=1
_

m

q
e

j2(k+q)m
L
+

q
e
j2(k+q)m
L
_

2
0
+

L
p
1
m=1

L
p
1
q=1
1
2
Re
_

m

L
k=1
e

j2(kq)m
L
_

2
0
= 0 +
1
4

L
p
1
m=1
[
m
|
2

2
0
=
1
4

L
p
1
m=1
[
m
|
2

2
0
(91)
BOKHARAIEE et al.: BLIND SPECTRUM SENSING FOR OFDM-BASED CR SYSTEMS 871
By noting that (1/L)(tr(

R
x
)) =
0
, the test statistics given
in (65) can be closely approximated as
T
G
( x)=
1
L
tr(

R
x
)
det(

R
x
)
1
L

=

0

0
e

1
4

L
p
1
m=1
|
m
|
2

0
2
=e
1
4

L
p
1
m=1

0
.
(94)
Taking the logarithm of the preceding test yields the follow-
ing equivalent test:

T ( x) =

L
p
1
m=1
|
m
|
2

2
0
. (95)
The ratio
m
=
m
/
0
can be easily obtained from (55) and
is given by

m
=

N
n=1

L1
i=0
x
n
(i)x

n
((i m) mod L)

N
n=1

L1
i=0
[x
n
(i)[
2
. (96)
Recall that
m
= Ex
n
(k)x

n
(k m)/Ex
n
(k)x

n
(k),
m = 0, . . . , L
p
1 is the correlation coefcients between the
ISI-free portions of the OFDM block. On the other hand,
m
represents the sample correlation coefcient corresponding to a
delay of m samples.
REFERENCES
[1] J. Ma, G. Y. Li, and B. H. Juang, Signal processing in cognitive radio,
Proc. IEEE, vol. 97, no. 5, pp. 805823, May 2009.
[2] H. A. Mahmoud, T. Yucek, and H. Arslan, OFDM for cognitive radio:
Merits and challenges, IEEE Wireless Commun. Mag., vol. 16, no. 2,
pp. 615, Apr. 2009.
[3] Z. Quan, S. Cui, A. H. Sayed, and H. V. Poor, Optimal multiband joint
detection for spectrum-sensing in cognitive radio networks, IEEE Trans.
Signal Process., vol. 57, no. 3, pp. 11281140, Mar. 2009.
[4] C.-H. Hwang, G.-L. Lai, and S.-C. Chen, Spectrum-sensing in wideband
OFDM cognitive radios, IEEE Trans. Signal Process., vol. 58, no. 2,
pp. 709719, Feb. 2010.
[5] S. M. Kay, Fundamentals of Statistical Signal Processing: Detection
Theory. Englewood Cliffs, NJ: Prentice-Hall, 1998.
[6] R. Tandra and A. Sahai, SNR walls for signal detection, IEEE J. Sel.
Topics Signal Process., vol. 2, no. 1, pp. 417, Feb. 2008.
[7] S. Kim, J. Lee, H. Wang, and D. Hong, Sensing performance of energy
detector with correlated multiple antennas, IEEE Signal Process. Lett.,
vol. 16, no. 8, pp. 671674, Aug. 2009.
[8] Y. Zeng and Y. C. Liang, Spectrum-sensing algorithms for cognitive
radio based on statistical covariances, IEEE Trans. Veh. Technol., vol. 58,
no. 4, pp. 18041815, May 2009.
[9] Y. Zeng and Y. C. Liang, Eigenvalue-based spectrum-sensing algorithms
for cognitive radio, IEEE Trans. Commun., vol. 57, no. 6, pp. 17841793,
Jun. 2009.
[10] F. Penna, R. Garello, and M. A. Spirito, Cooperative spectrum-sensing
based on the limiting eigenvalue ratio distribution in Wishart matrices,
IEEE Commun. Lett., vol. 13, no. 7, pp. 507509, Jul. 2009.
[11] S. Chaudhari, V. Koivunen, and H. V. Poor, Autocorrelation-based decen-
tralized sequential detection of OFDM signals in cognitive radios, IEEE
Trans. Signal Process., vol. 57, no. 7, pp. 26902700, Jul. 2009.
[12] J. Lunden, V. Koivunen, A. Huttunen, and H. V. Poor, Spectrum-sensing
in cognitive radios based on multiple cyclic frequencies, in Proc. Cogn.
Radio Oriented Wireless Netw. Commun., Aug. 2007, pp. 3743.
[13] M. Barkat, Signal Detection and Estimation. Norwood, MA: Artech
House, 1991.
[14] O. Bolkhovskaya and A. Maltsev, The performance of the GLRT for
the spatial signals detection with a small number of observations, IEEE
Signal Process. Lett., vol. 11, no. 10, pp. 841844, Oct. 2004.
[15] M. Novey, T. Adal, and A. Roy, Circularity and Gaussianity detec-
tion using the complex generalized Gaussian distribution, IEEE Signal
Process. Lett., vol. 16, no. 11, pp. 993996, Nov. 2009.
[16] A. Taherpour, S. Gazor, and M. Nasiri-Kenari, Invariant wideband
spectrum-sensing under unknown variances, IEEE Trans. Wireless
Commun., vol. 8, no. 5, pp. 21822186, May 2009.
[17] T. J. Lim, R. Zhang, Y. C. Liang, and Y. Zeng, GLRTBased spectrum-
sensing for cognitive radio, in Proc. IEEE Global Telecommun. Conf.,
Nov. 2008, pp. 15.
[18] J.-J. van de Beek, M. Sandell, and P. O. Borjesson, ML estimation of time
and frequency offsets in OFDM systems, IEEE Trans. Signal Process.,
vol. 45, no. 7, pp. 18001805, Jul. 1997.
[19] T. Hwang, C. Yang, G. Wu, S. Li, and G. Y. Li, OFDM and its wire-
less applications: A survey, IEEE Trans. Veh. Technol., vol. 58, no. 4,
pp. 16731694, May 2009.
[20] J. P. Burg, D. G. Luenberger, and D. L. Wegner, Estimation of structured
covariance matrices, Proc. IEEE, vol. 70, no. 9, pp. 963974, Sep. 1982.
[21] W. Zeng and G. Bi, Exploiting the multi-path diversity and multi-user
cooperation to detect OFDM signals for cognitive radio in low SNR with
noise uncertainty, in Proc. IEEE Global Telecommun. Conf., Dec. 2009,
pp. 16.
[22] W. Zeng and G. Bi, Robust detection of OFDM signals for cognitive
UWB in low SNR with noise uncertainty, in Proc. IEEE Int. Symp. Pers.,
Indoor, Mobile Radio Commun., Sep. 2008, pp. 15.
Simin Bokharaiee (S09) was born in Mashhad,
Iran, in 1980. She received the B.Sc. degree from
Sharif University of Technology, Tehran, Iran, and
the M.Sc. degree from Ferdowsi University of
Mashhad, both in electrical engineering. She is cur-
rently working toward the Ph.D. degree with the
Department of Electrical and Computer Engineering,
University of Manitoba, Winnipeg, MB, Canada.
Her current research interests include detection
and spectrum sensing in cognitive radio networks.
Ha H. Nguyen (M01SM05) received the B.Eng.
degree from Hanoi University of Technology, Hanoi,
Vietnam, in 1995, the M.Eng. degree from the
Asian Institute of Technology, Bangkok, Thailand, in
1997, and the Ph.D. degree from the University of
Manitoba, Winnipeg, MB, Canada, in 2001, all in
electrical engineering.
Since 2001, he has been with the Department of
Electrical Engineering, University of Saskatchewan,
Saskatoon, SK, Canada, where he is currently a
Full Professor. He holds adjunct appointments at the
Department of Electrical and Computer Engineering, University of Manitoba,
and TRLabs, Saskatoon, and was a Senior Visiting Fellow with the School
of Electrical Engineering and Telecommunications, University of New South
Wales, Sydney, Australia, during October 2007June 2008. He is a coauthor,
with E. Shwedyk, of the textbook A First Course in Digital Communications
(Cambridge University Press). His research interests include digital communi-
cations, spread spectrum systems, and error-control coding.
Dr. Nguyen is a Registered Member of the Association of Professional Engi-
neers and Geoscientists of Saskatchewan. He currently serves as an Associate
Editor for the IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS and
the IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY.
Ed Shwedyk received the B.Sc. (E.E.) and M.Sc.
(E.E.) degrees from the University of Manitoba,
Winnipeg, MB, Canada, in 1965 and 1968, respec-
tively, and the Ph.D. degree from the University of
New Brunswick, Saint John, NB, Canada, in 1974.
He was with the Department of Electrical and
Computer Engineering, University of Manitoba,
from 1974 to 2005, retiring as a full Professor.
His research interests are digital communications,
principally estimation and detection, and biosignal
processing.