Professional Documents
Culture Documents
Spectrum Sensing Techniques: A Comparison Between Energy Detector and Cyclostationarity Detector
Spectrum Sensing Techniques: A Comparison Between Energy Detector and Cyclostationarity Detector
=
y T
e H y T P
(7)
B. Cyclostationarity Detector
In CD, the cyclostationarity properties of the received
signal are used to detect the primary signal. A signal x(t) is
said to possess cyclostationarity property if its statistics, mean
and autocorrelation, are periodic with some period T
0
.
p
x
(t) = p
x
(t + I
0
) (8)
R
x
(t, ) = R
x
(t + I
0
, ) (9)
From Equation (9), it is clear that the Autocorrelation
Function is periodic in nature. So, it can be represented as a
Fourier series [3].
R
x
() = lim
1-
1
1
] x [t +
:
2
x [t -
:
2
-
c
-]2nut
Jt
1
(10)
R
x
u
() = E|x(n)x
-
(n - )c
-]2nun
] (11)
where R
x
u
() is the Cyclic Autocorrelation Function (CAF)
and is the cyclic frequency.
S(, o) = R
x
u
()c
-2n]:
:=-
(12)
where S(, o) is called the Spectral Correlation Function
(SCF) or Cyclic Spectrum Density (CSD).
For detecting the primary signal, CD computes the SCF of
the received signal and compares it with the predetermined
threshold. Here also, the NP approach is used to fix the
threshold. It is also assumed that the SCF of the received
signal is Gaussian in nature.
There are mainly two features which makes the CD more
superior than ED [4]. They are:
1) It is possible to distinguish White Gaussian Noise
(WGN) from the received signal as WGN is a WSS
process. So, when 0, SCF is ideally zero.
2) Cyclostationarity features are different for different
signals. This enables signal classification.
) ( ) ); ( (
0
y dT H y T P P
fa
389
Generally, the SCF function of the received signal y(n) can
be expressed as:
I(y) = S
(, o) = _
S
w
(, o) E
0
S
x
(, o) + S
w
(, o) E
1
(13)
A series of algorithms are available to estimate the SCF of
a given signal. In this paper, an FFT based time smoothing
algorithm called the FFT Accumulation Method (FAM) is
used to compute SCF [5].
Fig. 2.Block diagram for the determination of test statistic (SCF) for
Cyclostationarity Detector (using FAM)
III. SIMULATION RESULTS
In this paper, the input BFSK, BPSK and GMSK signals
(with length N) are generated using Simulink. Here, ROCs of
Energy Based Detector and Cyclostationarity Based Detector
for both BFSK, BPSK and GMSK signals with variable SNR
are compared and robustness of each method is analyzed. The
performance of energy based detector against cyclostationarity
based detector under noise uncertainty is also evaluated.
The magnitude of SCF of BFSK and BPSK signals are
plotted in Fig. 5 and Fig. 6 respectively. The presence of
discrete peaks in the plot represents the frequency components
of the signals. The peaks at 0 are used for detecting the
presence of signal in a CD.
On comparing the ROCs of ED and CD for BFSK, BPSK
and GMSK signals respectively, we can observe that for lower
SNR values the value of P
d
is higher for CD when compared
to ED. This can be observed by comparing the Fig.3 with
Fig.4, Fig. 7 with Fig. 8 and Fig. 9 with Fig. 10.
Fig. 3.Energy Based Detector for BFSK signal (N=100)
Fig. 4. Cyclostationarity Based Detector for BFSK signal (N=100)
Fig. 5. Magnitude of SCF of BFSK Signal
Fig. 6. Magnitude of SCF of BPSK Signal
390
Fig. 7.Energy Based Detector for BPSK sign
Fig. 8. Cyclostationarity Based Detector for BPSK
Fig. 9 .Energy Based Detector for GMSK sig
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Pfa
P
d
Cyclostationarity Based Detector - BPSK signa
nal (N=100)
K signal (N=100)
gnal (N=100)
Fig. 10. Cyclostationarity Based D
The evaluation of the perfo
sensing technique is based on t
which that particular technique
P
fa
. To find this minimum SN
for a constant P
fa
. Fig. 11 and
results are summarized in Ta
uncertainty, the SNRs require
100% and 90% are -4 dB a
achieves the same values of
respectively. This means that
signals with P
d
90% when com
uncertainty, CD achieves 100%
dB respectively. But in this ca
becomes worse; it achieves t
-2.8 dB respectively. Similarly
dB, the performance of ED
compared to CD. Similarly,
compared and the results are su
Fig. 11. Pd vs SNR plot for Energy
0.8 0.9 1
l (N=100)
SNR = 0dB
SNR = -2dB
SNR = -4dB
SNR = -6dB
SNR = -8dB
SNR = -10dB
Detector for GMSK signal (N=100)
ormance of a particular spectrum
the value of minimum SNR with
e can achieve the required P
d
and
NR we use a plot of P
d
vs. SNR
d Fig. 12 are compared and the
able I. In the case of no noise
ed by a CD to achieve a P
d
of
and -6.8 dB respectively. ED
f P
d
at -0.8 dB and -3.75 dB
CD can detect 3.1 dB weaker
mpared to an ED. For 1 dB noise
% and 90% P
d
at -2 dB and -4.35
ase the performance of the ED
these values at 0.996 dB and
y, for a noise uncertainty of 3
D degrades furthermore when
Fig. 13 and Fig. 14 can be
ummarized in Table II.
y Based detector (BFSK) with Pfa=10%
391
Fig. 12. Pd vs SNR plot for Cyclostationarity Based detector (BFSK) with
Pfa=10%
TABLE I PERFORMANCE EVALUATION SUMMARY (BFSK)
BFSK SENSING
NOISE
UNCERTAINITY
SNR REQUIRED TO ACHIEVE 90% Pd
CYCLOSTATIONARITY
BASED DETECTOR
ENERGY
BASED
DETECTOR
0dB -6.8dB -3.75dB
1dB -4.35dB -2.8dB
2dB -3.15dB -1.6dB
SNR REQUIRED TO ACHIEVE 100% Pd
0dB -4dB -1dB
1dB -2dB 0.996dB
2dB -0.01dB 1dB
Fig. 13. Pd vs SNR plot for Energy Based detector(BPSK) with Pfa=10%
Figure 14. Pd vs SNR plot for Cyclostationarity Based detector (BPSK) with
Pfa=10%
TABLE II PERFORMANCE EVALUATION SUMMARY (BPSK)
BPSK SENSING
NOISE
UNCERTAINITY
SNR REQUIRED TO ACHIEVE 90% Pd
CYCLOSTATIONARITY
BASED DETECTOR
ENERGY
BASED
DETECTOR
0dB -6.86dB -3.75dB
1dB -5.88dB -2.7dB
2dB -4.88dB -1dB
SNR REQUIRED TO ACHIEVE 100% Pd
0dB -4dB -0.8dB
1dB -3dB 0dB
2dB -2dB 1.5dB
IV. CONCLUSION
In this paper, we have analyzed the performance of ED and
CD for BFSK, BPSK and GMSK signals. From the simulation
results we can observe that in the case of high noise
uncertainty ED cannot be considered as an optimal detector.
But if we analyse the performance of CD for different noise
uncertainties we can observe that its performance is not much
affected by high noise uncertainty. Thus we can conclude that
even though Cyclostationarity based detector has a higher
computational complexity, its performance is superior to
Energy based detector in case of high noise uncertainty and so
Cyclostationarity based detector can be considered as a more
robust and reliable detector when compared to Energy based
detector.
ACKNOWLEDGEMENT
The authors thank Mr. S. Krishnakumar, Joint Director,
CDAC, Thiruvananthapuram and Mr. S. Sagar, Joint Director,
CDAC, Thiruvananthapuram for their support and guidance.
392
REFERENCES
[1] Simon Haykin, Cognitive Radio: Brain-Empowered Wireless
Communications -IEEE Journal on selected areas in
communications, Vol. 23, No. 2, February 2005.
[2] Steven M Kay, Fundamentals of Statistical Signal Processing
Vol II Detection Theory.
[3] Alfateh M. Mossa and VarunJeoti, Cyclostationarity- Based
Spectrum Sensing for Analog TV and Wireless Microphone
Signals, 2009 First International Conference on Computational
Intelligence, Communication Systems and Networks.
[4] W. A. Gardner, Cyclostationarity in Communications and
Signal Processing, New York: IEEE Press, 1993.
[5] EvandreLuiz da Costa, Detection and Identification of
Cyclostationary Signals, March 1996.
393