Вы находитесь на странице: 1из 4

Centralized and Distributed Spectrum Sensing

System Models performance analysis based on


three users
Fang Chen

Runhe Qiu

Institute of Information Science and Technology


Donghua University
Shanghai, China
Email: fangchen_1@yahoo.com.cn

Institute of Information Science and Technology


Donghua University
Shanghai, China
Email: qiurh@dhu.edu.cn

AbstractOn the condition of scarce spectrum resources, the


cognitive radio, as the next generation of radio technology, has
incomparable superiority. Cooperative spectrum sensing
cognitive radio has two kinds of systems: the centralized system
and the distributed system. In this paper, we propose the
centralized and the distributed spectrum sensing system models
based on three users. The theoretical derivation and analysis
about two models are conducted. Moreover, we simulate and
compare the performance of two models. Simulation results show
that when the report error probability is small, the overall
performance of distributed spectrum sensing is superior to a
centralized spectrum sensing, but the distributed cooperative
spectrum sensing in the structure is more complex.
Keywords-Cognitive radio; Cooperative spectrum sensing; The
centralized system; The distributed system

I. INTRODUCTION
Cognitive radio[1] is an intelligent wireless communication
system. It can sense the surrounding environment and use the
"understanding to building" approach to obtain information
from the surrounding environment. In addition the transmission
parameters are changed in real time to adapt to variation in
environment. Cognition and reconstruction are two major
features of cognitive radio. Spectrum sensing is an important
prerequisite for cognitive radio communication systems.
Cognitive radio has been able to sense changes in the
surrounding environment, because cognitive radio can find the
spectrum holes [2] through the spectrum sensing function and
enable the cognitive radio users be adapted to the surrounding
environment.
There are two ways [3] in cooperative spectrum sensing
cognitive radio: centralized and distributed system. For the
distributed system, fixed relay, optional relay and incremental
relay strategies are proposed by Laneman, Tse and Wornell
[4]. Babarossa and P. A. Anghel [5] [6] have explored the
application of space-time code to the cooperative diversity over
white Gaussian noise channel and Rayleigh fading channel.
Space-time block coding and convolutional coding are
proposed by Yulong Zou and Baoyu Zheng to complete the

cooperation diversity [7]. For the centralized system, Shanker


S. [8] has proposed the deployment of two different networks:
a sensor network deployed in the target area and a data
transmission running network. The sensors through
cooperation among nodes sense the spectrum and the central
controller together and deal with the information gathered from
sensors. The calculation formulas of the false alarm probability
and detection probability under the OR and AND criteria are
shown by Liang, Yi-Chang, Peh, Edward [9]. Under the two
criteria detection performance were compared and they put
forward the optimal information fusion criterion is the ChairVarshney Guideline. Hongjian Sun, DI Laurenson and JS
Thompson [10] proposed a novel strategy for centralized
spectrum sensing, and simulation results show that the strategic
performance is better under the same conditions.
The previous analysis are based on the same collaborative
manner. Based on three users system models, this paper
discusses and analyzes spectrum sensing under the two kinds of
collaborative approaches in the same system environment.
This paper is organized as follows. In Section II, we
describe two system models under two different kinds of
collaborative approaches. Simulation results and analysis of the
two system models are given in Section III. Finally,
conclusions are drawn in Section IV.
II. SYSTEM MODELS
Collaborative detection refers to merge the information
from multiple cognitive radio users and then to determine
whether the first user (authorized users) exists, so the detection
can greatly reduce uncertainty by a single user. Moreover in
severe shadow environment, we can decrease the multi-path
fading and shadow effect. To prevent "hidden terminal" at the
same time, the spectrum detection probability is improved.
Therefore collaborative detection is more accurate than the
single local spectrum detection [11]. In a distributed
collaborative sensing system, the users involved in the
collaborative detection exchange information from each other,
and then the final decision is done by one user. However in the

978-1-4244-3709-2/10/$25.00 2010 IEEE

centralized collaborative sensing system, Sensing Information


Fusion Center (wireless access point or base station) will
aggregate the local spectrum sensing results, and make a final
decision.
To analyse cooperation spectrum sensings detection
probability in Rayleigh fading channel, Energy Detection
Algorithm is selected as the local spectrum detection method
[12]. Local spectrum detection methods include the matched
filter detection, cyclostationary feature detection and energy
detection. For the matched filter detection algorithm although it
can maximize signal to noise ratio and the time required is
relatively small when achieving a higher processing gain, it
need to know a priori knowledge of the main users signal and
match a special receive for different user types, which makes
the matched filter algorithm be difficult to apply in the actual
cognitive radio networks. Cyclostationary feature detection
method can distinguish the noise energy and the main users
energy, but the computing complexity and too long observing
time become a major bottleneck in application. Energy
detection algorithm is relatively simple and easy to implement,
while do not need to know a priori knowledge of the signal.
Thus this paper uses detection probability and false alarm
probability formulas in the Rayleigh channel [12] in the
following analysis.
A. A centralized spectrum sensing system analysis
In figure 1, a centralized spectrum sensing system model
based on the three users is presented. Three spectrum sensing
users CR1, CR2, CR3 are on an equal footing, which will own
the local decision information about PU (the first user) to the
information fusion center. This paper selects BS (base station)
as an information fusion center to deal with the three
judgments information from cognitive radio users.

saving way. According to detection probability pd , ray ,


CR1,CR2,CR3 send the local judgments H0CR or H1CR to base
station. The BS gets BS0CR (first user does not exist) or BS1CR
(first user exists) by decoding. In the process of sending H0CR
or H1CR to BS, H0CR or H1CR will be disturbed by the influence
of fading and noise. In this paper, pn is report error probability
between base station and the cognitive radio users. Assuming
that the channel between base station and the cognitive radio
users is binary symmetric channel, then we obtain

Pn = P ( BS 0CR \ H 1CR ) = P ( BS1CR \ H 0CR )

(1)

H0PU and H1PU mean that the first user does not exist and exist
in actual situation respectively. Using the property of Markov
chain, we obtain

P ( BS0CR \ H1CR , H1PU ) = P ( BS 0CR \ H1CR ) = Pn


P ( BS0CR \ H 0CR , H1PU ) = P ( BS0CR \ H 0CR ) = 1 - Pn

(2)
(3)

From (2) and (3), we can obtain

P ( BS 0CR \ H1PU ) = P ( BS0CR \ H 0CR , H1PU ) P( H 0CR \ H1PU )


+ P ( BS0CR \ H1CR , H1PU ) P( H1CR \ H1PU )
= Pn Pd ,ray + (1 Pn )(1 Pd ,ray )
(4)

P ( BS 0CR \ H 0PU ) = P ( BS0CR \ H1CR , H 0PU ) P( H1CR \ H 0PU )


+ P ( BS0CR \ H 0CR , H 0PU ) P( H 0CR \ H 0PU )
= Pn Pf ,ray + (1 Pn )(1 Pf ,ray )
(5)
Assuming that sensing ability of three users are in the same
level, from (4) (5) we can obtain the detection probability pd
and false alarm probability p f of a centralized spectrum
sensing system based on three users.

[
= 1 [P P

]
)]

Pd = 1 P Pd ,ray + (1 Pn ) (1 Pd ,ray )

Pf

Figure 1. A centralized spectrum sensing system model based on three users

To analyze conveniently, we suppose CR1,CR2,CR3 have


the same detection probability and false alarm probability,
which we call pd , ray and p f , ray respectively on the condition
of Rayleigh fading. The control channel sends the local
spectrum detection results to the fusion center is band- limited.
Assuming that the results of the local ruling H0CR (first user
does not exist) or H1CR (first user exists) with one bit to
represent, and this is the minimum cost and most bandwidth-

f ,ray

+ (1 Pn ) (1 Pf ,ray

(6)
(7)

B. A distributed spectrum sensing system analysis


A distributed spectrum sensing system model based on the
three users is shown in figure 2. The function of base station is
weakened, the spectrum sensing function is realized through
the exchange of information among three sensing users. In
order to simplify the analysis, this paper only selects CR1 and
CR3 as the relay of CR2. CR2 acts as information fusion center
to calculate the detection probability and false alarm
probability in order to analysis and compare with the result of
the centralized system.

As the same as the centralized system, energy detection


algorithm is selected as the local spectrum detection in the
Rayleigh fading channel. pn is report error probability among
the cognitive radio users, assuming that the channel between
the cognitive radio user CR1(or CR3) and the cognitive radio
user CR2, is binary symmetric channel, CR1,CR2,CR3 have
the same detection probability and false alarm probability,
which we call pd , ray and p f , ray respectively on the condition
of Rayleigh fading.

Assuming that sensing ability of three users are in the same


level, From (11) (12) we can obtain the detection probability
pd and false alarm probability p f of a distributed spectrum
sensing system based on three users.

{ [
{1 [Pn P

P d = P d ,ray 1 Pn P d ,ray +(1 Pn) (1 P d ,ray )


Pf = Pf ,ray

f ,ray

+ (1 Pn) (1 Pf ,ray

] } (13)
)] } (14)
2

III. SIMULATIONS AND RESULTS


Based on the above system models analysis, this part
presents simulation results. Figure 3 depict the performance of
the false alarms probability about the two systems, we can find
the false alarm probability of the distributed system is higher
than the centralized system when the same local false alarm
probability p f , ray is selected . Let pn =0.0001.

Figure 2. A distributed spectrum sensing system model based on three users

In a three users distributed sensing system, CR2 is selected


as information fusion center. Assuming that the results of the
local ruling H0CR (first user does not exist) or H1CR (first user
exists) with one bit to represent too. According to detection
probability pd , ray , CR1and CR3 send the local judgments H0CR
or H1CR to CR2. CR2 gets CR2 0CR (first user does not exist) or
CR21CR (first user exists) by decoding. In the process of
sending H0CR or H1CR to CR2, H0CR or H1CR will also be
disturbed. So we can obtain
CR
CR
CR
Pn = P (CR 2 CR
0 \ H 1 ) = P (CR 21 \ H 0 )
PU

(8)

PU

H0 and H1 mean that the first user does not exist and exist
in actual situation respectively. Using the property of Markov
chain, we obtain

P (CR 2

CR
0

\H

CR
1

,H

PU
1

) = P(CR 2

CR
0

\H

CR
1

) = Pn

CR
PU
CR
CR
P (CR 2CR
0 \ H 0 , H 1 ) = P (CR 2 0 \ H 0 ) = 1 - Pn

(9)
(10)

From (9) and (10), we can obtain


PU
CR
CR
PU
P (CR 2CR
0 \ H 1 ) = P (CR 2 0 \ H 0 , H 1 )
CR
PU
CR
P ( H 0CR \ H1PU ) + P(CR 2CR
\ H1PU )
0 \ H1 , H1 ) P ( H1

= Pn Pd ,ray + (1 Pn )(1 Pd ,ray )


(11)
PU
CR
CR
PU
P (CR 2CR
0 \ H 0 ) = P (CR 2 0 \ H 1 , H 0 )
CR
PU
CR
PU
P ( H1CR \ H 0PU ) + P(CR 2CR
0 \ H 0 , H 0 ) P( H 0 \ H 0 )

= Pn Pf ,ray + (1 Pn )(1 Pf ,ray )


(12)

Figure 3. The false alarm probability analysis of two system

Figure 4 shows the performances of the detection


probability about the two systems, we also let pn =0.0001. We
can find the detection probability of the centralized system is
lower than the distributed system when the same local
detection probability pd , ray is selected.
In Figure 5, the average SNR of local detection is 10db, a
interested bandwidth multiplied by observing time is 5. pn is
equal to 0.0001,0.001,0.01,0.1 respectively, the corresponding
diagram is between false alarm probability and detection
probability of two kinds of systems. The distributed spectrum
sensing system is presented by DSSS. The centralized
spectrum sensing system is presented by CSSS. When pn =
0.0001, the false alarm probability in the large range of 0.0003
to 1, the distributed system sensing performance curve is
higher than the centralized system sensing performance curve.
Moreover the difference of corresponding detection probability
is relatively large. When pn is equal to 0.0001,0.001,0.01,0.1,
respectively, the centralized sensing system performance

curves change little. Especially when pn is equal to 0.0001 and


0.001 respectively, the centralized system performance curves
are almost unchanged. However the distributed system sensing
performance curves have changed greatly. This shows that the
centralized spectrum sensing system is relatively stable in
sensing network, while the distributed structure is vulnerable to
external interference and make the wrong decision. On the one
hand as pn increases, the detection performance of two
spectrum sensing systems are reduced. Especially for the
distributed spectrum sensing system, this effect is more
apparent. On the other hand, when pn is small, detection
performance of the distributed spectrum sensing system is
superior to a centralized spectrum sensing system. But when
pn increases, the advantage of the distributed system in
relation to the centralized system is decreased. This is because
the report error probability from the relay users to information
fusion center is increased. In a word, we can improve overall
system performance by reducing pn .

IV. CONCLUSION
In this paper, the centralized and the distributed spectrum
sensing system models based on three users are proposed.
Through simulation spectrum sensing system models are
analyzed and compared. The distributed system has a small
probability of missed detection and large probability of false
alarm. The centralized system is at the expense of the missed
detection probability in exchange for reducing false alarm
probability. In terms of the comprehensive performance, when
the report error probability is small. The distributed system is
superior to a centralized system. But for the specific systems,
the distributed system increases the complexity of the design.
Each sensing user has the ability of sharing and fusing the
information, which increases costs. On the other hand,
when pn increases sensing performance of the distributed
system would rapidly decline. How to reduce pn needs to be
paid more attention in the future.
REFERENCES
[1]

Figure 4. The detection probability analysis of two systems

Figure 5 The integrated performance analysis of two system

FCC, Facilitating Opportunities for Flexible,Eficient and Re-liable


spectrum Use Employing Cognitive Radio Technologies,ET Docket
No.03-332,2003.
[2] Hakin S. Cognitive radio:Brain-empowered wireless communications,
IEEE Journal on Selected Areas in Communications, 23(2):201220,2005.
[3] I.F.Akyiildiz, W.-Y. Lee, M.C Vuran, and S. Mohanty, Next
generation /dynamic spectrum access/cognaiive radio wireless
networks:a survey,Computer Networks,vol.50,pp.2127-2159,2006
[4]
J. N. Laneman, D. N. C. Tse, G. W. Wornell. Cooperative diversity in
wireless networks: Efficient protocols andoutage behavior, IEEE
Transactions on Inform Theory, vol.50, no.12, pp. 3062-3080, 2004.
[5]
S. Barbarossa, L. Pescosolido et al. Cooperative wireless networks
based on distributed space time coding, International Workshop on
Wireless Ad-Hoc Networks, June 2004,PP.30-34.
[6]
P. A. Anghel, M. Kaveh. On the diversity of cooperative system,
IEEE International Conference on ICASSP, vol. 4, 2004, pp. 577 580.
[7]
Yulong Zou, Baoyu Zheng, Xianglin Zhu.A New Cooperative
Diversity Scheme for Next Generation Wireless Network, 5th IEEE
Consumer Communications and Networking Conference, 2008. Las
Vegas, USA, Jan. 2008, pp. 938-942.
[8] Shanker S. Spectrum agile radios:utilization and sensing system, First
IEEE International Symposium on DySPAN,2005,PP.160-169.
[9] Liang,Yi-Chang;Peh,Edward,optimization for cooperative sensing in
cognitive radio network, Wireless communications and Networking
conference ,2007, pp:27-32.
[10] Hongjian Sun, D.I. Laurenson, J.S. ThompsonA Novel Centralized
Network for Sensing Spectrum in Cognitive Radio, IEEE International
Conference on Communications,2008,PP.4186-4190.
[11] Jia Zhu,Baoyu Zheng, Detection Probability Analysis of Cooperative
Spectrum Sensing in Rayleigh Fading Channels, 2009 Eigth
IEEE/ACIS International Conference on Computer and Information
Science,2009,PP.177-182.
[12] Fadel F. Digham, Mohamed-Slim Alouini, Marvin K. Simon,On the
Energy Detection of Unknown Signals Over Fading Channels,IEEE
Transactions on Communications, vol. 55, no.1, PP.21-24,Jan. 2007.

Вам также может понравиться