74 136-143

Comparative Analysis of Information Fusion
Techniques for Cooperative Spectrum Sensing in
Cognitive Radio Networks
O. P. Meena1
and Ajay Somkuwar2
Dept. of Electronics and Communication Engineering,
MANIT, Bhopal, M.P., India-462051
1
Email: opm@manit.ac.in
2
Email: ajaysomkuwar@manit.ac.in
Abstract— In Cognitive Radio Networks (CRN), Cooperative Spectrum Sensing (CSS) is
used to improve performance of spectrum sensing techniques used for detection of licensed
(Primary) user’s signal. In CSS, the spectrum sensing information from multiple unlicensed
(Secondary) users are combined to take final decision about presence of primary signal. The
mixing techniques used to generate final decision about presence of PU’s signal are also
called as Fusion techniques / rules. The fusion techniques are further classified as data
fusion and decision fusion techniques. In data fusion technique all the secondary users
(SUs) share their raw information of spectrum detection like detected energy or other
statistical information, while in decision fusion technique all the SUs take their local
decisions and share the decision by sending ‘0’ or ‘1’ corresponding to absence and presence
of PU’s signal respectively. The rules used in decision fusion techniques are OR rule, AND
rule and K-out-of-N rule. The CSS is further classified as distributed CSS and centralized
CSS. In distributed CSS all the SUs share the spectrum detection information with each
other and by mixing the shared information; all the SUs take final decision individually. In
centralized CSS all the SUs send their detected information to a secondary base station /
central unit which combines the shared information and takes final decision. The secondary
base station shares the final decision with all the SUs in the CRN. This paper covers
overview of information fusion methods used for CSS and analysis of decision fusion rules
with simulation results.
Index Terms— Cognitive radio, spectrum sensing algorithms, cooperative spectrum sensing,
data fusion, decision fusion.
I. INTRODUCTION
Dynamic spectrum sensing is a challenging and necessary task in Cognitive Radio Networks (CRN). It can
detect presence of PU who is having legacy right on licensed spectrum. The SU continuously or periodically
senses the PU’s spectrum and when it finds the spectrum idle, it starts transmitting its own data in the
spectrum. When the SU detects presence of the PU in the spectrum, it stops transmission or switches to
another idle frequency spectrum. The SU must maintain its transmission parameters like power level,
frequency band used for data transmission etc., in such a way that it must not cause any interference in PU’s
transmission [1].
DOI: 02.ITC.2014.5.74
© Association of Computer Electronics and Electrical Engineers, 2014
Proc. of Int. Conf. on Recent Trends in Information, Telecommunication and Computing, ITC

137
The existing spectrum sensing techniques can be broadly divided into three categories [1], [2]: energy
detection, matched filter detection, and cyclostationary detection. Among them, energy detection has been
widely applied since it does not require any a priori knowledge of the primary signals and has much lower
complexity than the other two schemes [3]. Hence the energy detection technique is considered in section III
of this paper during soft combination [4] and weighted combination techniques [5] for CSS.
In Cognitive Radio Networks fading, hidden nodes, shadowing effect etc. degrade performance of the
spectrum sensing techniques that may lead to false and missed detection of the spectrum by SUs that result in
missed opportunity and interference to PU respectively. Therefore to improve performance of the spectrum
sensing techniques, cooperative spectrum sensing is used. In the CSS, the spectrum sensing information from
multiple SUs is mixed to take final decision about presence of PU’s signal [6], [7]. The CSS is further
classified as distributed CSS and centralized CSS [1]. In distributed CSS all the SUs share the spectrum
detection information with each other and by mixing the shared information all the SUs take final decision
individually. In centralized CSS all the SUs send their detected information to a central unit/ base station of
secondary network which combines the shared information and takes final decision about presence of PU’s
signal as illustrated in Fig. 1. The base station of secondary network shares the final decision with all the
SUs. The mixing techniques used to generate final decision about presence of PU’s signal are also called
Fusion techniques / rules. The fusion techniques are further classified as data fusion and decision fusion
techniques. In data fusion techniques all the SUs share their detected raw information like detected energy
or other statistical information while in decision fusion technique all the SUs take their local decisions and
share the decision by sending ‘0’ or ‘1’ corresponding to absence and presence of PU’s signal respectively
[4]. The fusion rules at fusion centre can be OR, AND or majority rule which can be generalized as K-out-of-
N rule [8], [9], [10] respectively.
This paper is intended to provide a generic and comprehensive over view of fusion techniques used for
cooperative spectrum sensing in CRN, as well comparative analysis of decision rules with simulation results.
. The section II of the research paper covers different type of fusion techniques used in CSS. In section III,
system model and mathematical formulation is explained. The comparative analysis of various fusion
techniques and simulation results are covered in section IV. In section V, the discussion of fusion techniques
is concluded with future research scope in the field.
Figure 1. Illustration of CSS in CRN
II. INFORMATION FUSION TECHNIQUES IN CSS
In the CSS techniques, the spectrum sensing information from multiple SUs is mixed to take final decision
about presence of PU’s signal. Based on the way the sensed information is mixed, the CSS is classified as
distributed CSS and centralized CSS as shown in Fig. 2. In distributed CSS all the SUs share the spectrum
detection information with each other and by mixing the shared information all the SUs take final decision
individually. In centralized CSS all the SUs send their sensed information to a base station / central unit of
secondary network which combines the shared information and takes final decision about presence of PU’s
signal. The secondary base station shares the final decision with all the SUs in the CRN.
The fusion techniques are further classified as data fusion and decision fusion techniques [3] as shown in
Fig.3. In data fusion techniques, all the SUs share their detected raw information like detected energy or
other statistical information while in decision fusion technique all the SUs take their local decisions and
generally share the decision by sending ‘0’ or ‘1’ corresponding to absence and presence of PU’s signal
respectively. The decision fusion rules generally used at fusion centre are AND, OR and K-out-of-N rules.

138
Figure 2. CSS classification based on the way final decision achieved. Figure.3. Classification of Fusion techniques for CSS in CRN
A. Soft combination /data fusion technique
In soft combination technique, also known as data combination technique [4] in which the accurate sensed
energies from different SUs are combined to make a final decision. This technique needs larger overhead to
transmit the accurate sensed energies to the fusion center as compare to decision fusion technique. The
fusion strategy for soft combination is further classified as Equal Gain Combination (EGC) and Weighted
Gain Combination (WGC) [5]. In EGC, the fusion centre treats all the SU’s sensed data equally (assigns
equal gain) while in WGC the fusion centre assigns weights to sensed data in proportion to the level of
detected energy by the secondary users. The WGC gives better spectrum detection performance as compare
to EGC and hard combination techniques by assigning more weight to the SUs who detect higher energies.
B. Hard combination / decision fusion technique
Figure 4. Principle of one-bit hard combination scheme Figure 5. Principle of 2-bit hard combination scheme
In conventional 1-bit hard combination technique the detected energy of a SU is compared with a threshold
value and local decision is achieved about presence of the PU’ signal as shown in Fig. 4. The other spectrum
detectors like matched filter detector, cyclostationary detector etc., also can be used by a SU to achieve the
local decision. Conventionally the presence and absence of the PU’s signal is represented as ‘1’ and ‘0’
respectively. The SUs share their local decision with fusion centre to achieve the final decision about
presence of PU’s signal. The shared local decisions of the SUs are combined using any of the logical AND,
OR and K-out-of-N rules to take the final decision. This technique needs smaller overhead to share the local
decision to the fusion center as compare to soft combination technique [11].
Since the Soft combination technique gives better detection performance as compare to the hard combination
technique [4]. Therefore to improve the detection performance of hard combination technique the whole
range of the observed energy can be divided into more regions and allocated larger weights to the upper
regions and smaller weights to the lower regions. The more number of regions can be presented by
combination of multiple bits. In [4], a softened 2-bit hard combination technique shown in Fig. 5 gives better
performance than the 1-bit hard combination technique. Hence a softened K-bit hard combination can be

139
developed which would give better detection performance than the 1-bit and softened 2-bit hard combination
techniques but at the cost of increased overhead.
III. SYSTEM MODEL AND FORMULATION
In CSS, the spectrum sensing information from multiple secondary users is combined for primary user
detection. The SUs use any of the basic spectrum sensing method like energy detector, cyclostationary
detector, matched filter detector etc. [1] for local sensing of PU’s signal but the final decision is achieved by
mixing the information received from the SUs for cooperative sensing of PUs [11] as demonstrated in Fig 6.
The cooperative detection can be implemented either in a centralized (infrastructure based) or in a distributed
manner (infrastructure less) [1]. In the centralized method, the secondary base-station plays a key role to
collect sensing information from all the SUs but in the distributed method the sensed information shared
among SUs in the CRN and the SUs take their final decision individually. The benefits of CSS are reduced
time delay, improved agility and accuracy in spectrum detection [12], [13], [7].
For mathematical analysis of the CSS, let us assume that the received signal has the following form
y(n) = s(n) + w(n) (1)
Where y(n) is the received signal, s(n) is the signal to be detected, w(n) is the additive white Gaussian noise
and n is sample index. The decision metric for energy detector can be written as
M = |y(n)| (2)
By comparing the decision metric M with a threshold λE the spectrum occupancy decision is made. The
decision is made by distinguishing between following hypothesizes:
H0 : y(n) = w(n), (3)
H1 : y(n) = s(n) + w(n). (4)
In this paper, let pf, pd, and pm respectively denote the false alarm probability, detection probability and miss
detection probability for individual sensing decision. Where the pd is probability of detecting a primary user
signal on considered frequency spectrum when the signal is truly present by a secondary user. Thus a large
detection probability is desired. It can be formulated a
pd = Pr( M > λE|H1) . (5)
pf = Pr( M > λE|H0) (6)
The pf is false detection probability which indicates that the test incorrectly decides that the primary user
signal is present in the considered frequency spectrum. Hence the pf should be kept as small as possible in
order to prevent underutilization of transmission opportunities. The decision threshold λE can be selected in
such a way that gives optimum values of pd and pf.
The missed detection probability:
pm = 1- pd (7)
Let Pf, Pd and Pm denote corresponding probabilities of the final decision after information fusion. In
cognitive radio networks, SUs are required to sense the presence of the primary user’s signal as shown in Fig.
1. Cooperative spectrum sensing is modelled in Fig. 6, where SUs share their spectrum sensing information
with secondary base station which also works as fusion center in case of centralized CSS [11]. The CSS can
be illustrated as Fig. 6. The primary objective of CSS is to determine presence of the PU’s signal. As
mentioned in the above paragraph that the absence and presence of the primary signal is presented by H0 and
H1 hypotheses respectively and determined by analyzing the received signal Y. All secondary users make
individual judgment according to pre-setup threshold λ(.), and pass their judgment J to fusion centre. In case
of data fusion, J = actual sensed energy, while in case of decision fusion J= 0 or 1 corresponding to H0 and
H1. After collecting all judgments, the secondary base station utilizes the received judgments to make a final
decision Φ.

140
Figure 6. Cooperative spectrum sensing model
For decision fusion rule (DFR), the fusion centre makes final decision according to the number of SUs
claiming the presence of the PU’s signal. Let L denote the number of SUs claiming the presence of primary
user. In information fusion centre, the final decision strategy (.) is given by
=
, <
, ≥
(8)
Where K = 1, K = N and 1˂ K ˂ N are corresponding to OR-rule, AND-rule and K-out-of-N rule
respectively.
i. OR rule:
Like in logical OR operation, the output is high if at least one of the input is high. Similarly the fusion centre
also declares presence of the PU’s signal if at least one SU reports presence of the signal. The probability of
detection and false alarm detection for OR rule are given as:
P = ∑ 1 − (9)
P = ∑ (1 − ) (10)
ii. AND rule:
Like in logical AND operation, the output is high if all the inputs are high. Similarly the fusion centre also
declares presence of the PU’s signal if all the SUs report presence of the signal. The probability of detection
and false alarm detection for AND rule is given as:
= (11)
= (12)
iii. K-out-of-N rule :
In this case if K-out-of-N inputs are high of a logical operation, the corresponding output is high. Similarly
the fusion centre also declares presence of the PU’s signal if at least K out of N SUs report presence of the
signal. The probability of detection and false alarm detection for K-out-of-N rule are given as:
P = ∑ 1 − (12)
P = ∑ (1 − ) (13)
IV. COMPARATIVE ANALYSIS OF FUSION TECHNIQUES
In table 1, centralized and distributed cooperative sensing methods are compared. In centralized CSS, a
secondary base station is required for sensed information fusion to make final decision about presence of
primary signal while in distributed CSS base station is not required. Since in centralized CSS, the SUs share
their sensed spectrum information with the secondary base station and the base station replies the final
decision after the information fusion. While in distributed CSS, all the SUs take their final decision
themselves after shared information fusion. Hence distributed CSS is faster method because it consumes less
time as compare to the centralized CSS.

141
TABLE I, COMPARISON OF CENTRALIZED AND DISTRIBUTED CSS
Parameters Centralized CSS Distributed CSS
Secondary Base station required Yes No
Time delay to make final decision about presence of primary signal more less
Robustness slow fast
TABLE II, COMPARISON OF DATA AND DECISION FUSION TECHNIQUES
Parameters Data Fusion Decision Fusion
Type of information shared for spectrum
sensing
Raw information of spectrum
sensing
Local decisions of SUs
Loss of detected information No Yes
Size of required frame overhead Large Small
Weighted data fusion possible Not possible in one bit decision fusion but
possible in multi bit decision fusion [4]
In table 2, it’s clear that in data fusion technique the raw information of spectrum sensing is shared by SUs
and combined to make final decision about presence of PU’s signal which results in no loss of the sensed
information. The data fusion can support weighted combination of the raw information but need large frame
overhead during information sharing among SUs. In decision fusion, generally one bit is required to share
local decisions of SUs which leads to sensed information loss. In decision fusion, weighted combination is
not possible using one bit decision combination but possible with multi-bit decision combination.
For simulation, let total number of SUs in CRN are N=10, probability of detection of a SU pd and K is a
intermediate integer having values from 1 to N. The K is used in K-out-of-N fusion rule. If K=1 the fusion
rule becomes OR rule and for K=N the fusion rule becomes AND rule. For simplicity assuming that pd is
identical for all the SUs.
Figure 7. Comparative analysis of OR, AND, K-out-of-AND rules for K=3
Fig. 7 shows the probability of detection of fusion centre (Pd) versus the probability of detection of a
secondary user (pd), for K=3. For OR rule the probability of detection of fusion centre increases rapidly as
the probability of detection of a secondary user increases from pd = 0 to 1. But for AND rule the probability
of detection of fusion centre remains almost zero up to pd = 0.6 and increases rapidly as pd increases from 0.7
to 1. The probability of detection of the fusion centre for K-out-of-N rule falls between the probabilities of
the OR rule and the AND rule for all values of K.
Fig. 8 shows the probability of detection of fusion centre (Pd) versus the K for K-out-of-N rule for pd = 0.1,
0.5 and 0.9. For pd = 0.1 the Pd decreases rapidly as the K increases. But for pd = 0.9 the Pd remains almost 1
for K= 1 to 7 and decreases rapidly for K= 8 to N. For intermediate values of pd between 0.1 to 0.9, the Pd
decreases slowly and smoothly for increasing K.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Probability of detection (pd)of individual secondary user for K=3
Probabilityofdetection(Pd)offusioncenter
Graph between probabilities of detection of individual secondary user versus Fusion center
OR rule
AND rule
K-out-of-N rule

142
Figure 8. Analysis K-out-of-N rule for pd = 0.1, 0.5, and 0.9
V. CONCLUTION AND FUTURE RESEARCH WORK
In this paper, we have reviewed cooperative spectrum sensing techniques and analyzed decision fusion
techniques used in cognitive radio networks. The simulation results shows that probability of detection of the
fusion centre for K-out-of-N rule falls between the probabilities for OR and AND rules. The OR and AND
fusion rules at the fusion centre applies very relax and strict condition respectively during information fusion
at the fusion centre. While the K-out-of-N rule gives intermediate performance. Since the equations of the
false alarm probabilities for OR, AND and K-out-of-N rule are identical to spectrum detection probabilities,
hence the comparative analysis of false alarm probability and detection probabilities is also identical.
The simplicity and small overhead of hard combination technique needs further research to improve spectrum
detection performance. We are working to develop a softened hard k-bit algorithm which would be applied in
decision fusion in CSS. The other potential challenges like interference avoidance techniques for primary
users, sharing the sensed spectrum optimally among secondary users, identifying possible attacks on CRN (a
malicious secondary user can mislead CSS), developing mitigating solutions of the attacks, can be considered
as some of the open research areas in Cognitive Radio Networks.
REFERENCES
[1] Tevfik Yucek, Huseyin Arslan “A survey of spectrum sensing algorithms for cognitive radio Applications”, IEEE
communications survey & tutorials, vol. 11, no. 1, 2009, pp. 116-129.
[2] Erik Axell, greet leus, Erik G. Larsson, H. Vincent Poor, “Spectrum Sensing for Cognitive Radio”, IEEE signal
Processing Magazine, 2012, pp. 101-116.
[3] Saman Atapattu, Chintha Tellambura, Hai Jiang, “Energy Detection Based Cooperative Spectrum Sensing in
Cognitive Radio Networks”, IEEE Transaction on wireless communications, Vol. 10, No.4, 2011, pp.1232-1241.
[4] Jun Ma, Guodong Zhao, Ye (Geoffrey) Li, “Soft Combination and Detection for Cooperative Spectrum Sensing in
Cognitive Radio Networks” IEEE Transaction on wireless communications, Vol. 7, No. 11, 2008, pp.4502-4507.
[5] Edward C. Y. Peh, Ying-Chang Liang,Yong Liang Guan, Yonghong Zeng, “Cooperative Spectrum Sensing in
Cognitive Radio Networks with Weighted Decision Fusion Schemes”, IEEE Transaction on wireless
communications, Vol. 9, No.12, 2010, pp.3838-3847.
[6] Ghurumuruhan Ganesan, Ye (Geoffery) Li “Cooperative Spectrum Sensing in Cognitive Radio, Part I: Two user
networks” IEEE Transaction on wireless communications, Vol. 6, No. 6, 2007, pp. 2204-2212.
[7] Ghurumuruhan Ganesan, Ye (Geoffery) Li “Cooperative Spectrum Sensing in Cognitive Radio, Part II: Multiuser
networks” IEEE Transaction on wireless communications, Vol. 6, No. 6, 2007, pp.2214-2222.
[8] Kenta Umebayashi, Janne J. Lehtomäki, Takanao Yazawa, Yasuo Suzuki, “Efficient Decision Fusion for
Cooperative Spectrum Sensing Based on OR-rule”, IEEE Transaction on wireless communications, Vol. 11, No.7,
2012, pp.2585-2595.
0 1 2 3 4 5 6 7 8 9 10
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
values of K for K-out-of-N rule
Probabilityofdetection(Pd)offusioncenter
Graph between K and probability of detection of Fusion center for various pd =0.1,0.5,0.9
K-out-of-N rule for pd=0.1

143
[9] Krishnan Kalimuthu, Ramamoorthy Kumar, “Capacity maximization in spectrum sensing for Cognitive Radio
Networks thru Outage Probability”, International Journal of Electronics and Communications (AEÜ), 67 (2013),
pp.35-39.
[10] Wei Zhang, Khaled Ben Letaief, “Cooperative Spectrum Sensing with Transmit and Relay Diversity in Cognitive
Radio Networks”, IEEE Transaction on wireless communications, Vol. 7, No. 12, 2008, pp.4761-4766.
[11] Weijia Han, Jiandong Li, Zhi Tian, and Yan Zhang “Efficient Cooperative Spectrum Sensing with Minimum
Overhead in Cognitive Radio”, IEEE Transaction on wireless communications, Vol. 9, No.10, 2010, pp.3006-3011.
[12] Husheng Li, Huaiyu Dai, Chengzhi Li “Collaborative Quickest Spectrum Sensing via Random Broadcast in
Cognitive Radio Systems”, IEEE Transaction on wireless communications, Vol. 9, No.7, 2010, pp.2338-2348.
[13] Hang Hu, Hang Zhang, Hong Yu “Delay QoS guaranteed cooperative spectrum sensing in cognitive radio
networks”, International Journal of Electronics and Communications, Vol. 67, 2013, pp. 804-807.

74 136-143

More Related Content

What's hot

Similar to 74 136-143

More from idescitation

Recently uploaded

74 136-143