Anda di halaman 1dari 14

Wireless Pers Commun (2012) 64:93106

DOI 10.1007/s11277-012-0519-4

Adaptive Counting Rule for Cooperative Spectrum


Sensing Under Correlated Environments
Nuno Pratas Nicola Marchetti
Neeli Rashmi Prasad Antnio Rodrigues
Ramjee Prasad

Published online: 7 February 2012


Springer Science+Business Media, LLC. 2012

Abstract Spectrum sensing is the cognitive radio mechanism that enables spectrum awareness. It has been shown in the literature that spectrum sensing performance can be greatly
improved through the use of cooperative sensing schemes. This paper considers and proposes a data fusion based cooperative spectrum sensing scheme based on data fusion, where
an adaptive counting rule is used to implement the data fusion. The proposed scheme is
evaluated against other common counting rules (e.g., 1-out-of-c and c-out-of-c) found in the
literature and the optimum counting rule, while under different correlation conditions. The
impact of correlation on the performance of the considered counting rules is then studied.
It is concluded that the proposed adaptive counting rule detection performance reaches in
some cases the one of the optimum counting rule, and therefore it adapts to the correlation
conditions which the network nodes are experiencing.

The authors would like to acknowledge the FCT which supported this work under the grant
SFRH/BD/36454/2007 and IT/LA.
N. Pratas (B) N. Marchetti N. R. Prasad R. Prasad
Center for TeleInFrastruktur (CTIF), Aalborg University (AAU),
Niels Jernes Vej 12, 9220 Aalborg st, Denmark
e-mail: nup@es.aau.dk
N. Marchetti
e-mail: MARCHETN@tcd.ie
N. R. Prasad
e-mail: np@es.aau.dk
R. Prasad
e-mail: prasad@es.aau.dk
N. Pratas A. Rodrigues
Instituto Superior Tcnico (IST/UTL), Instituto de Telecomunicaes (IT),
Av. Rovisco Pais, 1, 1049-001 Lisbon, Portugal
A. Rodrigues
e-mail: ar@lx.it.pt

123

94

Keywords

N. Pratas et al.

Cooperative spectrum sensing Data fusion Counting rule

1 Introduction
Spectrum sensing is the cornerstone of the cognitive radio paradigm [10] and has therefore
been the focus of intensive research. One of the main conclusions from the research community is that spectrum sensing performance can be greatly enhanced through the use of
cooperative sensing schemes.
Spectrum sensing, covered extensively in the cognitive radio literature [1], falls into the
realm of detection theory, discussed in detail in [17]. Spectrum sensing is realized as a physical and medium access control (MAC) layer mechanism [7]. The MAC layer part of spectrum
sensing focuses on when to sense and which range of spectrum to sense. The physical layer
part of spectrum sensing focuses on detecting signals and the detection methods used can be
classified into two groups:
Coherentwhere a signal can be coherently detected by comparing the received signal
or the extracted characteristics of the signal with a priori knowledge about the type of
that same signal, e.g., Pilot-Based Coherent Detection, [8];
Non-coherentno prior information needed, e.g., Energy Detector, [16].
The modelling of the performance of these detectors is done by applying the binary hypothesis testing problem. In the literature two approaches to model the binary hypothesis testing
problem are commonly followed [1,17], being those:
The Bayesian approach (here the conditional densities under the two hypotheses have to
be known a priori as well as the cost of the action that can be followed).
The NeymanPearson (NP) test where a priori information is not needed.
This paper focuses on the effect of correlation on the detection performance of, counting
rule based data fusion schemes, i.e., Hard Combining schemes [1]. It is assumed there is no
access to any prior knowledge about the signals that are being detected, therefore the NP
test formulation is used.
This study is performed on an ad-hoc network, where the nodes are organized in a cluster
and one of the nodes is the cluster head, which besides being responsible by the management
of the cluster nodes is also the node on which the data fusion process occurs. This paper is
an extended version of the paper [12].
The remainder of this paper is organized as follows. In Sect. 2, the State of the Art is
shown. Section 3 gives the problem definition, methodology and the evaluation metrics used.
In Sect. 4 the system design is discussed. In Sect. 5, the data fusion scheme is explained
under certain correlation conditions. Finally, Sect. 6 concludes the paper.

2 State of the Art


Spectrum sensing requirements are set by the channel conditions that depend on path loss,
multipath, shadowing and local interference. The combination of these phenomena can result
in regimes where the signal SNR is below the detection threshold of the detector. To overcome this limitation, the use of cooperation in spectrum sensing was proposed in [1,2,5,9,19].
Since signal strength varies with sensor location, the worst fading conditions can be avoided
if multiple detectors share their sensing measurements, i.e., taking advantage of the spatial

123

Adaptive Counting Rule

95

diversity between the detectors. Most of the proposed cooperative methods are based on data
fusion techniques [1]. These allow the combination of the sensing measurements of multiple
detectors to achieve a global decision on what is the actual state of the sensed spectrum.
A simulation based study on the effect of correlation on cooperative spectrum sensing was
presented in [13], where the main conclusion was that cooperation is not always worthwhile.
When data fusion schemes are used, they cannot perform the data fusion on the sensing results
obtained from all the sensing nodes, and therefore need to be subdivided in smaller subsets
nodes. This effect as also been studied in [6], where the asymptotic performance (in the limit
of large number of users) of collaborative spectrum sensing under exponentially-correlated
log-normal shadowing was modeled. It was established that the probability of missed-opportunity is lower-bounded and the effective number of users depends only on their distance
spread and the characteristics of the propagation environment.
In [3], the authors proposed an adaptive algorithm to solve the optimal fusion problem
when sensors are dependent from one another, i.e., when they are correlated. The proposed
algorithm does not require a priori knowledge about the sensors and source, adapting the
weights from time to time. In the algorithm, the log-likelihood ratio function is expressed
as a linear combination of ratios of conditional probabilities and local decisions, where the
estimations of the conditional probabilities are adapted by reinforcement learning.
In [4] the authors derived the optimal likelihood ratio test (LRT) in the presence of correlation for a scenario where all the detectors have the same detection performance. It was later
observed and proved, in [18], that the derived LRT is equivalent under certain correlation
conditions to a counting rule. This is the foundation of the work presented in this paper and
therefore is further ellaborated in Sect. 5.

3 Preliminaries
3.1 Problem Definition and Methodology
The problem tackled in this paper is the following:
In a spectrum sensing context, what is the effect of correlation on the detection performance of a counting rule based data fusion scheme?
This problem is studied on an ad-hoc cognitive radio network where a MAC layer distributed spectrum sensing mechanism is introduced. This mechanism allows for a centralized as
well as a decentralized approach, and has been further elaborated in [14,15]. The proposed
distributed mechanism allows each of the sensing nodes to sense any channel of the targeted
spectrum while ensuring that an accurate detection of the spectrum state is achieved, i.e., it
checks whether the channel is vacant or not.
3.2 Evaluation Metrics
The goal of spectrum sensing is to detect what is the state of the range of spectrum being
monitored. This process can be described as a mechanism which corresponds to an imperfect
and simplified mapping of the real radio environment conditions, to a representation in the
sensing node.
The root mean square error (RMSE) is frequently used to measure the differences between
values predicted by a model or an estimator and the values actually observed from the phenomenon being modeled or estimated. Therefore RMSE is a measure of accuracy, [11]. Here

123

96

N. Pratas et al.

we use it to quantify the error between the detected state and the real state of the sensed
channel.
So consider that the sm,i and sm,i represent the detected state and real state of channel m in
sensing session i, then the individual accounted differences, sm,i sm,i , are called residuals,
and the RMSE aggregates these individual residuals to quantify the error between sensed
and real state of the channel during a set of sucessive sensing sessions.
The RMSE of the channel m state is given by,

2
N 
i=1 sm,i sm,i
RMSE =
(1)
N
where N is the number of sensing sessions.

4 System Design Overview


As stated in the previous sections, the considered network architecture is an ad-hoc network,
where the network nodes are organized in a cluster. Each of these network nodes are assumed
to have available two logical types of channel, as shown in Fig. 1, being those:
The control channel (CCH) where all control information is exchanged and where the
access method is of the carrier sense multiple access with collision avoidance (CSMA/CA)
type.
The data channel (DCH) through which the users data are exchanged and of which the
access method is out of scope.
It is assumed that each sensing node performs the sensing through the use of an energy
detector (ED) [16] a non-coherent sensing scheme and that they are already designed to
achieve a certain detection performance. Specifically that they are designed to detect signals
above a predefined SNR threshold. It is also assumed that these detectors have a homogeneous
performance, i.e., all the detectors were designed to detect the same SNR threshold.
The distributed spectrum sensing mechanism, depicted in Fig. 2, ensures that all the
network nodes have updated and synchronized information about the state of the targeted
spectrum. The considered mechanism was first proposed on [14,15]. The steps of the depicted
flow are the following:
Control Frame

(a)

Sensing
Exchanges

Control Exchanges
Data Frame

(b)

Sensing

Data

Sensing
Reporting

Processing

Orchestration
Reporting

Sensing Reporting

Fig. 1 a Control frame. b Data frame

123

(a.1)

(a.2)

Adaptive Counting Rule

Spectrum
Sensing

97

Local
Decision
Reporting

Local
Decisions
Fusion

Knowledge
Base
Update

Node
Selection

Fig. 2 Distributed spectrum sensing mechanism flow

Spectrum SensingEach node uses a detector to perform the spectrum sensing and therefore reaches a local decision regarding the status of the sensed channel, i.e., if either the
channel is vacant or not;
Local Decision ReportingEach node shares the local decisions with the rest of the network, the intended recipients depends on the considered network topology, which can be
centralized, decentralized or relay based. In this paper we consider a centralized topology;
Local Decisions FusionHere the recipients nodes perform the data fusion with the recieved the shared local decisions. In a multi-channel sensing scenario the data fusion is
performed separately for each channel;
Knowledge Base UpdateHere the network combines the reached global decision in the
previous step with the past decisions so to create a continuous knowledge base to be used
by the network for deciding which resources to use;
Node SelectionFinally in this step the network chooses which channel should each of
the network nodes sense.
This paper focuses on the Local Decisions Fusion step. In this step the cluster head node
combines all the received local decisions from the other network nodes with its own local
decision. Note that this fusion process is done separately for each sensed channel.
A synchronous hard decision type data fusion scheme is considered. Where the data
fusion is performed through the use of counting rules, which take into account the radio
channel properties and the correlation experienced between the nodes, by adapting to
it. The mechanism which allows this adaptation is presented and discussed in the next
session.
Note that the use of soft combining schemes, [1], instead of hard decisions would of course
lead to higher detection performance, but at the expense of the reporting channel bandwidth
and therefore are not considered in this paper.

5 Spectrum Sensing Results Fusion


5.1 Introduction to Data Fusion
A two-level distributed detection system has been considered consisting of a number of local
detectors and a fusion center, depicted in Fig. 3. The detectors reach a local decision based on
the underlying binary hypothesis testing problem. The local decisions are then transmitted
to the fusion center where a global decision is made.
The local decisions are combined in the fusion center, so to achieve a global decision. The
design described next is of a binary hypothesis based fusion rule.

123

98

N. Pratas et al.

Fig. 3 Two-level distributed


detection system

y1

Detector 1

u1

y2

Detector 2

u2
Fusion
Center

yn

Detector n

uf

un

5.2 Data Fusion Methodology


Consider the binary hypothesis testing problem with hypotheses H0 and H1 being the cases
where the sensed channel is vacant and occupied, respectively. The fusion center implements
the NP test by using all the local decisions, which the detectors have communicated to it.
The NP test, (u), is formulated as the following likelihood ratio (LR) test:
(u) =

P(u 1 , u 2 , . . . , u c |H1 ) H1

P(u 1 , u 2 , . . . , u c |H0 ) H0

(2)

where u = [u 1 , . . . , u c ]T , is the vector formed by the set of local decisions (where u i = 0


when the sensor i decides H0 , and u i = 1 when the sensor i decides H1 ) corresponding to
the c local detectors and is the decision threshold. is computed by setting an upper bound
on the probability of false alarm at the fusion center [17].
Consider the case where all the detectors have the same performance and there is a quantifiable degree of correlation between their decisions. Then, according to [4], the (LRT) can
be expressed as:
(u) =

P(u|H1 )
= (m)
P(u|H0 )

(3)

where m out of c detectors are in favor of H0 (i.e., there are m zeros in the vector u).
Elaborating further (u) [18]:
 
i m
i=0 (1) i Pd

m

cm+i2

1 (k+1Pd )+Pd
1+k1
 
cm+i2 0 (k+1P f a )+P f a
i m
i=0 (1) i P f a
k=0
1+k0

(u) = 
m

k=0

, 0m n2

(4)

where Pd and P f a , are respectively the probability of detection and probability of false alarm,
and the correlation indices 0 0 1 and 0 1 1 are defined by:
k = 

E[u i u j |Hk ] E[u i |Hk ]E[u j |Hk ]


E[(u i E[u i ])2 |Hk ]E[(u j E[u j ])2 |Hk ]

, i, j i  = j, k = 0, 1

(5)

Note that although (u) was not explicitly considered in [4] for the cases when m = c 1
and m = c, it was considered and derived in [18].
The optimum decision rule, i.e., the one which maximizes the global probability of detection, for a given upper bound of the global probability of false alarm, is obtained by the LRT
defined by:

123

Adaptive Counting Rule

99

r uleoptimal (u) = r uleoptimal (m) =

(6)

if (m) >
if (m) =
if (m) <

(7)

H1
,
H1 with probability ,

H0
,

where is a randomization constant. Both and are greater than zero, and are defined by
the upper bound of the global probability of false alarm.
The implementation of the LRT leads to complex iterative algorithms, and in [4] it was
shown that the LRT can be expressed as a function of m, i.e., the number of detectors that
decide in favor of H0 .
In [18] it was then proposed to use a counting rule, i.e., a rule that counts m, and decide
H1 when m is smaller than a given integer threshold. This counting rule is defined as:

r ulecount (u) = r ulecount (m) =

(8)

if m < m 0
if m = m 0
if m > m 0

(9)

H1
,
H1 with probability ,

H0
,

The equivalence between the LRT and the counting rule given by:
(m)  m  m 0 ()

(10)

is only valid if (m) is a decreasing function of m, as thoroughly demonstrated in [18].


The choice of the counting rule threshold, m 0 , depends on the detectors performance,
the correlation between the decisions of the detectors, and finally on the upper bound set for
the global probability of false alarm. Considering that the correlation between the decisions
changes over time, when the detectors are mobile, there should be a mechanism which adapts
the m 0 threshold.
5.3 Data Fusion Through an Adaptive Counting Rule
The need for an adaptive counting rule is motivated by the examples given in Figs. 4 and 5,
where one shows the RMSE of the global decision at the fusion center. The k is the decision
threshold defined as k = n m 0 . The k that minimizes the RMSE depends on the average
correlation index between the local detectors. Therefore there must be a mechanism which
allows selecting the best k according to the correlation index. Note also, that the optimum
k depends on the detectors performance as well as the number of detectors put in place to
detect the channel state. Finally, note that the results depicted in Figs. 4 and 5 were obtained
in presence of a signal with a 50% duty-cycle. Note that the obtained RMSE will vary with the
considered duty cycle, i.e., when the duty cycle is high the error introduced in the detection is

60%
50%

RMSE

Fig. 4 RMSE versus k according


to correlation index, with 10 local
detectors

40%
30%
20%
10%
0%

10

123

100

N. Pratas et al.

60%

RMSE

50%
40%
30%
20%
10%
0%
1

k
Fig. 5 RMSE versus k according to correlation index, with 5 local detectors
GPd

Probability

100%

GPfa

80%
60%
40%
20%
0%
1

10

k
Fig. 6 Variation of the G Pd and G P f a with the k threshold and c = 10

mostly due to miss detections, while when the duty cycle is low the error introduced is mostly
due to the false alarms (false positives). To better understand how the choice of k affects the
performance of the data fusion scheme, consider that there c network nodes and that their
decisions are uncorrelated. The global probability of detection, G Pd , and false alarm, G P f a ,
for a counting rule in these conditions is given by,
GPd =

c

c
i=k

Pd i (1 Pd )ci

(11)

P f a i (1 P f a )ci

(12)

and
GP f a =

c

c
i=k

where k is the threshold of the counting rule, i.e., k-out-of-c. The behaviour while varying
the k of the G Pd and G P f a is depicted in Fig. 6, where the Pd = 0.6 and P f a = 0.05.
From these it can be observed that an higher k minimizes both probabilities while a lower
k maximizes it and that in these conditions there is a k which minimizes the RMSE, which
is most likely different from the one where the correlation conditions between the network
nodes are different.
To find the best k according with network correlation conditions an algorithm is proposed.
Since it is not possible for the data fusion center to know what is the correlation between the
network nodes before they perform the local detection, then the correlation information is not
available. So an alternate route was taken which was to adapt the k threshold continuously,
according to the feedback from the previous global decisions. This approach allows to adapt
to the correlation conditions without having implicit information about it by instead using the
information from the feedback. For example in the case the sensed channel is deemed vacant

123

Adaptive Counting Rule

101

then at least one of the network nodes will try to transmit on the channel and the sucess of
that action is then used as feedback to tune the k.
If the nodes try to transmit and the channel is not vacant then it means that there was a
misdetection and therefore the k should be decreased since the G Pd needs to increase. At
the same time there should be an inverse mechanism which increases k, so to ensure that the
upper bound of the G P f a is respected.
The algorithm which implements the adaptable counting rule is depicted in Algorithm 1.
This algorithm adapts the decision threshold k, so that for a detection to occur there have to
be at least k positive detections out of the n detectors. Where the algorithm metrics are the
following:

Algorithm 1 Adaptive counting rule mechanism


if U F = H1 then
O DC + +
else
if C S  = H0 then
M DC + +
end if
end if
if M DC > M DT hr s then
k
end if
if O DC > OC T hr s then
k++
end if
Update M DC and O DC and discard observations outside the O W

Misdetection counter (M DC)Counts every time the fusion center decides that the
channel is free, and the network tries to use the channel but the channel is occupied.
Occupied detection counter (O DC)Counts every time the fusion center decides the
channel is occupied.
Observation window (O W )Is the time interval of the observations used to make the
decisions, i.e., it serves as the memory of what occurred in the previous sensing sessions.
Misdetection threshold (M DT hr s)Is the amount of misdetections permitted within
the observation window. When this value is overcome, it triggers an event to decrease k.
This threshold can be obtained by multiplying the 1 G Pd target by with O W , where the
G Pd target is the G Pd targetted for the network.
Occupied detection threshold (OC T hr s)Is the amount of occupied detections occurring during the OW, which, if exceeded, triggers the increase of k. It is obtained by
multiplying the OW with the upper bound of the allowed global false alarm probability.
Channel state (C S)Is the actual channel state, which is found when accessing the
channel.
The system feedback, used to tune the k parameter, can also be obtained, e.g., by the statistics of the successful transmissions, failed transmissions, and on the case of a CSMA/CA
based system, on the times the CA mechanism was put in place, etc.

123

102

N. Pratas et al.

60%

RMSE

50%
40%
30%
20%
10%
0%

Fig. 7 RMSE according to fusion rule

5.4 Adaptive Counting Rule Performance Comparison


To evaluate the performance of the adaptive counting rule mechanism, the two extreme cases,
the OR rule (1 out o f c) and the AND rule (c out o f c) are considered. The
results are shown in Fig. 7.
The RMSE obtained while using the different rules is depicted in Fig. 7. The adaptive
rule RMSE is close to the one obtained with the optimum rule, while outperforming both
the OR and AND rule, except for the case where correlation is 0.3 with c = 5. In this case
the performance of the adaptive rule is worse than the optimum rule because it is unable
to maintain the k equal to the value of the lower bound (k = 1) or upper bound (k = c).
This is due to algorithm being unstable at the extremities of the k domain, i.e., the k is only
able to maitain a value between ]1, c[, otherwise it switches between 1 and 2 in the case
of the lower bound or c 1 and c in the case of the upper bound. Note that the cause and
solution of this unstability was not addressed in this paper. The proposed adaptive fusion
rule algorithm achieves almost the same performance as the optimum rule, without having
implicit knowledge about the correlation between the local detectors.

6 Conclusion
In this paper, the results from a study on the effect of correlation between detectors on the
performance of a cooperative spectrum sensing data fusion based scheme were discussed.
As an outcome of this study, an adaptive counting rule has been proposed, which adapts its
parametersthe k threshold in k-out-of-c counting ruleto achieve the best performance
according to the experienced correlation. The adaptive rule accomplishes this by adapting the
decision threshold dynamically according to the correlation experienced by the underlying
sensors, while not having implicit information about the correlation, i.e., acting therefore as
an adaptive feedback mechanism.
The proposed adaptive counting rule performance has been compared with other common
counting rules as well the optimum counting rule, according to the underlying correlation
between the sensing nodes. It was shown that the proposed counting rule in some cases
reaches the performance of the optimum counting rule.
The future steps of this work will be to further study and refine the proposed data fusion
algorithm, specifically in the case where the correlation index between the detectors varies

123

Adaptive Counting Rule

103

over the time and also address the the instabilities of the algorithm in the upper and lower
bound of the k domain.

References
1. Akyildiz, I. F., Lo, B. F., & Balakrishnan, R. (2011). Cooperative spectrum sensing in cognitive radio
networks: A survey. Physical Communication, 4(1), 4062. doi:10.1016/j.phycom.2010.12.003.
2. Cabric, D., Tkachenko, A., & Brodersen, R.W. (2006). Experimental study of spectrum sensing based
on energy detection and network cooperation. In Proceedings of the First International Workshop on
Technology and Policy for Accessing Spectrum, TAPAS 06. ACM, New York, NY, USA. doi:10.1145/
1234388.1234400.
3. Chen, J. G., & Ansari, N. (1998). Adaptive fusion of correlated local decisions. IEEE Transactions
on Systems, Man, and Cybernetics, 28(2), 276281. doi:10.1109/5326.669570.
4. Drakopoulos, E., & Lee, C. C. (1991). Optimum multisensor fusion of correlated local decisions. IEEE
Transactions on Aerospace Electronic Systems, 27, 593606. doi:10.1109/7.85032.
5. Ghasemi, A., & Sousa, E. (2005). Collaborative spectrum sensing for opportunistic access in fading
environments. In 2005 First IEEE International Symposium on New Frontiers in Dynamic Spectrum
Access Networks, 2005. DySPAN 2005 (pp. 131136). doi:10.1109/DYSPAN.2005.1542627.
6. Ghasemi, A., & Sousa, E. (2007). Asymptotic performance of collaborative spectrum sensing under
correlated log-normal shadowing. IEEE Communications Letters, 11(1), 3436. doi:10.1109/LCOMM.
2007.060662.
7. Kim, H., & Shin, K. (2008). Efficient discovery of spectrum opportunities with mac-layer sensing
in cognitive radio networks. IEEE Transactions on Mobile Computing, 7(5), 533545. doi:10.1109/
TMC.2007.70751.
8. Ma, J., Li, G., & Juang, B. H. (2009). Signal processing in cognitive radio. Proceedings of the
IEEE, 97(5), 805823. doi:10.1109/JPROC.2009.2015707.
9. Mishra, S., Sahai, A., & Brodersen, R. (2006). Cooperative sensing among cognitive radios. In IEEE
International Conference on Communications, 2006. ICC 06 (Vol. 4, pp. 16581663). doi:10.1109/
ICC.2006.254957.
10. Mitola, J., & Maguire, G. (1999). Cognitive radio: Making software radios more personal. IEEE
Personal Communications, 6(4), 1318. doi:10.1109/98.788210.
11. Mood, A. M. (1974). Introduction to the theory of statistics, (3rd ed.). McGraw-Hill series in probability
and statistics. New York, NY: McGraw-Hill.
12. Pratas, N., Marchetti, N., Prasad, N., Prasad, R., & Rodrigues, A. (2010). Adaptive counting rule
for cooperative spectrum sensIng under correlated environments. In Proceedings of Wireless Personal
Multimedia Communications Symposia 2010, (pp. S111).
13. Pratas, N., Marchetti, N., Prasad, N., Prasad, R., & Rodrigues, A. (2010). System capacity limits
introduced by data fusion on cooperative spectrum sensing under correlated environments. MTA
Review, XX(4), 245262.
14. Pratas, N., Marchetti, N., Prasad, N., Rodrigues, A., & Prasad, R. (2010). Centralized cooperative
spectrum sensing for ad-hoc disaster relief network clusters. In IEEE International Conference on
Communications (ICC), 2010 (pp. 15). doi:10.1109/ICC.2010.5502710.
15. Pratas, N., Marchetti, N., Prasad, N. R., Rodrigues, A., & Prasad, R. (2010). Decentralized cooperative
spectrum sensing for ad-hoc disaster relief network clusters. In 2010 IEEE 71st on Vehicular Technology
Conference (VTC 2010-Spring) (pp. 15). doi:10.1109/VETECS.2010.5494012.
16. Urkowitz, H. (1967). Energy detection of unknown deterministic signals. Proceedings of the
IEEE, 55(4), 523531. doi:10.1109/PROC.1967.5573.
17. Varshney, P. K. (1996). Distributed detection and data fusion (1st ed.). New York, NY, Secaucus,
NJ: Springer.
18. Vergara, L. (2007). On the equivalence between likelihood ratio tests and counting rules in distributed
detection with correlated sensors. Signal Processing, 87(7), 18081815. doi:10.1016/j.sigpro.2007.01.
023.
19. Visotsky, E., Kuffner, S., & Peterson, R. (2005). On collaborative detection of tv transmissions in
support of dynamic spectrum sharing. In 2005 First IEEE International Symposium on New Frontiers
in Dynamic Spectrum Access Networks, 2005. DySPAN 2005 (pp. 338345). doi:10.1109/DYSPAN.
2005.1542650.

123

104

N. Pratas et al.

Author Biographies
Nuno Pratas received his B.Sc. and M.Sc. in Electrical Engineering
from Insttuto Superior Tcnico of the Technical University of Lisbon
in Portugal. He is currently pursuing his Ph.D. degree in Wireless Communications at Aalborg University in Denmark. His research focuses
on cooperative spectrum sensing methods for cognitive radio and in the
development of analysis tools for communication systems based on discrete time driven simulators. He has authored and co-authored several
conference and journal papers, as well as a book chapter. He is IEEE
student member since 2005.

Nicola Marchetti received the Ph.D. in Wireless Communications and


the M.Sc. in Mathematics from Aalborg University in the years 2007
and 2010 respectively, and the M.Sc. in Electronic Engineering from
University of Ferrara, Italy in 2003. He worked as a Research Assistant at the University of Ferrara during July 2003 to April 2004. He
also worked at Aalborg University as a Ph.D. fellow during May 2004
to May 2007 and as a Post-doctoral Research Fellow and Assistant
Professor from June 2007 till December 2010. Dr. Nicola Marchetti
is the Co-Founder and Vice-President of the International Conference
on Cognitive Radio and Advanced Spectrum Management (CogART)
and the International Symposium on Applied Sciences in Biomedical
and Communication Technologies (ISABEL). His research interests
include: Multiple Antenna Technologies, Single & Multi Carrier Modulations, Advanced Radio Resource Management Techniques, Cognitive Radios and Networks, Self-organizing Networks, e-Healthcare,
and Mathematics applied to Wireless Communication.
Neeli Rashmi Prasad Ph.D., IEEE Senior Member, Head of research
at the Center for TeleInfrastruktur (CTIF) at Aalborg University and
Director of CTIF-USA, Princeton, USA. She has over 14 years of management and research experience both in industry and academia. She
has gained a large and strong experience into the project coordination of EU-funded and Industrial research projects. She joined Libertel (now Vodafone NL), The Netherlands in 1997 as radio engineer
and team leader. Till May 2001, she worked as systems architect and
corporate member of architecture team at Wireless LANs in Wireless Communications and Networking Division of Lucent Technologies, The Netherlands. From June 2001 to July 2003, she was with
T-Mobile Netherlands, The Netherlands as Senior Core Network architect. Subsequently, from July 2003 to April 2004, at PCOM:I3, Aalborg, Denmark. She has been involved in a number of EU-funded R&D
projects, including FP7 CIP-PSP LIFE 2.0, FP7 CIP-PSP ISISEMD,
FP7 IP ASPIRE, FP7 IP FUTON, FP6 IP eSENSE, FP6 NoE CRUISE,
FP6 IP MAGNET and FP6 IP Magnet Beyond as the latest ones. She is currently the project coordinator of
the FP7 IST IP ASPIRE, FP7 CIP-PSP LIFE 2.0 and was project coordinator of FP6 NoE CRUISE. She was
also the leader of EC Cluster for Mesh and Sensor Networks and is Counsellor of IEEE Student Branch,
Aalborg. Her current research interests are in the area of QoL, SON, IoT, Identity Management, mobility,
network management and monitoring; practical radio resource management; cognitive learning capabilities
and modelling; Security, Privacy and Trust. Experience in other fields includes physical layer techniques,
policy based management, short range communications. Her publications range from top journals, international conferences and chapters in books. She has also co-edited and co-authored two books and has over 50

123

Adaptive Counting Rule

105

peer reviewed papers in international journals and conferences. She is also very active in several conferences
as chair and as program committee member.
Antnio Rodrigues received the B.S. and M.S. degrees in electrical
and computer engineering from the Instituto Superior Tcnico (IST),
Technical University of Lisbon, Lisbon, Portugal, in 1985 and 1989,
respectively, and the Ph.D. degree from the Catholic University of Louvain, Louvain-la-Neuve, Belgium, in 1997. Since 1985, he has been
with the Department of Electrical and Computer Engineering, IST,
where is currently an Assistant Professor. His research interests include
mobile and satellite communications, wireless networks, spread spectrum systems, modulation and coding.

Ramjee Prasad has been holding the Professorial Chair of Wireless Information and Multimedia Communications at Aalborg University, Denmark (AAU) since June 1999. Since 2004 he is the Founding Director of the Center for TeleInfrastruktur (CTIF), established as
large multi-area research center at the premises of Aalborg University.
Ramjee Prasad is a Fellow of IEEE, the IET and IETE is a worldwide established scientist, which has given fundamental contributions
towards development of wireless communications. He achieved fundamental results towards the development of CDMA and OFDM, taking the leading role by being the first in the world to publish books in
the subjects of CDMA (1996) and OFDM (1999). He is the recipient
of many international academic, industrial and governmental awards
and distinctions of which the most recently is the cross of the order of
chivalry (Ridderkorset af Dannebrogordenen) from the Danish Queen
due internationalization of top-class telecommunication research and
education. He has published a huge number of books (more than 25),
journals and conferences publications (together more than 750), more than 15 patents, a sizeable amount of
graduated Ph.D. students (over 60) and an even larger amount of graduated M.Sc. students (over 200). Several of his students are today worldwide telecommunication leaders themselves. He is the founding chairman
of the Global ICT Standardization Forum for India (GISFI) and was the founding chairman of the European
Center of Excellence in Telecommunications known as HERMES of which he is now the honorary chairman.
Recently, under his initiative, international M.Sc. and Ph.D. programmes have been started with the Sinhgad
Technical Education Society in India, the Bandung Institute of Technology in Indonesia and with the Athens
Information Technology (AIT) in Greece. Ramjee Prasad has a long path of achievements until to date and
a rich experience in the academic, managerial, research, and business spheres of the mobile and wireless
communication area. Namely, he played an important role in the success that the Future Radio Wideband
Multiple Access Systems (FRAMES) achieved. He was the leader of successful EU projects like the MAGNET and MAGNET Beyond, among others, as well as the driver of fruitful cooperation with companies
in projects, like Samsung, Huawei, Nokia, Telenor, among others. He is advisor to several multinational
companies. He started as a Senior Research Fellow (19701972) and continued as an Assistant Professor
(19721980) at the Birla Institute of Technology (BIT), Mesra, Ranchi, India. He was appointed as an Associate Professor in 19801983 and head of the Microwave Laboratory there. From 19831988 Ramjee Prasad
worked at the University of Dar es Salaam (UDSM), Tanzania, where he became Full Professor of Telecommunications in the Department of Electrical Engineering in 1986. From February 1988 till May 1999 Ramjee
Prasad worked at the Delft University of Technology (DUT), The Netherlands at the Telecommunications
and Traffic Control Systems Group. He was the founding head and program director of the Centre for Wireless and Personal Communications (CWPC) of the International Research Centre for TelecommunicationsTransmission and Radar (IRCTR) at DUT, The Netherlands. Prior to founding CTIF, Ramjee Prasad was the
Co-Director of the Center for Person Kommunikation until December 2002. He became the research director of the department of Communication and Technology in 2003. Prof. Prasad has authored or co-authored

123

106

N. Pratas et al.

more than 700 high cited scientific articles published in peer-reviewed conference proceedings and international journals. Since 1999, he has published 8 monographs, 22 books, 18 book chapters and more than 70
and 190 articles in journals and conference proceedings. Furthermore he has 15 patents within his research
areas.

123

Anda mungkin juga menyukai