Aerospace and Electronic Systems Magazine May 2018 - 20

Feature Article:

DOI. No. 10.1109/MAES.2018.170098

Real-Time Adaptive Spectrum Sensing for
Cyclostationary and Energy Detectors
Antoni Ivanov, Technical University of Sofia, Bulgaria
Albena Mihovska, Aarhus University, Herning, Denmark
Krasimir Tonchev, Vladimir Poulkov, Technical University of Sofia, Bulgaria

INTRODUCTION
Multiple spectrum measurement campaigns around the world
have shown that even though the frequencies below 6 GHz are
very crowded, enough portions of the spectrum remain unutilized
most of the time. Even the most heavily employed bands, such as
the ones used for television broadcasting and cellular communications, have shown an average occupancy between 25 and 50% [1].
Together with the ever-increasing growth of connected devices in
the scope of the Internet of Things (IoT) concept, these have been
the main motivational factors for the intensive research in the field
of cognitive radio (CR), which empowered by the abilities of the
software-defined radio (SDR) devices, can enable the delivery of
IoT services, as they operate together with the incumbent users
of the existing wireless networks. Consequently, the CR can be
utilized for a variety of applications within IoT and the modern
systems for delivery of intelligent services, or the traditional standards for wireless data transmission including mobile networks
[2]. Therefore, it can be implemented in diverse appliances and
professional equipment for integrated navigation, sensing, and
communications.
One of the enabling functions of the CR device (also called
a secondary user) is the spectrum sensing, which allows for the
detection of unused frequency resources. The CR device must free
the spectrum once the incumbent or primary user (PU) starts to use
it to avoid causing interference; therefore, the time and accuracy
for detecting an empty spectrum slot are a very important condition
Authors' current addresses: A. Ivanov is with Technical
University of Sofia, Sofia, 1000, Bulgaria, Email: astivanov@
tu-sofia.bg; A. Mihovska is with BTECH, Aarhus BSS, Aarhus
University, Herning, 7400, Denmark; K. Tonchev is with Technical University of Sofia, Sofia, 1000, Bulgaria; V. Poulkov is
with Technical University of Sofia, Sofia, 1000, Bulgaria, Email:
vkp@tu-sofia.bg.
This paper was supported by Contract DN 07/22/2016 of
the Bulgarian Research Fund of the Ministry of Education.
Research project: "Self-coordinating and Adaptive Wireless
Cyber-Physical Systems with Human in the Loop".
Manuscript received May 30, 2017, revised November 1, 2017,
and ready for publication December 20, 2017.
Review handled by L. Ligthart.
0885/8985/18/$26.00 © 2018 IEEE
20

of any spectrum sensing algorithm. There are two main parameters, which characterize the spectrum sensing process-the probability of detection (Pd) and probability of false alarm (Pfa). The
former defines the probability that the SU will detect the presence
of the PU correctly. Alternatively, the probability of miss-detection
(Pmd) represents the likelihood that the SU will not discover the PU.
The probability of false alarm describes the possibility that the SU
will decide that the PU is present, when in reality the band has not
been occupied. An incorrect decision on the spectrum occupancy
will result in either interference to the PU (if its presence is not
detected) or a loss of opportunity for using unutilized spectrum (in
case of a false alarm). That is why the efficiency of the detector
is central to most of the research efforts. The traditional spectrum
sensing techniques' energy detection, matched filter, cyclostationary detection, and wavelet detection [3] form the basis for most of
the known solutions.
In the past decade, the amount of research in relation to spectrum sensing has increased substantially. However, the majority
of it is performed theoretically and verified only via computer
simulations without testing their performance in practical real-life
scenarios. Moreover, no adaptivity or trade-off have been applied
in the expressions defining the detection performance in most of
the works. There have also been many practical realizations of CR
systems ([2], [4]-[17]) with a different range of scientific contributions but they are much fewer in comparison to the works which
present solutions verified only via simulations. This highlights the
need for more practical implementations. Recently proposed CR
applications ([18]-[21]) consider only simple energy detection as
the local spectrum sensing method which is the basis for their overall architectures, and in some cases, the role of spectrum sensing
has not been considered at all. In addition, there is a substantial
evidence in the literature ([22]-[24]) that suggests that the time,
which is used for the spectrum sensing process should be adapted
to the patterns of the PU transmission, for achieving an optimal
trade-off between the sensing accuracy (proportional to the sensing
period) and the loss of opportunity for the utilization of the unused
spectrum. These facts indicate the need for examinations of robust
detectors in real-life environments.
In this article, we explore the details of some practical implementations of energy and cyclostationary detectors, which take
into account the specific radio channel impairments (like noise uncertainty and fading), using the Universal Radio Serial Peripheral

IEEE A&E SYSTEMS MAGAZINE

MAY - JUNE 2018

Table of Contents for the Digital Edition of Aerospace and Electronic Systems Magazine May 2018