Aerospace and Electronic Systems Magazine May 2018 - 23

Ivanov et al.
tained graphical results for the desired values of Pfa and Pd. With
the use of an analytic expression, the sensing time is precalculated
on the basis of the known PU parameters. In a similar manner, in
[5], the bounds of the threshold are chosen empirically and the
probabilities of miss-detection and of false alarm are compared for
different thresholds within the range until an optimal value was
found. This study examined signals with various frequency ranges,
which was enabled by the use of the more advanced USRP B200
model. A combination of the two types of energy detectors was
presented in [6]. The edge ED integrated into the traditional ED
was used to locate the edges of the active subcarriers in the Power
Spectrum Density in order to discriminate the occupied from the
available portions of the spectrum. This method is less dependent
on whether the decision threshold of the standard detector is chosen
accurately or not. By using a signal generator as a PU, the received
power and SNR can be estimated precisely because the generator
defines clearly the transmitted power [10], [11], while this may not
be so simple when the PU is implemented via a USRP. In [10], the
received signal is obtained by calibration with the generator. The
noise uncertainty, though studied extensively in the works involving solely computer simulations, has not been considered in the
majority of the practical realizations. The authors in [11] included
the noise uncertainty into the standard expressions for Pd and Pfa
and added bounds to the threshold, which were dependent on the
desired Pfa. This study further uses a signal generator and examines
the cooperative spectrum sensing case with multiple USRP SUs to
increase the efficiency of the detector.
The detector implemented in [7] compares three methods,
which calculate the test statistics of the received signal using only
the autocorrelation function. Such an approach makes the process
less computationally intensive than a cyclostationary detector and
requires only the cyclic prefix length and the Fast Fourier Transform (FFT) size of the PUs orthogonal-frequency division multiplex (OFDM) signals to be known. In this form, however, the
detector is not applicable to other types of signals and the threshold
can be defined only empirically and for the specific case.
An example of the implementation of a cyclostationary spectrum sensing is provided in [8]. The method uses the symmetry
property of the CAF to obtain a decision. A scheme for compressed
sensing is also applied to increase the computational efficiency of
the calculations. The proposed detector is blind, in the sense, that
it does not require any prior information about the signal, neither
were the Pd and Pfa determined analytically, which makes the detector less flexible for introducing adaptability into it. In [9], the same
cyclostationarity method was proposed and compared to a modified
ED method, which compared the energy of sequential chunks of the
gathered samples until above the upper boundary of a threshold, or
below the lower boundary. This approach can improve the speed
of the ED, however the method did not consider a fading scenario.
A comprehensive survey on implementation of cyclostationary detectors using the FPGA platform is reported in [12]. The examined detectors show robustness in low SNR under a substantial
number of samples (over half a million). In spite of this the detection time is small because of the processing speed of the FPGA.
A more recent study [13] using the same type of platform shows
that cyclostationary detectors can be used for reliable signal recogMAY - JUNE 2018

nition under specific radio impairments such as carrier frequency
offset, phase noise and others even though some of them lead to
a noticeable decline in the detection performance. Similar results
are achieved in [14] but for a signal with much higher bandwidth.
A novel algorithm for cyclostationary detection was introduced in
[15]. It employs the concept of tunneling which provides the possibility to detect and classify different types of signals even though
they are under-sampled. That is why it can achieve faster and
more accurate processing than the typical cyclostationary method
if modulation classification is necessary. The authors in [16] proposed a cognitive radio OFDM system implemented on a FPGAbased SDR card equipped directly to a computer. This device supports simultaneous transmission and reception which allows the
SU to utilize the portions of the spectrum which are available.

COGNITIVE RADIO ARCHITECTURES
Other works, such as [18]-[21] have proposed complete architectures for cognitive communications and spectrum sharing where
the spectrum sensing is an essential functionality.
The study in [18] examined a framework for spectrum sharing
which enables a base station with both licensed networks and cognitive radio capabilities, to allocate resources to the two types of
users. A simulation of video transmission within a single cell with
primary and secondary users served by the same base station is
investigated. The emphasis is on developing a system for optimal
resource allocation between the users of the two radio access technologies, which minimizes the cost (in energy consumption) for
the CR terminals. A simple ED model is assumed for the spectrum
sensing function, but it is adaptive in the sense that the sensing duration is determined according to the priority of the video packets.
Therefore, this adaptation is not a part of the sensing function itself
but is performed together with the optimization of the rest of the
parameters of the framework.
Another application of spectrum sensing for dynamic spectrum
access was presented in [19]. The authors studied the opportunity
for a Wi-Fi network to utilize the spectrum used in general for
radar transmission. A framework was proposed that employs the
random idle periods, which generally would occur in a Wi-Fi system. The study is concentrated on finding the balance between the
amount of transmitted packets and the delay in detecting a radar
pulse (which can be detected only during the idle periods of the
Wi-Fi). It is assumed that detection will always be correct and so
the spectrum sensing function is not examined at all.
Cooperative spectrum sensing with emphasis on security was
studied in [20]. Threats to the accurate spectrum assessment, of different kinds, are considered and the performance of the proposed
solution for distributed cooperative decision-making is verified via
simulations. The shadowing effect is considered for determining the
efficiency of the overall cooperative detection scheme. The SU nodes
perform local spectrum sensing using the traditional energy detector.
Lagunas et al. [21] proposed a resource allocation framework,
which achieves spectrum sharing between Fixed Satellite Service
terminals (secondary users) and terrestrial microwave links (primary users). The study centered on carrier, power and bandwidth
allocation under interference constraints for the transmissions of

IEEE A&E SYSTEMS MAGAZINE

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