% IMPORTANT: The following is UTF-8 encoded. This means that in the presence
% of non-ASCII characters, it will not work with BibTeX 0.99 or older.
% Instead, you should use an up-to-date BibTeX implementation like “bibtex8” or
% “biber”.
@PHDTHESIS{Bollig:679117,
author = {Bollig, Andreas},
othercontributors = {Mathar, Rudolf and Thomä, Reiner},
title = {{S}pectrum sensing in cognitive radio},
school = {Rheinisch-Westfälische Technische Hochschule Aachen},
type = {Dissertation},
address = {Aachen},
reportid = {RWTH-2016-10925},
pages = {1 Online-Ressource (vi, 115 Seiten) : Diagramme},
year = {2016},
note = {Veröffentlicht auf dem Publikationsserver der RWTH Aachen
University 2017; Dissertation, Rheinisch-Westfälische
Technische Hochschule Aachen, 2016},
abstract = {Reliable spectrum sensing is the main enabler for
opportunistic access to the underutilized wireless spectrum.
The task of a spectrum sensing algorithm is to decide
between two hypotheses, the one that the spectral band under
observation is free and can be used by a secondary system
(H0), and the one that the primary system is transmitting on
the band, such that the secondary system needs to refrain
from accessing it (H1). The goal in the design of spectrum
sensing algorithms is to maximize the probability of
detecting a present primary system transmission (probability
of detection Pd) given a fixed probability of wrongly
determining the band under observation to be occupied when
it is not (probability of false alarm Pfa ). In the case of
a missed detection, i. e., when the primary system is
transmitting but the spectrum sensing algorithm decides that
the band is free, the secondary system might also start a
transmission, by which it will disturb the primary system.
When a false alarm happens, the secondary system misses a
chance to use the spectrum. In this thesis, contributions
have been made to three types of spectrum sensing
algorithms. The first type of spectrum sensing we consider
is cyclostationarity detection. Cyclostationarity is a
stochastic feature present in all man-made signals, e.g.,
wireless communication signals, but is absent in pure
stationary noise. Due to this property it can be used to
decide between H0 and H1, which makes it a good fit for
spectrum sensing. The problem arising is that in order to
determine the presence or absence of cyclostationarity in a
received signal, it has to be known beforehand which cycle
frequency is affected. In blind spectrum sensing it is
assumed that the secondary system possesses no knowledge
about the primary system signal, which, for the above
reasons, rules out the use of cyclostationarity. Based on
methods from the field of compressed sensing, two algorithms
for tackling this problem are proposed. In a second step, a
modification of a classic test for cyclostationarity is
devised to estimate the test statistic. This modification is
necessary to work around the problem that when using the
compressed sensing cyclic autocorrelation estimation
algorithms, information required for estimating the spectrum
sensing test statistic is lost. Furthermore, to assess the
cyclic autocorrelation estimation performance of the
aforementioned algorithms, a closed-form expression of the
discrete-time cyclic autocorrelation of linearly modulated
signals with a rectangular pulse shape is derived.
Eigenvalue-based spectrum sensing builds on the idea that a
communication signal induces either correlation in time or
correlation between different receivers, while pure i.i.d.
noise does not. The eigenvalues of a received signal’s
covariance matrix are used to define various test statistics
for spectrum sensing. One of these is the condition number
used in the maximum-minimum-eigenvalue (MME) detector. The
MME detector is independent of uncertainty regarding the
receiver noise power. In contrast, this uncertainty has been
shown to lead to an SNR-wall in the energy detector. An
SNR-wall constitutes the SNR-value that separates the regime
where a detector can robustly detect a primary system signal
and the regime where it cannot. Obviously, not exhibiting an
SNR-wall is a desired feature of spectrum sensing
algorithms. Unfortunately, the MME detector does not possess
this feature. Indeed, in this work we show that the MME
detector suffers from an SNR-wall induced by uncertainty
regarding the amount of coloring of the receiver noise. A
lower bound on this SNR-wall is derived and examples for
different types of covariance matrices are given. Moreover,
it is shown that low amounts of man-made impulsive noise
already lead to enough uncertainty in the noise coloring
that an SNR-wall considerably far above the desired regime
of operation is brought about. Furthermore, two new test
statistics for spectrum sensing based on the eigenvalues of
the received signal’s covariance matrix are proposed. One
of the oldest test statistics used in spectrum sensing is
the received signal power. The corresponding method goes by
the name of energy detection. It consists of measuring the
received energy in a spectral band and comparing it to a
predefined threshold. One of the problems occurring in
spectrum sensing is the so-called hidden terminal problem,
which leads to an SNR between the active node of the primary
system and the secondary system sensor that is too low for
reliable detection. In order to avoid the problem, a set of
spatially distributed sensors is deployed. To exploit the
spatial diversity, the sensors have to transmit either a
local decision on the spectrum occupancy or their
measurement data to a fusion center for combined analysis
and decision making. To minimize the resulting overhead in
spectrum usage, compressed sensing methods are utilized.
Finally, the architecture of the simulation framework used
for most numerical evaluations presented in this work is
described. It facilitates the reuse of code and benefits its
stability.},
cin = {613410},
ddc = {621.3},
cid = {$I:(DE-82)613410_20140620$},
typ = {PUB:(DE-HGF)11},
urn = {urn:nbn:de:hbz:82-rwth-2016-109259},
doi = {10.18154/RWTH-2016-10925},
url = {https://publications.rwth-aachen.de/record/679117},
}
guest ::