الفهرس 
chapter 1Only 14 pages are availabe for public view
 |  | from | 126 |  |  |
| | | from | 126 | | |
Abstract
Providing the dramatic increase in the number of radio devices due to
the current advances in Internet of things (IoT), the shortage of frequency
resources became a challenge to the new communication systems.
To tackle this problem, cognitive radio (CR) has suggested to
reutilize frequency bands left unoccupied by their licensed users raising
the need to e↵ective estimation techniques that can detect these
frequency bands. Additionally, if the spatial domain is investigated
along with the spectral domain, a larger number of CRs can share the
limited vacant bands simultaneously. In this thesis, the problem of
estimating the special and spectral information of the existing users is
considered for CR. The considered estimation problem should be handled
blindly as CR should not have any prior information about the
existing radio devices. Moreover, not only is the blindness the challenge
of the considered problem, but also the need to instantaneously
search a wideband spectrum, which needs high Nyquist rates. In order
to overcome the latter issue, sub-Nyquist methods have been proposed
in the literature. While these methods were capable to detect the desired
parameters from reduced number of samples, they require a large
number of relaxed analog-to-digital converters (ADCs) leading to increasing
hardware complexity. In contrast to sub-Nyquist methods,
which are carried out in the temporal domain, the proposed algorithms
here are applied in the spatial domain of the employed array
reducing the number of required samples at each array element to one.
In addition, the proposed algorithms employ nonlinear Kalman filters
(KFs), which are applied on a proposed spatial state space model, in
two di↵erent scenarios; one of which estimates carrier frequencies and
the corresponding direction of arrivals (DoA) of band-limited source
signals, and the other one concerns with two-dimensional DoA (2DDoA).
In each scenario, two di↵erent types of nonlinear KFs are implemented;
the first is extended Kalman filter (EKF) and the other is
unscented Kalman filter (UKF). Since nonlinear KFs are sub-optimal estimators, their performance can be deteriorated by several factors
such as filter tuning and initialization, the variance of the estimated
variables, and the value of the inter-elements spacing in the employed
array. Using simulations, the e↵ects of these factors on the filter performance
are examined and discussed. Overall, relying on one time
sample in the proposed algorithms eliminates the high-sampling-rate
requirements, whereas exploiting the spatial domain in detecting the
unknown parameters results in a gradual decline in the degrees of
freedom.