|dc.description.abstract||Synchronization is a challenging but very
important task in communications. In digital communication systems, a hierarchy of synchronization problems has to be considered: carrier synchronization, symbol timing synchronization and frame synchronization. For bandwidth efficiency and burst transmission reasons, the former two synchronization steps tend to favor non-data aided (NDA or blind) techniques, while in general, the last one is usually solved by inserting repetitively known
bits or words into the data sequence, and is referred to as a data-aided (DA) approach.
Over the last two decades, extensive research work has been carried out to design nondata-aided timing recovery and carrier synchronization algorithms. Despite their importance and spread
use, most of the existing blind synchronization algorithms are derived in an ad-hoc manner without exploiting optimally the entire available statistical information. In most cases their
performance is evaluated by computer simulations, rigorous and complete performance analysis has not been performed yet. It turns out that a theoretical oriented approach is indispensable for
studying the limit or bound of algorithms and comparing different methods.
The main goal of this dissertation is to develop several novel signal processing frameworks that enable to analyze and improve
the performance of the existing timing recovery and carrier synchronization algorithms. As byproducts of this analysis, unified methods for designing new computationally and statistically efficient (i.e., minimum variance estimators)
blind feedforward synchronizers are developed.
Our work consists of three tightly coupled research directions. First, a general and unified framework is proposed to develop optimal nonlinear least-squares (NLS) carrier recovery scheme for burst transmissions. A family of
blind constellation-dependent optimal "matched" NLS carrier estimators is proposed for synchronization of burst transmissions fully modulated by PSK and QAM-constellations in additive white Gaussian noise
channels. Second, a cyclostationary statistics
based framework is proposed for designing computationally and statistically efficient robust blind symbol timing recovery for time-selective flat-fading channels. Lastly, dealing with the problem of frame synchronization, a simple and efficient data-aided approach is
proposed for jointly estimating the frame boundary, the frequency-selective channel and the carrier frequency offset.||en