Examination Committee

Prof Ke YI, CSE/HKUST (Chairperson)
Prof Daniel PALOMAR, ECE/HKUST (Thesis Supervisor)
Prof Chee Wei TAN, Department of Computer Science, City University of Hong Kong (External Examiner)
Prof Ross MURCH, ECE/HKUST
Prof Jun ZHANG, ECE/HKUST
Prof Qi QI, IELM/HKUST
 

Abstract

Reliable channel estimation is crucial to achieving high-data-rate transmissions in popular wireless communication techniques such as multiple-input multiple-output (MIMO) and orthogonal frequency-division multiplexing (OFDM). Although plenty of research has thoroughly investigated relevant channel estimation, some practical aspects that are critical to power efficiency and system performance have not been properly studied in this context. The two specific aspects considered herein are the peak-to-average-power ratio (PAR) and phase noise. Mathematically, these two issues can be well represented by sequences that share some structural similarity and enable a common optimization approach.
 
Transmitted sequences of low PAR (or unimodulus as a special case) are frequently desired to satisfy hardware requirements and maximize power efficiency. Numerous works have studied unimodular sequences with good correlation properties, yet reliable channel estimates may not be available even if we have access to prior knowledge of the channel. First, the problems of unimodular sequence design for MIMO channel estimation are formulated by optimizing the minimum mean square error (MMSE) and conditional mutual information (CMI) criteria, respectively. The obtained optimization problems are non-convex, for which efficient algorithms based on the majorization-minimization (MM) framework are proposed. The developed algorithms are then extended to design tailored low PAR sequences.
 
The other practical issue is phase noise, whose correction is essentially necessary to exploit full advantage of OFDM systems. OFDM channel estimation with simultaneous phase noise compensation has therefore drawn much attention and stimulated continuing efforts. Existing methods, however, are only able to provide estimates of limited applicability due to their heuristic nature and considerable computational complexity. In this thesis, the joint estimation problem is reformulated in the time domain as opposed to the popular frequency-domain approaches. In doing so, much more computationally efficient algorithms are developed based on the MM framework. Furthermore, dimensionality reduction and regularization are also introduced to deal with the under-determined nature in the original estimation.