Loading...
Compressive sampling of communication signals
Lim, Chia Wei
Lim, Chia Wei
Citations
Altmetric:
Advisor
Editor
Date
Date Issued
2015
Date Submitted
Collections
Research Projects
Organizational Units
Journal Issue
Embargo Expires
Abstract
The theory of Compressive Sensing (CS) has recently enabled the efficient acquisition of signals which are sparse or compressible in some appropriate domain. The central theme of this dissertation addresses the utility of CS sampling protocols for the efficient acquisition of communication signals in various end applications. In particular, this dissertation first begins with a theoretical framework of the class of Temporal Higher Order Cyclostationary Statistics (THOCS) estimators when the CS nonuniform sampler is used to acquire low rate nonuniform samples of cyclostationary signals. The resulting class of estimators are referred to as Compressive Temporal Higher Order Cyclostationary Statistics (CTHOCS). In the sequel, its utility is demonstrated in the compressive inference problem, where one seeks to infer characteristics of the underlying signal from compressive measurements of the signal, in the form of a modulation classification application. This dissertation then addresses the recovery of communication signals in low traffic Time Division Multiple Access networks from compressive measurements assuming a priori knowledge of the structure of the support of such signals. Specifically, such communication signals belong to the Periodic Clustered Sparse (PCS) signal model and the utility of model-based CS is demonstrated where a sampling bound is derived using subspace counting arguments for the recovery of PCS signals. In addition, a heuristic subspace approximation algorithm is proposed over the optimal subspace approximation algorithm which improves overall solver runtime latency. Last but not least, this dissertation addresses the recovery of blind analog Frequency Hopping (FH) signals from Multi-Coset (MC) samples, an existing gap in previous works focusing on blind analog multiband signal reconstruction. In particular, such works have mostly focused on a segment-less recovery framework which is inadequate for the FH signal. Necessitated by the FH signal model, a segment-based recovery framework anchored on a salient aspect of the MC sampler is proposed. In turn, the utility of classical CS reconstruction techniques for the segment-based recovery framework is demonstrated. Further, a Discrete Prolate Spheriodal Sequence based dictionary is proposed to reduce the complexity of the segment-based recovery framework.
Associated Publications
Rights
Copyright of the original work is retained by the author.