Abstract
Reliable communication in shared, congested, and interference-limited spectrum increasingly requires receivers that can operate without coordination, without accurate channel state information, and without assuming that the interference is weak, predictable, or known a priori. This dissertation develops and evaluates a family of signal alignment methods for such settings. The central idea is to embed the same desired information-bearing waveform into two or more views, where each view is a matrix-valued received observation formed from multi-antenna samples over a repeated time, frequency, frequency-hopping, or space-time-coded block. Because the desired signal repeats while unpredictable interference, which may be comparable to or stronger than the signal of interest, changes across views, a multi-antenna receiver can recover the desired shared component through canonical correlation analysis (CCA) or generalized CCA (GCCA), thereby aligning the signal of interest rather than attempting to align or explicitly decode the interference.
The dissertation advances signal alignment along several dimensions. First, it studies binary signal alignment and shows how the binary decoding problem can be solved optimally in polynomial time, while also developing a linear-time solution that is quasi-optimal. Second, it develops adaptive CCA algorithms that exploit shift structure to make sliding-window synchronization computationally scalable. Third, it adapts signal alignment to frequency-hopped low-power wide-area IoT networks, including single-user and multi-user detection. Fourth, it extends the framework to broadband OFDM systems by using frequency-domain repetition for synchronization and decoding under strong unpredictable interference. Fifth, it couples signal alignment with Alamouti space-time block coding, using a structured prefix and CCA-based subspace nulling to suppress co-channel interference before channel estimation and Alamouti combining. Finally, it analyzes multi-view GCCA through a geometric corruption limit and proposes entropy-regularized weighted MAXVAR GCCA to provide robustness when some views are unreliable due to fading or nonuniform view gains.
Across these contributions, the dissertation shows that repetition, when paired with multi-view subspace processing, is not merely redundancy. It is a structural resource that enables interference-resilient synchronization, detection, and decoding in practical wireless systems ranging from ultra-narrowband IoT to multicarrier and transmit-diversity links. By combining theoretical analysis, algorithm design, simulation, and software-defined-radio validation, this dissertation establishes a theory-to-practice path showing how signal alignment can move from mathematical principle to practical wireless implementation.
