Multichannel blind deconvolution in underwater acoustic channels

Abstract

This thesis developed new techniques for solving the multichannel blind deconvolution problem and implemented these techniques in acoustic waveguide multiple environment. We developed a systematic way to build an efficient and accurate channel models incorporating a priori information about the expected Channel Impulse Responses' (CIRs) arrival-time structure. Based on the linear and bilinear channel models in the underwater acoustic applications, we solved the problem using two approaches. In the first approach, we formulated the problem as solving a system of bilinear equations, which in turn can be recast as recovering a low-rank matrix from a set of linear observations. In the second approach, we formed a cross-correlation matrix from the channel outputs and solved the problem by minimizing a quadratic function over a non-convex set. We demonstrated the efficiency and robustness of both multichannel blind deconvolution methods on realistic acoustic channels in ocean waveguides and experimentally validated the methods using at-sea data. In the end, we investigated methods to learn the subspace of CIRs directly from multiple snapshots in the context of solving multichannel blind deconvolution.