Abstract
The Bayesian view provides a framework for underwater signal analysis by incorporating relevant and disparate streams of information into the inferential process. Presented here is a Bayes Factor Active Sonar (BFAS) inference scheme for high-frequency broadband analysis from relatively short vertical arrays. The BFAS properly accounts for environmental information regarding the refractive media, as well as surface and volume reverberation models. The BFAS addresses target depth uncertainty through proper marginalization rather than maximization as in the conventional Generalized Likelihood Ratio Test (GLRT). BFAS operates as a set of time-varying quadratic forms in beam-delay space, optimally balancing target, reverberation, and noise sub spaces. By using waveguide information, the system optimally, in the minimum average risk sense, attenuates reverberation subspaces while preserving the target subspace, effectively increasing Signal-to-Reverberation+Noise Ratios despite target depth uncertainty. Depth-Invariant Modes (DIM) are leveraged for a computationally fast BFAS factorization. Performance testing across various refractive and shallow water environments is demonstrated and lends credence to the approach. [Funded by the Office of Naval Research]
1st Place - ASA Student Paper Presentation Award for Signal Processing