Investigating Practical Underwater Acoustic Signal-Processing Applications of Information Geometry

Abstract

In this work, three novel signal-processing techniques for detection, matched field processing (MFP), and depth discrimination are applied to data from the SWellEX-96 experiment. The novelty in all three cases stems from the use of information geometry to derive processors that use full-rank, cross-spectral density matrices (CSDM). Information geometry is a unique approach to probabilistic reasoning wherein distributions are viewed as points on a Riemannian manifold. This allows for hypothesis testing and for the reformulation of parameter-estimation problems using the geometric notion of distances between distributions. Five metrics are used in this work to evaluate distances between CSDMs and for the case of detection, it is found that all of the distances fail to produce a better detector than the more conventional Kullback-Leibler divergence. For MFP and depth discrimination, the Euclidean, Wasserstein-2, and root-Euclidean metrics prove successful while the Fisher-Rao and Log-Euclidean metrics fail to produce reliable results.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Dec 16, 2022
Accession Number
AD1192366

Entities

People

  • Daniel J. Brooker

Organizations

  • United States Naval Research Laboratory

Tags

Communities of Interest

  • Energy and Power Technologies

DTIC Thesaurus Topics

  • Acoustic Signals
  • Acoustics
  • Algorithms
  • Ambient Noise
  • Applied Mathematics
  • Curvature
  • Data Sets
  • Detection
  • Detectors
  • Differential Equations
  • Differential Geometry
  • Distribution Functions
  • Engineering
  • Equations
  • Frequency
  • Gaussian Distributions
  • Geometry
  • Information Science
  • Information Theory
  • Military Research
  • Normal Distribution
  • Probability
  • Signal Processing

Readers

  • Acoustical Oceanography.
  • Adaptive Control and Estimation with Uncertainty in Dynamic Systems.
  • Graph Algorithms and Convex Optimization.