A computational Bayesian approach to machine learning of an acoustic scatterer's state of motion in a refractive propagation environment under a small aperture constraint

• Published in: The Journal of the Acoustical Society of America 2021-04, Vol.149 (4), p.A85-A85
• Main Authors: Gendron, Paul J., Barros, Abner C.
• Format: Article
• Language: English
• ISSN: 0001-4966; 1520-8524
• DOI: 10.1121/10.0004592

A computational Bayesian method for inference regarding the size and state of motion of a submerged mobile object from narrow band transmissions and a small receiver aperture is presented. The challenge of attendant closely spaced multipath arrivals is addressed by taking advantage of knowledge of the refractive environment. Acoustic phase fronts, their angles and Doppler compressed wavelengths are jointly inferred. We take full advantage of the analytic conditional densities of eigenray amplitudes as well as ambient and reverberant power under the conventional Gaussian-Inverse Gamma model and focus computational resources on conditional densities for the frequency-angle dependent wave vectors via sampling from conditional quantile functions. Inversion of the posterior probability density of the wave vectors to the scatterer's state space is accomplished quickly via eigen-ray interpolation. Examples are given from real ocean sound speed profiles as well as the classic Munk profile.