Sound Speed Profile Determination using a Large Database of Underwater Recordings of Vessel Traffic with Machine Learning
Abstract
Ocean acoustic tomography typically relies on precise travel time measurements of curated signals between a source and receiver to" measure ocean temperatures and currents. This method requires tightly-synchronized instrumentation across large distances, and computationally intensive waveform inversions. In addition to affecting sound velocities, ocean conditions also alter the frequency content of broadband signals through frequency-dependent acoustic absorption, scattering and interference. Deep learning algorithms are effective pattern recognition tools which may be ble to leverage information contained in the frequency content of received signals to make inferences aboutlocal oceanographic conditions, if trained on large numbers of observations. An existing dataset of thousands of broadband acoustic recordings of vessel passages in a shipping channel off of Southern California from 2014 to 2018, with coincident temperature and saliniy profiles, provides an opportunity to investigate the feasibility of this approach. The proposed project will explore the use of deep neural networks to make inferences about local ocean conditions based on received characteristics of vessel signatures as acoustic sources of opportunity, supplemented by AIS-derived vessel descriptions. The proposed project will potentially identify a method for estimating in situ water column properties using passive acoustic data collected at a single, non-profiling sensor location using opportunistic sound sources. The ability to estimate local water column properties using a passive acoustic point sensor would allow these properties to be measured more widely, inexpensively, and frequently than currently feasibl"e with traditional methods.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 20, 2019
- Source ID
- N000141912582
Entities
People
- Kaitlin E Frasier
Organizations
- Office of Naval Research
- United States Navy
- University of California, San Diego