Harnessing diversity of sounds in complex and vast oceans for efficient sensing and classification
Abstract
We propose to (1) understand, characterize and quantify the sounds produced by and/or reflected and scattered from the plethora of" underwater objects - biological, geological and man-mad entities, as well as boundaries and processes in the frequency range from 10 Hz to over 10 kHz (2) develop mathematical, physical and computer science based models and algorithms for rapidl detecting, classifying, identifying and localizing objects, boundaries and processes from the measured sound field over instantaneous wide areas of the ocean, (3) provide statistical inferences about properties of the natural and man-made objects and processes, including temporo-spatial dependencies and correlations, and (4) derive fundamental physical mechanisms and further develop/enhance statistical, stochastic and physics-based analytic and numerical models for broadband acoustic propagation and its moments in random range-dependent ocean waveguides, coherent and incoherent object scattering and reverberation, and natural sound generating mechanisms.We will examine both deterministic and random signals in large ocean acoustic data sets concurrently of biological, geophysical and man-made origins acquired using large-aperture coherent hydrophone arrays. A coherent hydrophone array typically detects several hundred thousand to several million significant acoustic signals per day in the 10 Hz to 10 kHz frequency range. We will develop novel and enhanced approaches for automatic signal detection, representation, feature extraction and clustering, classification and identification, localization and geographic mapping for diverse ocean sound sources, objects and boundaries. The strength and novelty of the approach stems from combining concurrent passive and active ocean acoustic sensing over instantaneous wide-areas, 100 km or more in diameter, along with inputs from other acoustic/ultrasonic and non-acoustic sensing modalities. The goal is to rapidly acquire and maximize information from both sound producers and sound reflectors/scatterers distributed over vast areas of the ocean. We aim to uncover and discover the oceans by application of the automated processing and analysis algorithms developed in this project to large coherent hydrophone array datasets acquired in several major ONR funded ocean acoustic experiments, including the Norwegian and Barents Seas Arctic 2014 and the US East Coast Gulf of Maine 2006 Experiments.Our approach utilizes and incorporates advances in ocean acoustics and remote sensing, marine biology and ecology, oceanography and geophysics, marine structural mechanics and dynamics, as well as computer science and engineering, making analysis of such large datasets efficient and possible in the time frame of this project. Physical mechanisms and stochastic-acoustic models for sound production, scattering, reverberation, attenuation and dispersion, including multiple scattering where applicable, based on Green~s theorem will be investigated and developed/enhanced using inputs from the measured acoustic data. The results from instantaneous wide area passive and active acoustic sensing will be fused along with those from other sensing modalities, such as ultrasound, optical and radar, to provide validation of the methods and theories we deve"lop, as well as expand wealth of information about the detected targets. The goal is to make the ocean more transparent.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Mar 11, 2020
- Source ID
- N000142012026
Entities
People
- Purnima Ratilal
Organizations
- Northeastern University
- Office of Naval Research
- United States Navy