SCRIPPS INSTITUTION OF OCEANOGRAPHY FACULTY CANDIDATE SEMINAR - Joint hire SIO/ECE: Signal Processing in Applied Ocean Science
DATE: May 24th, Friday, 12:45 p.m.
LOCATION: Eckart 227
SPEAKER: Shima Abadi
University of Washington
TITLE: Acoustic Array Processing in the Ocean
Any sound produced in the ocean is confined between the ocean surface and the ocean floor and can travel great distances. Hydrophone arrays are widely used in the ocean to collect sound in both temporal and spatial scales. Array signal processing is used to estimate spatial characteristics of the recorded signals. In this presentation, I describe an application of array signal processing in marine seismic surveys. Marine seismic surveys use horizontal hydrophone arrays to image the structure of the seafloor by generating repetitive high energy sound signals from airguns and recording their reflections off the ocean floor. I use experimental data to characterize short-range propagation of the airgun pulses and take data-driven approaches to predict acoustic energy radiated from airguns. I highlight limitations of classical signal processing in analyzing data recorded by unstructured ocean sensor arrays such as arrays of autonomous underwater vehicles, gliders, and drifting buoys. Classical signal processing methods, which have benefited ocean sciences in the past few decades, are not able to take the structure of the irregular arrays into account. Graph signal processing is an emerging field that overcomes the classical signal processing limitations in working with increasingly complex structures of arrays (high-dimensional data) and fast-growing volumes of data (big data). I introduce a framework for graph signal processing and use it to reformulate the direction-of-arrival (DOA) estimation problem and explore the practical advantages therein.