"Learning to Listen"
Shallow oceans are diverse. Variations in water depth, sound speed, and ocean floor composition all affect how sound propagates making it difficult to identify where sounds originate. Models for sound propagation can predict the frequency-dependent sound field for a given set of source-receiver positions and environmental parameters. The inverse problem is actually of more interest: how can the received sound provide information about a source’s location and the properties of the ocean environment? Optimization algorithms have tackled this problem for many years. The future, however, lies with machine learning. For successful application of machine learning techniques, important questions need to be addressed concerning how to train the algorithms such that they learn to “listen” for key features that are indicative of specific ocean environments. To answer these questions, care must be taken to determine the appropriate variability in training data, traditional ways of representing data and labels need to be re-evaluated in terms of the key features in the data, and results need to be interpreted taking into account parameter sensitivity and inherent uncertainty. As my research has focused on answering these questions, I am seeing parallels with how each of us must “learn to listen” in ways that allow for diversity and foster belonging in engineering.
Tracianne Beesley Neilsen received a doctoral degree in physics from the University of Texas at Austin and completed a postdoctoral appointment at the associated Applied Research Laboratories, where her research focused on optimization methods for source localization and environmental characterization from multi-source sound fields in the ocean. Her experience includes management and analysis of large acoustical data sets, numerical modeling of acoustic propagation, and development and employment of multi-dimensional inversion techniques. After working part-time as an adjunct professor for 14 years, Neilsen became a full-time Associate Professor in the Department of Physics and Astronomy in May 2018. With this new position, she has returned to underwater acoustics research. Her main projects include underwater acoustical measurement and the application of machine learning to classifying and locating sounds sources and uncovering seabed properties in variable ocean environments. Most of Neilsen’s professional service has been for the Acoustical Society of America where she served as the Chair of the Women in Acoustics Committee 2018-2021 and is currently a member of the Executive Council.