Listening to the World's Oceans: Searching for Marine Mammals by Detecting and Classifying Terabytes of Bioacoustic Data in Clouds of Noise

• Activity Number: 60
• Type: Topic Contributed
• Date/Time: Sunday, August 3, 2014 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistical Computing
• Author(s): Christopher W. Clark*+ and Peter J. Dugan
• Companies: Cornell University and Cornell University
• Keywords: Bioacoustics ; machine learning ; big data ; signal processing ; MATLAB

Abstract

The Cornell University Bioacoustics Research Program (BRP) is collecting terabytes of ocean acoustic data containing the sounds of large baleen whales and other marine mammals in order to understand how human activities affect the ocean's acoustic ecosystem. A dataset that would have taken months to process can now be processed multiple times in just a few days using different detection algorithms. In this paper we describe how BRP data scientists use MATLAB to develop high-performance computing software to process and analyze terabytes of acoustic data. This included the use of signal and image processing algorithms and machine learning techniques to detect and classify animal signals in the presence of various levels of background noise, much of it from commercial shipping and seismic airgun surveys prospecting for offshore oil & gas. To evaluate the algorithm accuracy we used statistical tools to compute a suite of performance curves. After optimizing the algorithms on small data sets, we ran them against several full archived data sets on a 64-node cluster. BRP also collaborated with Marinexplore and the Kaggle community to sponsor a worldwide competition in which more than