The “Dolphin Whistle Detection and Vocalization Display” project is being led by Dr. Heidi Lyn, University of South Alabama, and Peter Sugarman and aims to better understand the whistle repertoire of a group of dolphins located at the Ocean Adventures facility in Gulfport, Mississippi. Our role in this project relates to the software components used in detecting and displaying whistles. See how we are using accessible analytical software for this project below. Elizabeth Ferguson and our partner Jennifer Pettis Schallert and Gabi Alongi are involved in the programming elements of this project.
DeepSqueak is an open-access bioacoustic analysis software program currently run in MATLAB. This tool was originally developed to detect and classify ultrasonic vocalizations from rodents in a low noise, laboratory setting. We have trained networks in DeepSqueak to detect dolphin vocalizations in comparatively noisy, natural acoustic environments and are using the classification features to establish a repertoire.
The Seewave package available in R is used to display the detected whistles obtained from Ishmael. R and the Seewave package are both open-access and widely used in bioacoustic and statistical analysis. We anticipate generating an additional element to the R routine to assign a percentage match of each whistle to a predetermined set of parameters.
ARTwarp is a freely available MATLAB-based program that uses whistle contour similarity and an Adaptive Resonance Theory (ART) neural network to categorize tonal sounds. It was developed by Volker Deecke and Vincent Janik of the University of St Andrews. ARTwarp provided categorization of whistles which were subsequently analyzed to characterize the group's repertoire.