Determining spectral properties of Bowhead whale vocalizations through automated methods

Abstract

Determining spectral properties of whale vocalizations is valuable to the study of marine mammal behavior. However, manually measuring these properties is tedious, which motivates the search for automated methods. Linear regression and Radon transform methods were developed to automatically measure the slope of Bowhead whale frequency-modulated (FM) sweeps from spectrogram data. The Radon transform method was further expanded to include classification of whale vocalizations to aid in further analysis. The data collected using these methods were used to study the acoustic dispersive nature of marine mammal vocalizations in an ocean waveguide. Possible relationships were explored between the change in slope of the whale vocalization and range between each animal and the receiver.Determining spectral properties of whale vocalizations is valuable to the study of marine mammal behavior. However, manually measuring these properties is tedious, which motivates the search for automated methods. Linear regression and Radon transform methods were developed to automatically measure the slope of Bowhead whale frequency-modulated (FM) sweeps from spectrogram data. The Radon transform method was further expanded to include classification of whale vocalizations to aid in further analysis. The data collected using these methods were used to study the acoustic dispersive nature of marine mammal vocalizations in an ocean waveguide. Possible relationships were explored between the change in slope of the whale vocalization and range between each animal and the receiver.