Peter L. Tyack
Dept. of Biol., Woods Hole Oceanogr. Inst., Woods Hole, MA 02543
Most analyses of acoustic communication in animals lump sounds into qualitative categories. These are usually based upon aural impressions for human listeners or visual inspection of spectrograms. Quantitative analysis of acoustic features from beluga whale sounds raise serious questions about how discrete and robust the traditional categories of beluga calls are. Other categories, such as the contour of a dolphin whistle, use features that need not relate directly to absolute measure of time and frequency. Studies of vocal development and vocal imitation often benefit from a quantitative measure of similarity between sounds and putative models. Several different methods are compared to analyze whistle contours, including multivariate analysis of time-frequency features, dynamic time warping, and a signal compression approach. Ultimately, all such techniques need to be validated by studies of how each species perceives its own signals.