High level automatic structuration of ocean passive data : From click sequence modulations to whale behavior analyses

Abstract

This paper discusses on estimating the behavior of a sperm whale using hydrophones recordings thanks to a set of features extracted from a real-time passive underwater acoustic tracking algorithm for multiple emitting whales. Acoustic localization permits to study whales' behavior in deep water (several hundreds of meters) without infering with the environment. Here, we use a real-time multiple tracking algorithm, which provides a localization of one or several sperm whales. Thanks to the positions coordinates and the audio files, we are able to automatically label the signal and to extract different features such as the speed, energy of the clicks, Inter-Click-Interval (ICI)... These features allow us to cross-analyse the whale behavior (foraging , hunting, ingestion) and to see the influence of each parameters and the dependency between them. Finally, we generate a XML file for description and efficient access of deep ocean sound records for cetacean studies, which leads to index and structure the audio files. Thus, the behavior study is facilitated choosing and accessing the corresponding index in the recording. The complete method is processed on real data from the NUWC 1 and the AUTEC 2 . As an illustration, we process the algorithm on a one whale case and study the correlation between the features and the whale behavior during the diving.

Désentrelacement de clics par analyse du rythme

“…Afin d'automatiser ces analyses, la communauté bioacoustique a développé des outils de traitement du signal pour détecter les clics d'odontocètes, par exemple (Kandia, Stylianou, 2006 ;Gervaise et al, 2010). Le suivi des trains de clics a été majoritairement étudié à des fins de localisation des cachalots lors de leurs plongées (Laplanche et al, 2005 ;Giraudet, Glotin, 2006 ;Caudal, Glotin, 2008a ;Morrissey et al, 2006 ;Caudal, Glotin, 2008b ;Nosal, Frazer, 2007), mais aussi des baleines à bec (Ward et al, 2008) et plus récemment des belugas (Roy et al, 2010). Ces travaux utilisent la plupart du temps un réseau de plusieurs hydrophones et associent les clics appartenant au même train en calculant l'autocorrélation entre paires d'hydrophones.…”