Rhythmic analysis for click train detection and source separation with examples on beluga whales

Passive acoustic monitoring systems are used to study cetaceans through the sounds they produce. Among them, toothed whales emit sequences of acoustic impulses having a rhythmic pattern. As they generally live in pods, click trains from several individuals are often interleaved and recorded together with additional natural or anthropogenic impulsive sources. This paper presents an algorithm that uses the rhythmic properties of odontocete click trains for detecting rhythmic impulse trains embedded in other impulse sounds and de-interleaving click trains from simultaneous clicking odontocetes. The contributions of the article are: (1) a method to detect the presence/absence of rhythmic click trains and to characterize the time – Inter Click Interval (ICI) pattern of click trains; (2) an analytical analysis of the performances of this method (jitter, length of click trains); (3) the demonstration of its efficiency on real data with wild beluga whales recorded in Canada.