Classification of odontocete buzz clicks using a multi-hypothesis tracker

Abstract

Blainville's beaked whale (Mesoplodon densirostris) buzz clicks have been found to have characteristics that can vary significantly. While we have not succeeded to classify them individually, we find that their spectrum is very similar from one click to the next. In previous work, we showed that a multihypothesis tracker can be used to associate these clicks, and subsequently to classify the click sequence. This paper describes further tracker enhancements and shows improved performance results. Further, we find that the property of slowly varying buzz clicks spectrum is also true for other odontocete species,and apply our multi-hypothesis tracking algorithm to such data.