The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

MSc Thesis #9655

Title:Classification of Humpback Whalesong Units Using a Self Organising Feature Map
Date: 1996
Abstract:In studying animal vocalisation for the purposes of discovering more about both animal communication and behaviour, classifying vocalisations is particularly important. Since the acoustic signal is propagated so efficiently through water it is especially significant for marine mammals. Until now, traditional statistical techniques have primarily been used for clustering of marine mammals vocalisations. The following study attempts to examine the utility of the Self Organising Kohonen Map as a clustering technique for classifying humpback whalesong units. The system used to do this has been successful in terms of extracting units from whalesong. The Map was also found to be successful in classifying these units, although the degree of its success cannot be evaluated until either the utility of the map is assessed in terms of being able to classify relatively unknown classes, or until there are more "objective" criteria for establishing what classes do in fact exist.

[Search These Pages] [DAI Home Page] [Comment]