| Summary: | The songs produced by male humpback whales, Megaptera novaeangliae, during the breeding season have been increasingly studied in the last couple of decades with methods based mainly on their spectrographic characteristics. Songs were characterised as series of units - defined by Payne as continuous sounds between two silences - associated in specific patterns, similarly to the way in which bird songs are formed. Previous work highlighted the need for objective methods for humpback whale sound units classification; for this purpose, various automatic clustering algorithms were developed to study how units are associated to produce themes that are repeated throughout the song duration and to compare them across whale populations and from year to year. However, detailed analysis of the vocalisations showed that the features of a unit can change abruptly throughout its duration making it difficult to characterise and cluster them systematically. We propose a new approach for song segmentation based on the identification of subunits that are characterised by looking at the changes of their frequency content through time. Typically units within a song consist of several elements - which we refer to as subunits - that have distinct structure, i.e. units are formed as combination of subunits. The distinction between subunits and units should improve the accuracy of classification algorithms, especially for those vocalisations that present a complex structure that varies significantly with time.
|
|---|