Characterisation and classification method of burst pulses produced by Guiana dolphins (Sotalia guianensis)

Abstract

ABSTRACT Although burst pulses are the least studied sounds produced by odontocetes, their characterisation is essential to describe the acoustic repertoire of a species. This study aimed to characterise burst pulses produced by Guiana dolphins and propose a visual method to classify these sounds into structural categories. Data collection occurred during boat surveys in south-eastern Brazil. Five acoustic parameters were extracted: number of pulses, duration, interclick interval, click repetition rate, and minimum frequency. All sounds were visually classified into four categories and used in a discriminant function analysis. In addition, a K-means cluster was used to compare the visual classification with this unsupervised grouping method and confirm the representativeness of each category. According to the discriminant analysis, most sounds were correctly classified (77%) into the four categories. Both methods grouped burst pulses in similar patterns, confirming the use of the categories. Click repetition rate was the most important parameter to separate categories. The variation in this parameter could be related to different behaviours as already suggested for other species. Thus, each category could have a different function in Guiana dolphin communication. The visual classification method is efficient and could be used for other species that produce pulsed sounds.