Comparison of methods for rhythm analysis of complex animals' acoustic signals

摘要: Analyzing the rhythm of animals’ acoustic signals is interest to a growing number researchers: evolutionary biologists want disentangle how these structures evolved and what patterns can be found, ecologists conservation aim discriminate cryptic species on basis parameters such as temporal structures. Temporal are also relevant for research vocal production learning, part which animal learn structure. These structures, in other words, rhythms, topic this paper. How they investigated meaningful, comparable universal way? Several approaches exist. Here we used five methods compare their suitability interpretability different questions datasets test support reproducibility results bypass biases. Three very with regards recording situation, length context were analyzed: two social vocalizations Neotropical bats (multisyllabic, medium long isolation calls Saccopteryx bilineata, monosyllabic, short Carollia perspicillata) click trains sperm whales, Physeter macrocephalus. Techniques compared included Fourier analysis newly developed goodness-of-fit value, generate-and-test approach where data was overlaid varying artificial beats, inter-onset-intervals calculations normalized Pairwise Variability Index (nPVI). We discuss advantages disadvantages show suggestions best visualize results. Furthermore, decision tree that will enable researchers select suitable method data.