Quantifying the group specificity of animal vocalizations through relative entropies

Recordings of animal vocalization can lack of important information about sender and context, in particular in the increasing number of bio-acoustic monitorings and in studies on marine mammals. Here, we develop a framework to estimate group specificity without specific sender information. We introduce and apply the bag-of-calls-and-coefficients approach (BOCCA) to study ensembles of cepstral coefficients composed from vocalization signals recorded from a given animal group. Comparing distributions of such ensembles of coefficients by computing relative entropies reveals group specific differences. We illustrate the ensemble-based method by showing that differences of ensembles of calls within social groups of pilot whales (Globicephala melas) are significantly lower than inter-group differences.