Unsupervised Acoustic Analysis of the Vocal Repertoire of the Gray-Shanked Douc Langur (Pygathrix cinerea)

Unsupervised analyses of nonhuman vocal repertoires are important not only because they allow decoding species-specific displays, but also to understand the evolution of communication signals, which need objective quantitative investigations. We recorded more than 1000 vocalizations emitted by captive gray-shanked douc langur (Pygathrix cinerea). Putative vocal categories were first identified by their overall acoustic structure, considering duration and spectral properties. We processed each call using an automatic pairwise Dynamic Time Warping comparison, limiting the use of a priori assumptions. We then performed a cluster analysis that indicated the presence of 25 clusters, which we then labeled to seven vocal types according to spectrographic information.