Clustering Dolphin Signature Whistles with Dirichlet Process Mixtures
摘要
Smooth functions can represent dolphin signature whistle frequencies. Here, we consider each whistle evaluated at the same number of equally spaced time-points. These frequencies and the original sound-lengths are modeled by a mixture model based on the Dirichlet process estimated by MCMC. The results are promising and show the ability of the model to differentiate between different shapes of the sounds.