Learning About Animals From Unlabeled Acoustic Data

Marie A. Roch, San Diego State University

Stevenson Hall 1300
11:00 AM - 11:50 AM

Recent work has shown that the human preauditory and/or auditory cortex is likely to play a role in acoustic landmark processing, such as the recognition of syllable and phoneme boundaries. Neurons appear to track acoustic envelopes with neural activity corresponding well with acoustic landmarks. These structures have also been observed in non-human primates, suggesting that acoustic landmark processing could be present in non-human primates and have an evolutionary role. Should landmark processing occur in non-humans, it could provide new methods for approaching animal communication. We present the results of a biologically inspired system on a phoneme segmentation task and anecdotal evidence that plausible boundaries are detected for non-human primates. Many species of toothed-whales are poorly understood, with their ranges determined by handfuls of sightings or stranding’s. We demonstrate that properties of echolocation clicks can be used to provide hypotheses about the number of species using an area by analyzing acoustic data from a well-studied area, The Southern California Bight. We show that symmetric Kullback-Leibler similarity metrics from distributional models of toothed-whale encounters can be clustered into species-specific groups that show reasonable concurrence with groups constructed by analysts using known characteristics of echolocation clicks as measured by an adjusted Rand statistic.