Data mining in large audio collections of dolphin signals

Pattern discovery;Dolphin communication

Kohlsdorf, Daniel ; Starner, Thad Computer Science Essa, Irfan Isbell, Charles Herzing, Denise Beetz, Michael ; Starner, Thad

University:Georgia Institute of Technology

Department:Computer Science

关键词: Pattern discovery;    Dolphin communication;   

Others  :  https://smartech.gatech.edu/bitstream/1853/53968/1/KOHLSDORF-DISSERTATION-2015.pdf

美国|英语

来源: SMARTech Repository

【 摘 要 】

The study of dolphin cognition involves intensive research of animal vocal-izations recorded in the field. In this dissertation I address the automated analysisof audible dolphin communication. I propose a system called the signal imager thatautomatically discovers patterns in dolphin signals. These patterns are invariant tofrequency shifts and time warping transformations. The discovery algorithm is basedon feature learning and unsupervised time series segmentation using hidden Markovmodels. Researchers can inspect the patterns visually and interactively run com-parative statistics between the distribution of dolphin signals in different behavioralcontexts. The required statistics for the comparison describe dolphin communicationas a combination of the following models: a bag-of-words model, an n-gram modeland an algorithm to learn a set of regular expressions. Furthermore, the system canuse the patterns to automatically tag dolphin signals with behavior annotations. Myresults indicate that the signal imager provides meaningful patterns to the marinebiologist and that the comparative statistics are aligned with the biologists’ domainknowledge.