Screening large data sets and real‐time data streams for bioacoustic signals.

Abstract

This presentation will discuss a major challenge of passive‐acoustic monitoring systems: the analysis of large data sets to identify the occurrence of bioacoustic signals of interest. Triggered by the rapid development of digital audio technology and the increasing capacity of memory devices, it has become easier than ever to produce large, long‐term acoustic data sets. However, analyzing these data sets is challenging, as suitable automated detection and classification systems are needed to perform the analysis in a reasonable amount of time. One such large data set is currently produced by the Perennial Acoustic Observatory in the Antarctic Ocean. This passive acoustic observatory has been operated by the Alfred Wegener Institute since January 2006 and has generated more than 10,000 hours of data to date. The observatory features a data link via satellite, which allows analysis of the recorded hydroacoustic data in real time in Germany. However, to be able to run custom algorithms over the entire data set in several times real‐time speed, a distributed computing system was developed and applied. Here we provide a detailed description of this system and discuss further possible applications.