Singing in Sync: How synchronized microphones and machine learning scale up studies of wild bird behavior

Collecting information on individual animals’ vocal and movement behavior is important in many ecological and behavioral applications. However, gathering this information in wild animals may be especially challenging, time-consuming, invasive, or disruptive. This talk will describe a non-invasive method we engineered for gathering behavioral data for individual wild birds at large scales. We gather acoustic data using an array of 49 autonomous recorders. Audio is precisely time-stamped and synchronized using GPS receivers onboard the sound recorder. To estimate birds’ positions, we created tools for automated song detection and localization in OpenSoundscape, an open-source Python package for bioacoustic analyses. In particular, we create machine learning algorithms to detect the songs of each species, use a spatial-matching algorithm to match up detections of the same sound event recorded on multiple microphones, then estimate the position from which each sound event was sung. I will discuss several applications of these new techniques, including mapping individual birds’ home territories, estimating bird population sizes, observing birds’ movement and behavior, and creating libraries of individual birds’ distinctive voices. Speaker(s): Tessa Rhinehart, Agenda: mac Virtual: https://events.vtools.ieee.org/m/413370