Example annotations of vocal activity in multi-modal data from mixed-sex zebra finch pairs

This guideline provides an educational resource and a set of instructions for detecting and annotating vocalizations of paired zebra finches in multi-modal recordings from BirdPark (Rüttimann, et al., 2025). The main purpose of this guideline is to instruct the reader how to detect and annotate the onsets and offsets of zebra finch vocalizations and consolidate the vocalization using multi-modal information. By studying this document and the associated example annotations, we hope to minimize the annotation variance among annotators on larger datasets of this kind to be published. The published dataset is taken from experiments aimed at examining vocal and non-vocal behaviors during pair bonding. We recorded freely moving birds in a custom isolation chamber (“BirdPark”) equipped with multiple microphones, cameras, and radio antennas. Vocalizations were captured both with stationary microphones attached to the wall (4-5 microphones depending on the experiment setup) and with miniature sensors attached to birds' backs, measuring body vibrations with an accelerometer and wirelessly transmitting the vibration signals via FM radio. We acknowledge support of the National Science Foundation NCCR Evolving Language, Agreement #51NF40_180888.