ePoster

Analyzing vocalization behavior of jackdaws (Corvus monedula) using CrowTone, a suite of modern techniques including deep neural networks

Lutz Wehrland
FENS Forum 2024(2024)
Messe Wien Exhibition & Congress Center, Vienna, Austria

Conference

FENS Forum 2024

Messe Wien Exhibition & Congress Center, Vienna, Austria

Resources

Authors & Affiliations

Lutz Wehrland

Abstract

Recent research has shown that corvids have comparable cognitive abilities to primates and even human children. However, birds differ significantly from mammals in brain size and structure, making them an interesting but also important model organism for comparing cognitive abilities across species. Corvids learn their vocalizations form their parents and other tutors and do not lose the ability to learn or mimic sounds over their lives. Beside that very little is known about the vocal cognition and behavior of these birds.I will show initial evidence that corvid and human vocalizations share similarities in phonetic structure. Furthermore, I will show how we study the vocal behavior in Jackdaws using three fundamental approaches: Phonetic and spectro-analytical clustering of the calls, examining proto-syntactic structure like turn-taking and using our gathered data to train deep learning neural network models to analyze more data and with that higher complexity in vocal behavior. Our results serve as a basis, on the one hand, to investigate the communication behavior in more natural contexts such as flocks and, on the other hand, to work more on the neuronal cognition, for example to shape vocal behavior in experimental setups like “skinner boxes”. If we were able to understand communication and neural mechanisms at a higher level of complexity, corvids could compare favorably to human language due to comparable cognitive abilities and shared phonetic features.

Unique ID: fens-24/analyzing-vocalization-behavior-jackdaws-789c00e5