Measuring behaviors in songbird groups


Loading...

Author / Producer

Date

2024

Publication Type

Doctoral Thesis

ETH Bibliography

yes

Citations

Altmetric

Data

Abstract

To truly comprehend the brain's complexity, we must closely examine behavior in real-world settings. Neuroscience has traditionally focused on studying behavior in controlled lab environments, but this approach may not fully capture the brain's natural workings. By observing behavior in more natural contexts, we can gain valuable insights into brain function and understand it on a higher, computational level. This top-down approach can then guide our understanding of the brain's intricate mechanisms at lower levels, ultimately leading to a more comprehensive understanding of this complex system. However, to segment fast behaviors in naturalistic environments and to extract individual vocalizations from sound mixtures remain challenging problems. Promising approaches are multimodal systems that record behaviors with multiple cameras, microphones, and animal-borne wireless sensors. In this vein, we designed a modular system (BirdPark) for simultaneously recording the behavior of small animals wearing custom low-power frequency-modulated (FM) radio transmitters in a naturalistic environment. By disentangling the vocal interactions among up to eight birds, our work paves the way for researching complex social behavior. We built a recording arena inside a soundproof enclosure that features seven microphones and five video cameras to capture the entire scene. Moreover, all animals wear a miniature low-power transmitter device that transmits body vibrations from a firmly attached accelerometer via an analog FM radio signal and functions as a contact microphone. The transmitter devices transmit well-separated vocalizations unless these are masked by large body movements such as wing flaps, rare cases of crosstalk, or because of weak mechanical coupling between the bird and the device, which we were unable to completely eliminate. Due to these issues, we miss around 3.6 \% of vocalizations in the transmitter signals. Our custom radio receiver makes use of a multi-antenna demodulation technique that increases the signal-to-noise ratio of the received radio signals by 6.5 dB compared to single-antenna demodulation, thereby reducing signal losses due to fading. Digital acquisition in our system is driven by a single clock, allowing us to exploit cross-modal redundancies for dissecting rapid behaviors. With this system, we recorded a large hypothesis-free dataset of freely-behaving zebra finches. The dataset consists of a total of 369 days of behavioral recordings of birds in different social compositions. Additionally, we present the BenchSeg dataset, a small dataset of high-quality, hand-annotated vocal segment annotations that we used to benchmark the quality of our recordings and the AutoSeg1 dataset, a dataset of machine-annotated vocal segment annotations. Furthermore, we developed a method for detecting copulation events in BirdPark data and retrieved a set of 176 copulation attempts from our dataset that will serve as the basis for a study on the behavioral signatures of copulations.

Publication status

published

Editor

Contributors

Examiner : Hahnloser, Richard
Examiner : Mante, Valerio
Examiner : Schmidt, Marc

Book title

Journal / series

Volume

Pages / Article No.

Publisher

ETH Zurich

Event

Edition / version

Methods

Software

Geographic location

Date collected

Date created

Subject

NEUROETHOLOGIE, VERHALTENSNEUROLOGIE (BIOLOGIE); Songbird; SOFTWARE DEFINED RADIO, SDR (MOBILE COMMUNICATIONS); Zebra finch; Song learning

Organisational unit

03774 - Hahnloser, Richard H.R. / Hahnloser, Richard H.R. check_circle

Notes

Funding

182638 - The roles of vocal communication in pair formation and cultural learning in songbirds (SNF)

Related publications and datasets