Bird song attractors

Reconstructing hidden dynamical patterns in bird sounds

The project explores bird songs using nonlinear signal analysis. By reconstructing geometric patterns from audio recordings, it investigates whether bird vocalizations contain structure beyond what is visible in standard waveforms and spectrograms.

Bird sounds are shaped by interacting processes involved in vocal production. This project asks whether those processes leave recognizable dynamical signatures in the sound, and whether those signatures are visualized and compared across recordings and species.

Figure: From bird sound to reconstructed dynamical pattern. A single chirp can be viewed not only as a waveform or spectrogram but also as a geometric trajectory.

Project overview

Examples

Methods

Results

Demos

Code

Why this project?

Bird vocalizations are often studied through waveforms and spectrograms, which are extremely useful for describing sound. This project explores an additional perspective, whether methods from dynamical systems can reveal patterns in bird sounds that are less visible in conventional representations. The aim is not to replace standard bioacoustic analysis, but to provide another way to study how vocal behavior is organized in time.

What this project does?

Using time-delay based reconstruction methods, geometric patterns are built directly from bird sound recordings. These reconstructed structures can then be compared across different recordings, species, and simple sound production models.

This approach helps reveal fine-scale variation in vocal patterns and offers a new way to connect sound structure with the dynamics of sound production.

What can be seen?

In some recordings, the reconstructed trajectories form clear loops, spirals, or other repeated structures. These patterns can vary across species and even between consecutive chirps, suggesting that they may capture aspects of vocal organization that are not obvious in raw waveforms alone.