Sound source localization

This repository contains the code, the data and the readme to reproduce the figures of the paper : Lellouch & al. Sound source localization in a natural soundscape with autonomous recorder units based on a new time-difference-of-arrival algorithm

Table of Contents

• Sound source localization
  • Table of Contents
  • Installation
  • Purpose of the notebooks
  • Data Used for the notebooks
  • Contributing
  • Contact

Installation

1. Clone the repository:

   git clone https://github.com/ear-team/sound_source_localization.git

2. Navigate to the project directory:

   cd sound_source_localization

3. Install the required dependencies:

   • numpy >= 1.26.4
   • scipy >= 1.13.0
   • librosa >= 0.10.2.post1
   • matplotlib >= 3.8.4

Purpose of the notebooks

Acoustic sound source localization is an emerging approach in animal behavior and ecology to either monitor individuals or to describe the spatial and temporal structure of natural soundscapes. The localization of sound sources mainly rely on the deployment in the field of microphone arrays and on the subsequent pairwise comparison of signals to infer position through triangulation. However, most of the current methods face the challenge of separating or not the sound sources before estimating their spatial position. On the one hand, source localization with upstream separation is particularly challenging in noisy outdoor environments. On the other hand, no-separation sound source localization can be possible only when calling rates are low. Here, using four outdoor low-cost recorders, we propose a new method which is an intermediate between the separation and no-separation strategies. Our deterministic algorithm combines a rough call separation, the computation and self-consistency check of time difference of arrivals, the separation or aggregation of sound sources, the localization of the sources in Cartesian or polar coordinates, and the final identification by an expert. We tested this method on a 190 minute soundscape recording achieved in a temperate freshwater environment. The automated analysis revealed the space and time pattern of an amphibian, avian and human complex soundscape. This procedure opens the possibility to deploy similar passive acoustic programs to monitor either predetermined individuals or the composition of dynamic soundscapes.

Data Used for the notebooks

Two Python Notebooks are provided in this repository.

• The notebook sound_source_localization_example.ipynb applied the full process on a small part of the real data in order to get a result in a relatively short amount of time (few minutes depending on the machine).
• The notebook sound_source_localization_full.ipynb is the original notebook that process all the audio. It might take several hours to several days depending on the machine used.

Data used for the notebook sound_source_localization_example.ipynb is stored in the data/example_data directory of this repository. It consists on 30-second long mono audio files trimmed from the original audio recorded at the Parc Départemental de la Haute-Île.

Data used for the notebook sound_source_localization_full.ipynb is stored on Zenodo. The archive can be downloaded from Zenodo and then unzipped on you computer locally in the folder data from this repository such that you do not need to change the path in the notebook. Otherwise, you need to change the pass in the notebook.

Contributing

The original contribution of this work was done by Laurent Lellouch at the Muséum National d'Histoire Naturelle. The list of contribution is here

Contact

• Ear team - ear@mnhn.fr