Bioacoustic fundamental frequency estimation: a cross-species dataset and deep learning baseline

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68081766%3A_____%2F25%3A00636805" target="_blank" >RIV/68081766:_____/25:00636805 - isvavai.cz</a>

Nalezeny alternativní kódy

RIV/60076658:12310/25:43909759 RIV/60460709:41330/25:102117 RIV/00020702:_____/25:N0000077

Výsledek na webu

<a href="https://www.tandfonline.com/doi/full/10.1080/09524622.2025.2500380" target="_blank" >https://www.tandfonline.com/doi/full/10.1080/09524622.2025.2500380</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1080/09524622.2025.2500380" target="_blank" >10.1080/09524622.2025.2500380</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Bioacoustic fundamental frequency estimation: a cross-species dataset and deep learning baseline

Popis výsledku v původním jazyce

The fundamental frequency (F0) is a key parameter for characterising structures in vertebrate vocalisations, for instance defining vocal repertoires and their variations at different biological scales (e.g. population dialects, individual signatures). However, the task is too laborious to perform manually, and its automation is complex. Despite significant advancements in the fields of speech and music for automatic F0 estimation, similar progress in bioacoustics has been limited. To address this gap, we compile and publish a benchmark dataset of over 250,000 calls from 14 taxa, each paired with ground truth F0 values. These vocalisations range from infra-sounds to ultra-sounds, from high to low harmonicity, and some include non-linear phenomena. Testing different algorithms on these signals, we demonstrate the potential of neural networks for F0 estimation, even for taxa not seen in training, or when trained without labels. Also, to inform on the applicability of algorithms to analyse signals, we propose spectral measurements of F0 quality which correlate well with performance. While current performance results are not satisfying for all studied taxa, they suggest that deep learning could bring a more generic and reliable bioacoustic F0 tracker, helping the community to analyse vocalisations via their F0 contours.

Název v anglickém jazyce

Bioacoustic fundamental frequency estimation: a cross-species dataset and deep learning baseline

Popis výsledku anglicky

The fundamental frequency (F0) is a key parameter for characterising structures in vertebrate vocalisations, for instance defining vocal repertoires and their variations at different biological scales (e.g. population dialects, individual signatures). However, the task is too laborious to perform manually, and its automation is complex. Despite significant advancements in the fields of speech and music for automatic F0 estimation, similar progress in bioacoustics has been limited. To address this gap, we compile and publish a benchmark dataset of over 250,000 calls from 14 taxa, each paired with ground truth F0 values. These vocalisations range from infra-sounds to ultra-sounds, from high to low harmonicity, and some include non-linear phenomena. Testing different algorithms on these signals, we demonstrate the potential of neural networks for F0 estimation, even for taxa not seen in training, or when trained without labels. Also, to inform on the applicability of algorithms to analyse signals, we propose spectral measurements of F0 quality which correlate well with performance. While current performance results are not satisfying for all studied taxa, they suggest that deep learning could bring a more generic and reliable bioacoustic F0 tracker, helping the community to analyse vocalisations via their F0 contours.

Druh

J_imp - Článek v periodiku v databázi Web of Science

CEP obor

—

OECD FORD obor

10613 - Zoology

Projekt

<a href="/cs/project/GF21-04023K" target="_blank" >GF21-04023K: Je vnitropohlavní výběr hybatelem evoluce signálů individuální identity živočichů?</a><br>

Návaznosti

I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Rok uplatnění

2025

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název periodika

Bioacoustics-The International Journal of Animal Sound and Its Recording

ISSN

0952-4622

e-ISSN

2165-0586

Svazek periodika

34

Číslo periodika v rámci svazku

4

Stát vydavatele periodika

GB - Spojené království Velké Británie a Severního Irska

Počet stran výsledku

28

Strana od-do

419-446

Kód UT WoS článku

001501315800001

EID výsledku v databázi Scopus

2-s2.0-105007437974