Great reed warbler. Credit: Patrik Lauha

In a study conducted at the University of Helsinki, AI was trained to classify bird sounds with increasing accuracy. The results of the study have been used, among others, in the "Muuttolintujen kevät" (Spring of Migratory Birds) mobile application, which has become a substantial platform for collecting bird recordings.

In his doctoral thesis, Doctoral Researcher Patrik Lauha demonstrates that bird sound classification models can be improved by adapting them to specific monitoring contexts. Fine-tuning and localizing improves the accuracy of the classification models.

"An important by-product of my research was the mobile application called ‘Muuttolintujen kevät,’ for which I developed the AI model that identifies bird sounds. The app’s popularity proves that automated sound recognition models can be used to boost extensive citizen science–based collection of data," affirms Lauha.

Automation and AI revolutionize bird monitoring

Many bird species serve as bioindicators, meaning that changes in their populations broadly reflect the state of whole ecosystems. In recent years, passive acoustic monitoring with automated recording equipment has become a mainstream method in monitoring biodiversity and bird populations, thanks to the inexpensive and easy-to-use devices. Passive acoustic monitoring is primarily done by leaving automated microphones in the forest. The microphones record sound data without the need for human presence.

Passive acoustic monitoring generates large quantities of data, amounting to years’ worth of recordings. To analyze such data requires the introduction of automated species recognition methods.

"My aim was to improve the existing bird sound classification models and to tailor methods for sound data that have been collected through automated methods," explains Lauha.

The workflows developed in the thesis enable constructing more accurate classification models for various regions and monitoring scenarios. Increasingly reliable and accurate species classification offers a more solid foundation for audio-based bird monitoring and research.

Citizen science supports research with the help of an app

While the primary users of the methods developed in the study are scientists conducting bird research and collecting sound data around the world, a practical example of the application of research results is the mobile app "Muuttolintujen kevät," which makes state-of-the-art scientific research tools available to the public.

Users of the app can record bird sounds and obtain AI-provided species classifications for their recordings. The data collected through the app complements traditional bird monitoring in Finland, thereby opening up new research initiatives from now on. The recordings can be used to investigate, among others, bird populations, migratory behavior and singing activity.

"As inventions in many other fields, automated classification models are not intended to replace the surveying and monitoring work done by experts, but rather to enhance, facilitate and intensify their work so that we can monitor the state of our environment as reliably and efficiently as possible—a vital capacity in our world shaken by the ongoing biodiversity crisis," Lauha summarizes.

Patrik Lauha, M.Sc., will defend his doctoral thesis entitled "Improving bird sound classifiers for passive acoustic monitoring’ on 9 January 2026 at the Faculty of Biological and Environmental Sciences, University of Helsinki. Assistant Professor Aki Härmä from Maastricht University will serve as the opponent and Otso Ovaskainen as the custos.

Citation: AI boosts bird sound identification accuracy (2026, January 7) retrieved 7 January 2026 from https://phys.org/news/2026-01-ai-boosts-bird-identification-accuracy.html

This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.