2026-05-04T18:40:57.881Z http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai

ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai:3:11778 2026-01-05T14:31:04Z 3

Classification of South American Birds via Audio Analysis with Convolutional Networks Optimized by Adapted Firefly Algorithm, Chapter in LNNS Lecture Notes in Networks and Systems: ISBM 2025: Information Systems for Intelligent Systems, Springer, volume 1750 2026 http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai/record/3/11778 https://link.springer.com/chapter/10.1007/978-3-032-12993-2_17 T. Zivkovic M. Zivkovic B. Radomirovic M. Cajic S. Adamovic M. Stankovic V. Zeljkovic N. Bacanin Biological diversity has a crucial purpose in all ecological systems. Ecosystems with greater species variety tend to be more robust against environmental fluctuations and typically reflect better ecological well-being. Nevertheless, tracking biodiversity is often challenging, particularly when trying to avoid interference with natural habitats. Traditional methods like specimen collection or animal tagging can be intrusive and can cause damage to the ecosystem. This study investigates the feasibility of leveraging acoustic signal processing as a nonintrusive strategy for estimating biodiversity levels. A unified methodology is introduced, which combines Mel-frequency cepstral coefficients (MFCCs) with convolutional neural networks (CNNs), enabling classification of bird species through their vocalizations captured in audio recordings. To enhance the model’s predictive capabilities, an adapted variant of the firefly optimization algorithm (FA) is employed for fine-tuning the model parameters. The top-performing configurations achieved promising results on a real-world dataset, reaching an accuracy of 64.7%, indicating the method’s practical applicability in real environmental monitoring scenarios. bookPart Springer, Cham 169 181 10.1007/978-3-032-12993-2_17 LNNS Lecture Notes in Networks and Systems: ISBM 2025: Conference Proceedings of World Conference on Information Systems for Business Management, volume 1750