Advances in avian acoustic recognition through artificial intelligence: a systematic review of techniques and environmental applications

The accelerating loss of biodiversity has driven the adoption of automated technologies for environmental monitoring, including the acoustic recognition of birds using artificial intelligence. This study aimed to systematically review the primary methods employed in automatic recognition of bird vocalizations, with an emphasis on the evolution of techniques and their environmental applications. The integrative literature review covered publications from 2013 to 2025, with searches conducted on the databases Scopus, Web of Science, IEEE Xplore, and Google Scholar, using terms related to machine learning and bioacoustics. After screening 2,435 publications, 25 studies were selected for in-depth analysis. The findings indicate a methodological shift from Mel-Frequency Cepstral Coefficients to Convolutional Neural Networks, highlighting improvements in classification accuracy and noise robustness. It is concluded that neural networks are becoming increasingly effective tools for biodiversity conservation, although challenges remain regarding model generalization and computational cost.