Fully Connected Neural Networkfor the Classification of Calls from Two Species of Glass Frogs
Year: 2025
Type: article
Abstract: This document presents the application of machine learning (ML) for the classification of calls from glass frog species, Hyalinobatrachium fleischmanni (Hf) and Espadarana prosoblepon (Ep) based on audio recordings. For this, a dataset of acoustic data obtained through frequency manipulation (+4 semitones for Hf and -4 semitones for Ep) and the incorporation of environmental noise (white noise/pink noise) was used. The ML model was trained with original and modified frog calls in order to distinguish the calls under variations in the acoustic signals. Model evaluation was carried out using F1-score, precision, and recall metrics. The results show the model's ability to classify frog calls with high accuracy (98%).
Source: Revista Ciencia y Tecnología
Authors: Byron Jiménez Oviedo, Katalina Oviedo Rodríguez, Jorge Arroyo Hernández, Federico Mora Mora
Institution: Universidad Nacional
Language: English
Topic: Music and Audio Processing
Subfield: Signal Processing
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4409627214