Comparison of neural network topologies for the classification of frogs by their songs

Abstract

At present the capture and recognition of sounds from different animal species have great value in making biological and environmental research in certain areas affected by humans. In this paper, two topologies for five species of anurans are proposed using a database that contains samples of sounds emitted by them. For the development of systems is performed processing audio signals to digitize it; then, the samples to extract the main features that will be our input system for subsequent classification methods using back-propagation and self-organized maps are processed.