Análise do desempenho de diferentes vetorizadores e diferentes classificadores na identificação de espécies de anuros através do som

Abstract

Tail-less amphibians, known as anurans, are a diverse group that includes toads, frogs, and tree frogs, and when monitored, they serve as excellent indicators of ecosystem degradation and the presence of invasive species. This study determines the best combination in the form of a pair composed of a vectorization algorithm and a classification algorithm for the task of identifying anuran species through audio recordings. The main objective of the study is to termine accuracy, precision, and recall values for five vectorization algorithms combined with six classification algorithms. To this end, these algorithms were implemented in Python and iterated over every possible combination, with evaluation metrics calculated. The results show that short-time Fourier transform and mel-frequency cepstrum yield excellent results when combined with multilayer perceptron and random forest. It is concluded that there are several promising methods for identifying anurans through sound, and that these methods can be explored and selected according to the constraints and requirements of the research in which they are to be applied.