Classificação de Sinais de Sensores Acústicos Seletivos de Drones

Raoni Alcântara, Rosângela Coelho

DOI: 10.14209/sbrt.2024.1571031117

Evento: XLII Simpósio Brasileiro de Telecomunicações e Processamento de Sinais (SBrT2024)

Keywords: Acoustic source classification surrogates adaptive learning drone

Abstract

This article presents a study on the classification of acoustic source signals captured by drones. The main contributions are the use of the ALSS technique and the alpha-GMM classifier in this scenario. Experiments were conducted with six environmental sources and two ego-noises with different degrees of stationarity. The results showed that both techniques separately provided an increase in classification accuracy and that the best results with alpha-GMM occurred for alpha=-2. Furthermore, it was demonstrated that the joint use of alpha-GMM and ALSS surpassed the accuracy of the techniques individually.

Download