Análise comparativa do desempenho de algoritmos para localização de sinais sonoros sujeitos a ruído AWGN e impulsivo

Abstract

In recent years, with increasing processing capacity as well as the miniaturization of electronic devices, a number of systems based on the location of audio signals have emerged, such as commercial sniper detection systems (ShotSpotter) and voice monitoring systems ( voice care), and residential, among them, entertainment system and home automation (eg Google Home and Amazondot). In the literature, the traditional methods of Direction-of-Arrival (DoA) are commonly evaluated in acoustic environments submitted to White Additive Gaussian Noise (AWGN). However, acoustic environments are also subject to impulsive noise, under which the classical methods of DoA estimation have degraded performance. In this way, besides describing and evaluating the performance of conventional parametric and non-parametric DoA methods, and demonstrating their in acoustic environments submitted to the AWGN, this work also proposes to present their performance when subjected to the impulsive noise modeled by the Gaussian Mixture Model (GMM).