Environmental sound sources classification using neural networks

Abstract

Noise pollution is the greatest single environmental issue faced by many urban centres in the world. Current techniques used for monitoring sound do neither provide adequate information for designers and planners, nor determine many of the sound parameters that influence perception. The overall aim of this research is to provide new strategies for acoustic monitoring of complex urban environments. The specific aim of this research is to determine features of sound from commonly existing sources to enable automated source recognition. This paper reports the use of Fast Fourier Transforms in order to produce spectral data of sounds from different sources for the classification using neural networks.