Multiresolution analysis of environmental noise signals: an application to railway noise

The most part of acoustic signals associated to real world phenomena are non-stationary. A particularly important example is represented by signals associated to train transits, whose noise environmental impact is evaluated by considering the features associated to each of them. The traditional noise signal analysis techniques, based on Fourier transform and digital filtering, are generally not be able to adequately characterize such non-stationary signals. In this paper we'll discuss the use of multiresolution time-scale and time-frequency analysis, based on wavelet and Wigner-Ville transforms, to the study of noise signal produced by railway traffic, showing that the suitable employ of these techniques is able to properly characterize the trains feature and their environmental noise impact.