NOIDESc: A novel traffic noise description scheme

Abstract

A novel framework for the acoustic and psychoacoustic description of noise signals (railway and road) has been proposed in order to extend and improve traditional noise classification schemes. In addition to measures of averaged sound-pressure level and noise exposure duration, the description scheme includes psychoacoustic parameters such as spectral centroid, spectral spread, modulation spectrum, and further dynamic aspects of the sound events corresponding in part to the MPEG-7/4 standard. The dynamic changes between foreground and background are addressed in the model. The approach assumes the evaluation of several low-level acoustic parameters of the noise, to be integrated into few high-level features by means of statistical methods, such as Gaussian mixture models and cluster analysis. As the usage of noise monitoring systems is expected to increase in the near future, the cumulative collection of calibrated sound recordings will provide comprehensive regional and supraregional sound databases to serve as input for automated noise classification. [The work in progress is performed in cooperation with the division of Acoustics of the TGM Vienna and is supported by the Austrian FFG.]