Probabilistic modeling framework for multisource sound mapping

Abstract

The process of modeling noise maps is now well defined: long-term aggregated indicators are calculated based on a collection or estimation of road, air and rail traffic variables. This framework however disregards the sound levels variations, and hence prevents the production of statistical or emergence indicators, and does not allow for the study of competition between typical urban sound sources that can improve the characterization of urban sound environments. A modeling framework in four steps is proposed to answer these issues: (i) a spatial distribution of the potential sound source of interest, (ii) the calculation of a sound propagation matrix, (iii) the stochastic activation of a sound sources ratio for n iterations of the sound map, and (iv) the calculation of specific sound indicators. The stochastic approach proposed in this study enables the estimation of the temporal sound distribution per sound source. It permits in particular to deduce source-oriented indicators such as the percentage of the time when a given sound source emerges from an urban sound mixture. An example of application of this framework is exposed for a district in the city of Nantes, France. It shows the interest of such approaches, in particular for soundscape and urban sound environment studies.