Describing and classifying urban sound environments with a relevant set of physical indicators.

Abstract

Categorization is a powerful method for describing urban sound environments. However, it has only been applied, until now, to discrete noise data collection, whereas sound environments vary continuously both in space and time. Therefore, a procedure is developed in this paper for describing the variations of urban sound environments. The procedure consists of mobile measurements, followed by a statistical clustering analysis that selects relevant noise indicators and classifies sound environments. Analysis are based on a 3 days + 1 night survey where geo-referenced noise measurements were collected over 19 1-h soundwalk periods in a district of Marseille, France. The clustering analysis showed that a limited subset of indicators is sufficient to discriminate sound environments. The three indicators that emerged from the clustering, namely, the Leq, A, the standard deviation σL eq, A, and the sound gravity spectrum SGC[50 Hz-10 kHz], are consistent with previous studies on sound environment classification. Moreover, the procedure proposed enables the description of the sound environment, which is classified into homogenous sound environment classes by means of the selected indicators. Thus, the procedure can be adapted to any urban environment, and can, for instance, favorably enhance perceptive studies by delimiting precisely the spatial extent of each typical sound environment.