Selection of a sound propagation model for noise annoyance prediction: A perceptual approach

The LDEN is widely used to assess transportation noise annoyance, as recommended by the European Commission. However, other signal features (e.g. amplitude modulation) have proven to influence noise annoyance as well. It is not practical to determine the corresponding acoustical and psychoacoustical indices (e.g. roughness) when producing noise maps, since that would require numerous in situ recordings. An alternative might be to estimate the relevant index values knowing the LDEN value at a given receiver point M. To this end, a perceptually relevant sound propagation model needs to be selected. Therefore, ground transportation noises were simultaneously recorded at short (M1) and longer (M2) distances from the source. Three different propagation models were applied to the M1 recordings in order to simulate pass-by noises heard at M2. A simple level decrease based on geometrical divergence was also considered as a fourth model. After a physical comparison of the four models, a listening test was carried out to determine the best propagation model from a perceptual viewpoint. This test consisted of collecting dissimilarity ratings between simulated pass-by noises and a reference noise recorded in situ at M2.