EXPERIMENTAL VALIDATION OF A URBAN TRAFFIC NOISE ANNOYANCE MODEL

Résumé En

This study deals with noise annoyance in urban environments. It is based on a previous work that resulted in two main outputs. First, a measurement tool that identifies and extracts sonic properties of the main urban noises sources, i.e. the different types of vehicles. Second, a modeling tool that estimates the perceived annoyance level by adapting from the literature, and implementing, a current multi-class psychoacoustic model. The present work follows these outcomes and aims at validating the annoyance estimations given by the model with an experimental approach. A listening test is designed in order to be able to collect perceived annoyance in a virtual environment (in lab test). Sound scenes made of urban soundscapes (background) and vehicles passing-by (foreground) are built and encoded in a 2D-audio immersive format (2D to multichannel algorithm), in order to create a listening experience as ecological as possible. Within this frame, an important experimental protocol is designed in order to measure the perceived annoyance caused by each synthesized sound scene and to control the cognitive load (attention on a list of words to play back afterwards) of the participants. The first rounds of data analysis show consistent results, especially in terms of influence of sound level or source typicality on the measured annoyance. A deeper round of analysis is currently being processed in order to examine the fine correlation between measured and predicted annoyance values, with regards to psychoacoustic features of source signals. All the results will be presented and discussed in the course of the conference.