Noise annoyance from a mixture of multiple single sources: rating and prediction

Abstract

In this paper, noise annoyance from a mixture of multiple single sources is studied with emphasis on subjective evaluation and objective prediction. From 10 subjects, annoyance values for all single and artificially combined noise samples are collected using the semantic differential method with a suitable verbal scale. We propose a novel method to determine the utility weights of a multivariate linear regression model by comparing the total annoyance αT of the combined noise sample to every single annoyance αi from its componential single sound sample. This method predicts αT on the premise of given αi. Our results demonstrate that the multivariate linear regression model and the calculated utility weights provide a good and conceptually simple framework to predict the total noise annoyance.