A perceptual dissimilarities based nonlinear sound quality model for range hood noise.

Abstract

The application of sound quality in household appliances has gradually increased in recent years. In addition to modeling algorithms, appropriate acoustic metrics that characterize product sounds also play an important role in developing models. In this study, an artificial neural network based sound quality model for range hood noise was established with the combination of prior metric selection by multidimensional scaling (MDS) analysis of perceptual dissimilarities. First, sounds in different environments, speeds, and positions were recorded, and their annoyance was evaluated by grouped anchor semantic differential subjective jury testing. Then, the timbre space underlying dissimilarity judgments were analyzed by CLASCAL, an accurate MDS algorithm. Each dimension of the space was well explained by some metrics through stepwise regression. Finally, a sound quality model was established based on a back propagation neural network (BPNN). Results show that the combination of BPNN and CLASCAL can address the interpretation of the sound quality model and the ability to model nonlinearity for high accuracy. In addition, the application of noise control on range hoods showed that passive and active noise control (ANC) measures improve sound quality, especially ANC systems.