Deriving sound quality measures from a perceptual model

There is a growing need for objective measures that provide a reproducible and reliable characterization of sound quality in many practical areas of sound design engineering. In current practice and research, however, the commonly used measures are rather simple technical measures (e.g. weighted sound pressure levels) or psychoacoustic measures (e.g. loudness, sharpness, roughness). These measures need to be verified for every new class of signals, and usually a different and newly adapted combination of several metrics is required to predict more high-level percepts such as sound annoyance or preference. Another promising approach to develop more generally applicable models is to employ perceptual models that incorporate all of the basic mechanisms underlying human sound perception (hearing thresholds, limited temporal and spectral resolution, dynamic compression, etc.). In this study, the output of an existing perceptual model is used to derive measures for evaluating the perceived quality of different stimuli. This includes artificial stimuli from fundamental psychoacoustic experiments (e.g. roughness perception) as well as real product sounds from other sound quality studies. The results are compared to experimental data and existing quality measures in order to evaluate the potential of the psychophysical model as frontend for sound quality evaluation.