Objective models of subjective annoyance remain limited to specific noise types

Evidence

The results from listening tests can be compared with psychoacoustic attributes computed using auditory models, such as loudness, fluctuation strength, sharpness, roughness, tonality, and impulsiveness. A common approach is to attempt to create a model from these attributes to estimate the annoyance and disturbance caused by noise, which should be in agreement with results from listening tests. If successful, such a model could be used as an objective measurement system of subjective annoyance (Marquis-Favre et al., 2005a). Unfortunately, these models are usually valid only with a limited set of noise signals, and separate models have to be constructed for different types of noise signals, and in some cases the models do not explain the measured results (Waye and Öhrström, 2002). — Ville Pulkki, Matti Karjalainen, Communication Acoustics- An…, 2490

Categorization

• QUESTION: The stated goal of objective measurement of subjective annoyance — the failure mode worth understanding: why don’t these models generalize?
• XREF: Directly relevant to reader’s application goal — annoyance models as an acoustic-communication application domain.