Perceptually related analysis of time-frequency patterns via a hearing model (Sottek), a pattern-measurement algorithm (‘‘relative approach’’) and a window-deconvolution algorithm

Abstract

The presence of patterns (time structure and/or tonal structure) in sound situations elicits a different analytic process by the human hearing system than when sound situations are without pattern. In the presence of pattern, two issues can interfere with the effectiveness of conventional measurement methods: the human hearing is not acting as an absolute magnitude measurer so conventional measurements may not, or only poorly, resolve the important criteria; and pattern information that is subjectively highly significant is frequently of low magnitude amid general high-level structure that is not a direct part of the subjective pattern-response problem yet carries significance. Several automotive sound quality and information technology time-data recordings involving various patterns, of different strengths, were analyzed using conventional, psychoacoustic, and specialized pattern-sensitive techniques. In general, the advanced perception-related methods proved superior in quantifying ‘‘patterned’’ sound situations according to subjective impression.