Developing a timbrometer: perceptually-motivated audio signal metering

摘要

Early experiments suggest that a universally agreed upon timbral lexicon is not possible, and nor would such a tool be intrinsically useful to musicians, composers, or audio engineers. Therefore the goal of this work is to develop perceptually-calibrated metering tools, with a similar interface and usability to that of existing loudness meters, by making use of a linear regression model to match large numbers of acoustic features to listener reported timbral descriptors. This paper presents work towards a proof-of-concept combination of acoustic measurement and human listening tests in order to explore connections between 135 acoustic features and 3 timbral descriptors, brightness, warmth, and roughness.