Psychoacoustic model for calculating the sound quality of piano tones

Abstract

Assuming that only audible dissimilarities between different piano tones may lead to different quality judgments, experiments were conducted to determine the perceptual attributes listeners use to judge these dissimilarities. The psychoacoustic results showed that two major attributes contribute with over 90% to the explanation of the perceived dissimilarities: the psychoacoustic ‘‘sharpness’’ and a second attribute that was described with the scale ‘‘open–closed.’’ Both attributes depend on the specific loudness distribution of the sound signals and are defined as a weighted first moment of the critical-band rate distribution of specific loudness using different weighting functions. The weighting function used to calculate the second attribute emphasizes spectral components around 1 kHz. The model developed for calculating audible dissimilarities between piano tones showed a good agreement between psychoacoustically measured and calculated dissimilarities; the correlation was 0.90. The influence of the two attributes on sound quality judgments was verified by listening tests with appropriately modified piano sounds. The results showed that calculating the sound quality of musical instruments on the basis of their sound signals requires an aurally adequate signal processing. The model developed to calculate the sound quality of piano tones showed a good agreement between measured and calculated quality orders.