Methods for specification of sound quality applied to saxophone sound

Sammanfattning: Product specifications that include descriptors of sound qualities are most helpful when a description contains adequate detail and utilises understandable wording. Good specifications require that any descriptions used are familiar to users and that these descriptions are interpretable as acoustical quantities. The objectives of the studies reported upon in this thesis were to investigate how musicians use verbal descriptions of musical sounds and to interpret the descriptions in terms of some commonly used acoustical quantities. Interviews were made with saxophone players and the results were analysed with respect to how frequently different words were used. The most frequently used words were evaluated through listening tests using binaurally recorded test sounds of two saxophone players playing on two saxophones of different brands. Two groups of subjects participated in the listening tests, saxophone players and non-saxophone players. The subjects were asked to judge how well words selected from the interviews described the timbre of the test sounds. Data from the listening tests was analysed to examine how 3 factors: musician, saxophone, and type of listener influenced subject judgements. Principal Component Analysis (PCA) was used to find the most significant perceptual dimensions for the test sounds. Results from factorial analysis of variance and PCA were used to select the most appropriate verbal descriptions that best described significant perceptual dimensions. A small set of words (9 words) suitable for describing the variations in the analysed sounds was developed. To interpret the perceptual dimensions in terms of physically measurable indices, a number of acoustical quantities were added one by one to the set of perceptual variables. PCA of these extended data sets, resulted in a suggestion as to how the perceptual dimensions could be defined in terms of acoustical quantities. The procedure resulted in two significant perceptual dimensions being developed and described. A bipolar scale was used to describe Dimension 1 where items on the scale ranged from sharp/keen/rough to soft/warm/full-toned. Dimension 2 was described by a unipolar scale using the term E-like. The acoustical quantities sharpness, tonality, loudness, specific roughness (5-6 Bark and 21-22 Bark) and specific loudness (8-14 Bark) correlated with sharp/keen/rough. Roughness and specific roughness (13- 14 Bark and 20-21 Bark) correlated with soft/warm/full-toned. Specific loudness (10-11 Bark ) correlated negatively with E-like/large and specific loudness (9-10 Bark) and specific roughness (7-9 Bark) correlated positively with E-like/large. These findings demonstrate that it is possible to subjectively describe sound quality and to relate acoustical quantities to the descriptions. This suggests that it is practical to develop sound description terminology useful for work such as preparation of specifications and subsequent product standardisation.