FöRFATTARE
Nykänen, Arne

INSTITUTION
Arbetsvetenskap / Ljud & Vibrationer

SAMMANFATTNING
Product sound design has received much attention in recent years. This has created a need to develop and validate tools for developing product sound specifications. Elicitation of verbal attributes, identification of salient perceptual dimensions, modelling of perceptual dimensions as functions of psychoacoustic metrics and reliable auralisations are tools described in this thesis. Psychoacoustic metrics like loudness, sharpness and roughness, and combinations of such metrics into more sophisticated models like annoyance, pleasantness and powerfulness are commonly used for analysis and prediction of product sound quality. However, problems arise when sounds from several sources are analysed. The reason for this complication is assumed to be the human ability to separate sounds from different sources and consciously or unconsciously focus on some of them. The objective of this thesis was to develop and validate methods for product sound design applicable for sounds composed of several sources.

The thesis is based on five papers. First, two case studies where psychoacoustic models were used to specify sound quality of saxophones and power windows in motor cars. Similar procedures were applied in these two studies which consisted of elicitation of verbal attributes, identification of most salient perceptual dimensions and modelling of perceptual dimensions as functions of psychoacoustic metrics. In the saxophone experiment, psychoacoustic models for prediction of prominent perceptual qualities were developed and validated. The power window experiment showed that subjects may judge only parts of the sound. Power window sound consists of the motor sound and the scratching of a window sliding over the seal. The motor sound was filtered out and models developed using motor sound alone showed good agreement with listening tests. This demonstrated the human ability to separate sound from different sources and pointed out the importance of handling auditory stream segregation in the product sound design process.

In Paper III sound sketches (simple auralisations) was evaluated as a way to assess sounds composed of several sources. Auralisation allows control of the contributions of different sources to a sound at the listening position. This way, psychoacoustic analysis and listening tests may be carried out on the contributions from sources separately and as an ensemble. Sound sketches may also serve to specify a target sound for a product.

In Papers IV and V, the precision of auralisations related to intended use was investigated. Auralisations were made by filtering engine sounds through binaural transfer functions from source locations to the listening position in a truck cabin. In Paper IV simplifications of auralisations of one source were compared to artificial head recordings. For idling sounds auralisations through binaural transfer functions with a resolution of 4 Hz or better, or smoothed with maximum 1/96 octave moving average filters were found to preserve perceived similarity to artificial head recordings. In Paper V the effect of simplifications of transfer functions on preference ratings of auralisations was examined. This is of interest in applications where audible differences may be acceptable as long as preference ratings are unaltered, e.g. when auralisations are used as rough sound sketches. At 500 rpm idle speed, a resolution of 32 Hz or better, or smoothing with maximum 1/24 octave moving average filters showed no significant alteration of subject preference ratings. These figures may serve as guide for required accuracy in auralisations used for evaluation of idling sounds in truck cabins.

To conclude, psychoacoustic analysis of total sound may be used for prediction of perceived sound quality as long as the sound is generated by one source. When several sources generate sound, auditory stream segregation effects in combination with cognitive effects may deteriorate the results. Auralisation is a useful tool in such cases, since it makes it possible to analyse the effects of contributions from each source. It can also be used for making sound sketches which can serve as support in the design process.

ISSN 1402-1544 / ISRN LTU-DT--08/45--SE / NR 2008:45