Between location and place: A view of timbre through auditory models and sonopoietic space

The following work addresses problems of definitional vagueness surrounding timbre as well as reviewing previous research done in quantifying qualitative aspects of sound (e.g. speech pathology, psychology, psychophysics, voice conversion). In particular, the relationship between phase and timbre (monaural phase perception) is studied. Whereas previous research in monaural phase perception has only used synthetic stimuli, this study proposes a method of controlling the phase of instrumental tones. An auditory model of timbre is presented that demonstrates monaural phase effects similar to those exhibited by listeners. It is argued that a spectro-temporal auditory model, as opposed to a spectral (place code) model, must be used as the basis for a complete timbre model.;The concept of sonopoietic spaces is introduced in contrast to timbre spaces. Timbre spaces (e.g., produced by multidimensional scaling studies) are premised on generality. This generality comes from attempting to represent many listeners' pair-wise judgements of the similarity between musical tones. Sonopoietic space on the other hand is the space of listening constructed by a listener's interaction with a sonic event (e.g. natural sounds or musical compositions). Sonopoietic images articulate this space and are formed by a listener's attempts to construct a coherent interpretation of the sonic world. A sonopoietic image includes (i) listeners' understanding of a sound's underlying dynamics, (ii) how this image relates to other sonopoietic images (including both categorical and thematic relations) and (iii) a listener's sonic associations (e.g., contiguous events from past listening experiences).;McAulay and Quatieri's (1986) sinusoidal system provides a method for stimulus generation in which the phases, amplitudes, and frequencies of the spectral components can be individually controlled. Two types of stimuli were generated from instrumental tones---one with random quadratic phase (RQP) and another with measured cubic phase (MCP). The first of two pilot studies indicated that, as the fundamental frequency of the stimuli increases, the listener's ability to discriminate RQP from MCP decreases. The second pilot study demonstrated that a spectro-temporal auditory model can be used to predict monaural phase effects. As the fundamental frequency increased, the distance between the RQP and the MCP decreased.