Perceptual Considerations in Audio Morphing

Research Student: Dr Duncan Williams
Principal Supervisor: Dr Tim Brookes

Start date: 2004
End date: 2010

Project Outline

A prototype timbre morpher has been designed, developed, and evaluated. The prototype morpher is capable of adjusting the timbre of a source sound toward that of a target sound specifically in terms of its brightness, softness, and/or warmth, independently from other perceptual attributes, and provides a novel tool for sound design and composition.

A literature review revealed that several different paradigms of timbre exist. Consideration of the alternatives led to the conclusion that a multidimensional paradigm of timbre, comprising attributes with both meaningful descriptors and specific acoustic correlates was the most suitable for the development of the timbre morpher. In order to develop the prototype morpher, a range of attributes with spectral, temporal, independent, and overlapping acoustic correlations was required. The most suitable candidate attributes were brightness, softness, and warmth.

Various computer implementations of cross-synthesis, audio morphing, and sound hybridisation were investigated to inform the selection of a suitable platform for the prototype timbre morpher. Spectral Modelling Synthesis was found to be the most suitable alternative, partly as a result of its fast processing speed.

Three iterations of the prototype morpher were developed in MATLAB, creating a complete system capable of selectively adjusting the acoustic correlates of each of the chosen timbral attributes.

A review of literature relating to perceptual evaluation methodologies established that a combination of Multidimensional Scaling and Verbal Protocol Analysis was the most suitable approach with which to evaluate the prototype morpher. Through a series of listener evaluations involving stimuli created by the morpher, this approach, combined with a third quality control experimental stage, demonstrated that brightness, softness, and warmth can each be morphed independently of each other.

Using the procedures devised in this thesis, a complete morpher, controlling the full range of possible timbral attributes, can now be developed, coded and evaluated.