The Development of a Parametric Real-Time Voice Source Model for use with Vocal Tract Modelling Synthesis on Portable Devices

This research is concerned with the natural synthesis of the human voice, in particular, the expansion of the LF-model voice source synthesis method. The LF- model is a mathematical representation of the acoustic waveform produced by the vocal folds in the human speech production system. Whilst being used in many voice synthesis applications since its inception in the 1970s, the parametric capabilities of this model have remained mostly unexploited in terms of real-time manipulation. With recent advances in dynamic acoustic modelling of the human vocal tract using the two-dimensional digital waveguide mesh (2D DWM), a logical step is to include a real-time parametric voice source model rather than the static LF-waveform archetype.
This thesis documents the development of a parameterised LF-model to be used in conjunction with an iOS-based 2D DWM vocal tract synthesiser, designed with the further study of voice synthesis naturalness as well as improvements to assistive technology in mind.