Virtual Exploration of the Human Vocal Tract

Hypothesis: By simulating the acoustic field throughout the entire vocal tract the evolution of speech sounds within the tract can be directly and quantitatively related to physical variations in the tract geometry. This insight into speech production could then be applied to a variety of fields where the ability to alter or investigate speech characteristics in a targeted way could be useful for example in the teaching of speech science, in speech coaching, or as part of the planning of medical procedures. In this research, a bespoke acoustic simulation package has been produced using a continuous 3-dimensional Digital Waveguide Mesh (DWM) which can produce acoustic output throughout the entire simulation domain containing the tract at every time step. This package has been shown to reproduce formant frequencies for a variety of vocal tract shapes with an average mean absolute error of 10.12% at the lips, which is comparable to other research. These results have been investigated by comparing simulation output to recorded output from physical models. This simulation package has also been used to perform studies into the shifting of formant frequencies during speech sound production along the length of the tract, and into the effect on formant frequencies of the removal of geometric features of the tract such as the piriform fossae. These studies have been compared to physical internal measurements of vocal tract models from living subjects, showing preliminary agreement with further development required. A large emphasis has been placed on the accessibility of this research, with the production of several tools for visualisation of the data contained within, and with decisions made during the production of the simulation package itself.