Artificial speech for intensive care unit (ICU) patients and laryngectomees

Type of content
Theses / Dissertations
Publisher's DOI/URI
Thesis discipline
Electrical Engineering
Degree name
Doctor of Philosophy
Publisher
University of Canterbury. Electrical and Computer Engineering
Journal Title
Journal ISSN
Volume Title
Language
Date
2005
Authors
Lim, Marilyn Adeline Mei
Abstract

A method and prototype device to provide artificial speech for intensive care unit (ICU) patients and laryngectomees is presented. The method assists these patients to produce natural sounding speech by "mouthing the words". A review of the current communication techniques for these patients is presented. The limitations of these techniques suggests that there is a need for a device that produces natural sounding speech (pitch variation and glottal sound source that resembles the actual glottal pulse generated by the vibrating vocal folds) and a device that is user friendly. As vocal folds only vibrate during vowel production, only vowel sounds are considered. Since pitch variation plays a major role in the naturalness of a person's voice, a number of alternative (automatic) pitch control techniques were explored. A unique pitch control technique utilising the changes in jaw height when a person "mouth the words" is presented. The electroglottographic (EGG) signal is used as the glottal sound source signal for this research as the properties of the EGG signal offers a number of advantages compared with other glottal sound source measurement techniques. A new glottal source model known as the twin-bar model, based on EGG measurements from normal volunteers, is also introduced. This model changes the shape of the glottal pulse based on a single parameter: pitch. Perceptual testing of the simulated voice using the twin-bar glottal model and two other well-known models on volunteers showed that the twin-bar model produces more natural sounding voice than the other two models. A new artificial speech system combining the automatic pitch control technique (jaw height) and the glottal sound source (twin-bar model) was constructed. It also includes a number of extra functions that would further improves the speech produced with this system. Existing technology on a laptop (e.g. serial port communication, bluetooth transceivers and USB port) is utilised for the construction of the prototype, with the laptop as the signal processing unit. The prototype was tested on a normal subject.

Description
Citation
Keywords
Ngā upoko tukutuku/Māori subject headings
ANZSRC fields of research
Rights
Copyright Marilyn Adeline Mei Lim