Date of Award

5-2013

Degree Type

Report

Degree Name

Master of Science (MS)

Department

Electrical and Computer Engineering

First Advisor

Dr. Jacob H. Gunther

Abstract

A novel speech synthesizer is being developed which needs a source waveform that represents the sound created by the vocal folds before it is shaped by the rest of the vocal cavity. Methods already exist for extracting this waveform, but this report explores a new method. The method involves finding a model for the vocal tract. A system identification technique is applied that uses a white noise audio source emitted into the oral cavity via a tube as the input. The effects of the tube are characterized and accounted for to allow for greater accuracy in the estimation of the true vocal tract properties. The vocal tract model is then used to extract the source waveform from a vocalized speech recording. Common properties of the source waveform will also be characterized and synthesized. These properties include the changes in harmonic content of the source based on vocal effort, and the natural aperiodic punctuations in pitch and amplitude of the source waveform. All of these properties, when properly synthesized, will help to create a more natural-sounding glottal source waveform.

Share

COinS