Date of Award
1981
Document Type
Dissertation - Restricted
Degree Name
Doctor of Philosophy (PhD)
Department
Electrical and Computer Engineering
First Advisor
Russell J. Niederjohn
Second Advisor
James A. Heinen
Third Advisor
Steven E. Reyer
Fourth Advisor
Arthur C. Moeller
Abstract
The problem of accurate extraction of two features of the speech signal affected by additive white noise is investigated. These feati.res are the fundamental frequency of vocal cord vibration and voiced/unvoiced categorization of speech. The extracted features are used in an adaptive pitch processing method for the enhancement of speech intelligibility in noise. A pattern recognition approach is used for categorization of noisy speech. A value of confidence in categorization is used as a primary parameter for the adaptive pitch processing method.
The pitch extraction technique, the classification technique and the enhancement method have been extensively tested. Toe results obtained are discussed and suggestions for further improvements and research directions are presented.