Journal of Phonetics (1986) 14, 495-499
Session 6
Vocal fold disorder evaluation by digital speech
analysis
Germana Banci
Department of Internal Medicine , Medical Physics Lab. , II University of Rome, 00173 Rome,
Italy
Simonetta Monini
I Clinic O.R.L., Policlinic Umberto I, 00161 Rome, Italy
Alessandro Falaschi and Nicolo De Sario
Ugo Bordoni Foundation, 00144 Rome, Italy
By means of digital processing of vocal signals we have automatically
extracted some quantitative meaningful features which are related to
vibratory pattern perturbations of the vocal folds, when affected by
pathological processes. Here we give the methods and techniques of
vocal signal analysis and the procedure for the evaluation, in a
multidimensional space, of the parameters extracted, to estimate their
capability of discrimination between the normal and pathologic
classes.
1. Introduction
Since the appearance of the acoustic theory of speech production (Fant, 1960), the voice
acoustic analysis for diagnostic purposes has appeared to be a promising aid for the
traditional methods for the diagnosis of laryngeal pathologies, based mainly on visual
inspection of the larynx and on auditory and perceptual evaluation of the voice of the
patient. Later, by the application of new techniques of digital speech analysis, some
methods (still at experimental level) were developed, aimed at the automatic detection
of laryngeal pathologies through the extraction of several quantitative features. These
features represent some characteristic aspects present in pathological voices due to vocal
fold diseases; for instance, the alteration of the spectral structure, abnormal noise, pitch
perturbations and periodicity loss (Hecker & Kreul, 1971 ; Davis, 1976; Koike, Taka-
hashi & Calcaterra, 1977; Horii, 1980; Kitajima, 1981; Gubrynowicz, Kacprowski,
Mikiel & Zarnecki, 1981 ; Laver, Hiller & Hanson, 1982; Kim, Kakita & Hirano, 1982).
In this work, which is the continuation and integration of a previous study, after
having recalled the techniques and methods for the extraction of some physiologically
meaningful parameters related to vocal fold disfunctions and alterations, we propose a
statistical method for the evaluation of the feature "effectiveness" in a multidimensional
0095-4470/ 86/030495 + 05 $03.00/0 © 1986 Academic Press Inc. (London) Ltd.