Spectral noise levels and roughness severity ratings for vowels produced by male children.

Abstract

This study investigated the relationship between spectral noise level (SNL) and perceived roughness in sustained vowels phonated by children. Twenty male children, ten presenting normal vocal quality and ten presenting vocal roughness associated with benign vocal cord lesions, served as subjects. The children individually sustained each of the vowels (Formula: see text) at one intensity and at two pitch levels (comfortable and high). Each vowel recording was analyzed to produce a 3-Hz bandwidth, amplitude-by-frequency acoustic spectrum. The levels of inharmonic (noise) components were measured for the range 100 to 2600 Hz within each vowel spectrum and the mean of those measures provided an index of vowel spectral noise level. Each recorded sample was also rated for degree of roughness, by 11 trained judges, using a five-point, equal-appearing intervals scale. The fundamental vocal frequency (fo) of each production also was measured. The results indicated that four subjects with vocal pathology phonated their vowel productions with a relatively extreme fo when compared to subjects with normal voice quality. It was also found that vowel roughness and SNL differed predictably for individual vowels and that, especially for normal-speaking subjects, roughness and SNL decreased with an increase in fo. Further, for both subject groups, SNL measures were correlated positively and moderately highly with perceived roughness.