Aerodynamically-based parametric description of the noise envelope in voiced fricatives

Abstract

In voiced fricatives, the radiated sound is composed of a harmonic component associated with the vibrating larynx and a noise component generated at a constriction in the oral cavity. The sound from the two sources interacts in a nonlinear way to produce a noise signal with an amplitude envelope modulated at the fundamental frequency of voicing. While voiced fricatives synthesized as a linear combination of harmonic and noise components are identifiable, it is recognized that the inclusion of the modulation improves the perceived naturalness of a synthesized token. The depth of modulation of the radiated noise, for a range of aerodynamic and acoustic variables, was measured experimentally using a dynamic mechanical model of the larynx and vocal tract. Glottal excitation arose from driven shutters representing the vocal folds; frication noise was produced by an orifice plate with a sharp-edged obstacle downstream. Based on the empirical data, a parametric description was developed to predict the depth and phase of amplitude modulation of the noise from the aerodynamic and acoustic conditions.