The influence of dynamic flow conditions on sound generation in fricative consonants.

Abstract

Studies of speech production have often remarked that frication noise in voiced fricatives is weaker than that produced in homorganic voiceless fricatives. This is traditionally attributed to the reduction in mean flow (and, hence, Reynolds Number) produced by the glottal constriction needed to induce voicing. However, research in fluid mechanics has demonstrated that turbulence may develop quite differently in steady and unsteady background flows. Strong flow gradients may inhibit the transition to turbulence, or even relaminarize an already-developed turbulent flow. This paper describes a series of mechanical modelling experiments designed to investigate the effect of dynamic flow conditions on sound generation in fricatives. A feedback-regulated Kolvenbach flow valve, controlled by computer, is used to create carefully-shaped dynamic modulations of the air flow through a stereolithographic replica of the vocal tract, about a series of static flow conditions. A multitaper spectral analysis technique is then used in conjunction with sweeptone measurements to separate out the contribution of source and filter components to the far-field sound spectra recorded with a microphone, and source properties are compared across flow conditions. The observed variations cannot be explained by differences in mean flow alone, suggesting that dynamic effects may be important.