Analysis and simulation of a speaker's aerodynamic and acoustic patterns for fricatives

Covarying acoustic pattern features located near the boundary region between a frication noise segment and the adjacent vocoid (vocalic) segments seem likely to be important for the intelligibility of fricative consonants. Attempts to capture some of these naturally occurring effects through simulation with a computer-implemented composite model of speech production are described. The aim is to gain greater understanding of the production of fricatives, by trying to match quantitatively the detailed patterning for each of two speakers. An analysis-by-synthesis approach is adopted: parameters in the model are first given estimated values based on the analysis of some phonetically controlled [vowel-fricative-vowel] sequences produced by the speaker; in subsequent runs of the model the match can be improved. First results for one speaker's production of [asa] are described.