Simulation of disordered speech using a frequency-domain vocal tract model

In this paper, we address the issue of how the perception of disorderness in selected types of speech disorders may be correlated with the abnormal articulatory structure and with the related acoustic properties. As a first step towards this end we have developed an articulatory synthesizer based on frequency-domain simulation of vocal-tract wave propagation. The synthesizer has been implemented by three numerical methods — Runge-Kutta, ABCD matrix, and Finite difference, which provide frequency-domain solutions to the transmission-line equation characterizing a lossy vocal tract. The synthesizer is applied in preliminary experiments where the synthesizer’s outputs are used to match samples from a corpus of steady-state speech sound, obtained from a dysarthric speaker, uttered in the /hV/ context.