Inverse scattering for vowel articulation with frequency-domain data

TL;DR

This paper investigates whether different frequency-domain scattering data can uniquely determine the vocal tract's cross-sectional area during vowel articulation, analyzing conditions for uniqueness and additional information needed.

Contribution

It provides a mathematical analysis of inverse scattering problems in speech production, identifying data sets that ensure unique reconstruction of vocal tract geometry.

Findings

Pressure at microphone can determine vocal tract shape

Transfer function data may require additional info for uniqueness

Nonuniqueness cases are characterized and addressed

Abstract

An inverse scattering problem is analyzed for vowel articulation in the human vocal tract. When a unit amplitude, monochromatic, sinusoidal volume velocity is sent from the glottis towards the lips, various types of scattering data are used to examine whether the cross sectional area of the vocal tract can uniquely be determined by each data set. Among the data sets considered are the absolute value of the pressure measured at a microphone placed at some distance from the lips, the pressure at the lips, and the transfer function from the glottis to the lips. In case of nonuniqueness, it is indicated what additional information may be used for the unique determination.