Digital waveguide mesh modeling of the vocal tract acoustics

Abstract

The Digital Waveguide Mesh is a technique used in the modelling of room acoustics and musical instruments. This paper details a project that applies the theory of waveguide mesh acoustic modelling to the production of human-like vowel sounds. A 2D software mesh model is created that approximates the shape of the vocal tract in different vowel positions, and a glottal flow input is applied. The resulting signal bears similar resonant frequencies or formants to that of recorded speech. Recommendations are made towards extending the model to include some of the more complex features of the mouth, potentially constructing an acoustical model of the human vocal tract capable of creating speech sounds of increased naturalness.

Waveguide physical modeling of vocal tract acoustics: flexible formant bandwidth control from increased model dimensionality

“…The impedance discontinuity between two sections of tube arises from the change in cross-sectional area between two points, and acts to cause scattering of pressure signals through both the transmission and reflection of some amount of the propagating wave signal across the section discontinuity. The pressure and velocity components in each section can be directly related to the corresponding impedance and cross-sectional area using air density and the wavespeed (10) The KL scattering equation is a special case of the more general -connection waveguide junction (3) where . It describes, in discrete form, the signal flow out of a junction as …”

“…Physical models of the vocal tract [7]- [9], have shown that sounds approaching natural speech can be synthesized using a chain of one-dimensional (1-D) waveguides representing the length of the tract from the glottis to the lips. A two-dimensional (2-D) model extending the 1-D case by incorporating variable tract width along the length of the model has been introduced [10], and the software used to analyze the model has been discussed [11]. The model is examined here in greater detail, with particular reference to the control available over formant bandwidths of synthesized vowels.…”

Real-Time Dynamic Articulations in the 2-D Waveguide Mesh Vocal Tract Model

“…Previous work has shown that the 2-D mesh vocal tract model based on a widthwise area function implementation offers speech sound generation on a non real-time basis [11], [12]. Parameters are set within the model and the system is left to generate the output.…”

“…More recent research has been focused on improving the underlying propagational model to give better simulation of the resonant cavity [9], [10]. Research into higher dimensionality of the wave propagation model has shown that equivalent formant patterns to the existing 1-D models can be generated using a two-dimensional (2-D) waveguide mesh [11]. It has also been detailed that formants produced by the mesh model follow an approximately linear bandwidth variation in response to changes in boundary reflection parameters [12].…”

Evolution of adult male oral tract shapes for close and open vowels

“…It is based on computing techniques that are inspired by the principles of evolution, and therefore offers a direct alternative to placing subjects in an fMRI machine or LPC analysis. Oral tract shapes are evolved and tested using the two dimensional physical modeling synthesis techniques of Mullen et al [11,12], who have demonstrated that a digital waveguide mesh (DWM) provides a highly successful method for vowel synthesis. Currently, only oral sounds can be synthesized because there is no nasal tract (nose), but this is all that is required for the synthesis of isolated vowels.…”