The aeroacoustic mechanisms in human voice production are complex coupled processes that are still not fully understood. In this article, a hybrid numerical approach to analyzing sound generation in human voice production is presented. First, the fluid flow problem is solved using a parallel finite-volume computational fluid dynamics (CFD) solver on a fine computational mesh covering the larynx. The CFD simulations are run for four geometrical configurations: both with and without false vocal folds, and with fixed convergent or convergent-divergent motion of the medial vocal fold surface. Then the aeroacoustic sources and propagation of sound waves are calculated using Lighthill's analogy or acoustic perturbation equations on a coarse mesh covering the larynx, vocal tract, and radiation region near the mouth. Aeroacoustic sound sources are investigated in the time and frequency domains to determine their precise origin and correlation with the flow field. The problem of acoustic wave propagation from the larynx and vocal tract into the free field is solved using the finite-element method. Two different vocal-tract shapes are considered and modeled according to MRI vocal-tract data of the vowels /i/ and /u/. The spectra of the radiated sound evaluated from acoustic simulations show good agreement with formant frequencies known from human subjects.

Acoustic data has long been harvested in fundamental voice investigations since it is easily obtained using a microphone. However, acoustic signals alone do not reveal much about the complex interplay between sound waves, structural surface waves, mechanical vibrations, and fluid flow involved in phonation. Available high speed imaging techniques have over the past ten years provided a wealth of information about the mechanical deformation of the superior surface of the larynx during phonation. Time-resolved images of the inner structure of the deformable soft tissues are not yet feasible because of low temporal resolution (MRI and ultrasound) and x-ray dose-related hazards (CT and standard xray). One possible approach to circumvent these challenges is to use mathematical models that reproduce observable behavior such as phonation frequency, closed quotient, onset pressure, jitter, shimmer, radiated sound pressure, and airflow. Mathematical models of phonation range in complexity from systems with relatively small degrees of freedom (multi-mass models) to models based on partial differential equations (PDEs) mostly solved by finite element (FE) methods resulting in millions of degrees-of-freedom. We will provide an overview about the current state of mathematical models for the human phonation process, since they have served as valuable tools for providing insight into the basic mechanisms of phonation and may eventually be of sufficient detail and accuracy to allow surgical planning, diagnostics, and rehabilitation evaluations on an individual basis. Furthermore, we will also critically discuss these models w.r.t. the used geometry, boundary conditions, material properties, their verification, and reproducibility.

No abstract

No abstract