Numerical study of turbulent flow-induced sound production in presence of a tooth-shaped obstacle: towards sibilant [s] physical modeling

Julien Cisonni, Kazunori Nozaki, Annemie Van Hirtum, Shigeo WadaInterspeech 2010

TL;DR

This paper uses computational fluid dynamics (CFD) to simulate airflow through a simplified vocal tract with a tooth-like obstacle, investigating how the constriction's size affects the resulting sound. The findings suggest that narrower constrictions produce louder sounds, offering a step towards physically modeling the [s] speech sound.

What Problem Does This Paper Address?

Producing speech sounds like the sibilant [s] involves complex interactions between airflow within the vocal tract and the teeth. Specifically, a turbulent jet of air impacts the incisors, generating the characteristic hissing sound. Accurately modeling this phenomenon requires understanding how the geometry of the vocal tract and teeth influences the airflow and the resulting acoustic sources. Previous studies have explored this, but a clear, systematic analysis of how specific geometric parameters, like the aperture of the constriction formed by the teeth, affect the generated sound has been lacking. This paper aims to fill that gap by numerically studying the flow-induced noise produced by a simplified tooth-shaped obstacle in a channel, focusing on the impact of the constriction's height.

Methods

The researchers employed a Large-Eddy Simulation (LES) method to model the incompressible and isothermal airflow. LES is a powerful CFD technique that resolves large turbulent eddies while modeling the smaller, sub-grid scale ones. This approach is suitable for capturing the turbulent nature of the airflow relevant to speech production.

The study used a simplified geometric model consisting of a rectangular channel with a tooth-shaped obstacle placed within it. The key variable manipulated was the constriction height (Hc), which represents the gap between the obstacle and the vocal tract wall. Simulations were performed for various constriction heights, expressed as percentages of the main channel height.

To model the sound generated, the researchers used Curle's formulation of Lighthill's acoustic analogy. This method relates the sound produced to the distribution of acoustic sources (like turbulence and surface pressure fluctuations) on the obstacle. A virtual microphone was placed downstream to capture the simulated sound.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Constriction Aperture is Key: The study strongly suggests a direct link between the size of the constriction formed by the "tooth" and the intensity of the generated sound. Narrower openings lead to louder noise.
  • Vorticity as a Sound Source Indicator: Higher vorticity (a measure of fluid rotation) was observed near the obstacle, particularly in the upstream part of its tip. The magnitude of this vorticity increased as the constriction narrowed, correlating with increased sound intensity.
  • Flow Detachment and Jet Formation: The obstacle causes the airflow to accelerate and detach from its surface, forming a turbulent jet. The degree of this detachment and the jet's behavior are influenced by the constriction size. Narrower constrictions lead to stronger detachment and a wider jet relative to the constriction.
  • Simplified Model Limitations: The authors acknowledge that their simplified model has limitations. The simulated turbulence is lower than in real speech, and the constrictions are wider than those found in the human vocal tract.

Key Results

  • The mean sound pressure level of the flow-induced noise increases as the constriction aperture decreases.
  • For the narrowest aperture (Hc = H × 4%), the generated noise level was significantly higher than for wider apertures, although the spectral shape differed.
  • Vorticity magnitude was found to be substantially higher for narrower constrictions. For Hc = H × 4%, the maximum vorticity was approximately 22 times greater than for Hc = H × 30%.
  • Velocity profiles at the constriction exit showed increased asymmetry and a greater distance from the wall for narrower constrictions, indicating stronger flow detachment.

Limitations and Future Work

The authors explicitly acknowledge several limitations:

  • Turbulence Level: The simulated flow exhibited lower turbulence compared to real speech production. This is attributed to the simplified geometry and potentially the mesh resolution.
  • Constriction Width: The constrictions studied were wider than those typically found in the human vocal tract during sibilant production.
  • Acoustic Analogy: The use of Lighthill's analogy is a simplification. It cannot account for the influence of vocal tract resonance on the radiated sound.

Future work suggested includes:

  • Finer Meshing: To better capture turbulence and simulate smaller constrictions, a much finer mesh will be required.
  • Computational Aeroacoustics: Employing more advanced computational aeroacoustics methods would provide a more accurate modeling of the generated noise, incorporating resonance effects.
  • More Realistic Geometries: Incorporating more detailed and realistic vocal tract and tooth geometries is crucial for a complete physical model.

Summary updated

BibTeX
@inproceedings{cisonni2010numerical,
  author = {Julien Cisonni and Kazunori Nozaki and Annemie Van Hirtum and Shigeo Wada},
  title = {{Numerical study of turbulent flow-induced sound production in presence of a tooth-shaped obstacle: towards sibilant [s] physical modeling}},
  booktitle = {Interspeech},
  year = {2010},
}
Figure