Inverse estimation of the vocal tract shape based on a vocal tract mapping interface

Area of Science:

• Acoustic phonetics
• Speech science
• Biomedical engineering

Background:

• Vocal tract shape estimation is crucial for understanding speech production.
• Previous research developed an interface for generating vocal tract shapes based on mouse interaction.
• Interpolation methods were used to derive shapes for arbitrary positions.

Purpose of the Study:

• To apply a vocal tract mapping interface for inverse estimation of vocal tract shapes.
• To validate the interface's accuracy using formant frequency data.
• To compare estimated shapes with existing imaging and video data.

Main Methods:

• Utilized a vocal tract mapping interface for inverse estimation.
• Employed a coarse-to-fine algorithm to search for target formant frequencies.
• Interpolated vocal tract shapes based on mouse-pointer positions.

Main Results:

• Estimated vocal tract shapes closely matched magnetic resonance imaging (MRI) data.
• Estimated shapes also correlated well with lip area data from video recordings.
• Individual subjects exhibited unique trajectories on the interface window.

Conclusions:

• The vocal tract mapping interface is effective for inverse estimation of vocal tract shapes.
• The method shows potential for applications in speech synthesis and analysis.
• Individual variations in speech production can be observed through interface trajectories.