Modeling the glottal volume-velocity waveform for three voice types

Area of Science:

• * Speech Science
• * Acoustic Phonetics
• * Voice Production

Background:

• * Understanding the glottal volume-velocity waveform is crucial for characterizing voice production.
• * Previous models often lack detailed parameterization for different voice types.

Purpose of the Study:

• * To model glottal volume-velocity waveform features for modal, vocal fry, and breathy voices.
• * To identify key LF model parameters differentiating these voice types.
• * To develop a method for synthesizing these voice types using the glottal waveform model.

Main Methods:

• * Glottal inverse filtering to estimate the glottal volume-velocity waveform.
• * Fitting the estimated waveform to an LF (Liljencrants-Fant) model.
• * Statistical analysis including ANOVA and linear regression on four LF model parameters.

Main Results:

• * Three of the four LF model parameters showed significant differences across the three voice types.
• * Linear regression identified significant LF model parameters correlating with perceived voice quality.
• * A synthesis rule was developed based on the LF glottal waveform model.

Conclusions:

• * The LF model effectively captures distinct features of modal, vocal fry, and breathy voice waveforms.
• * Specific LF model parameters are key acoustic correlates of voice type.
• * The developed model and synthesis rule can accurately reproduce different voice qualities.