Spatiotemporal classification of vocal fold dynamics by a multimass model comprising time-dependent parameters

Area of Science:

• Biomechanics of speech
• Vocal fold dynamics
• Medical imaging and signal processing

Background:

• Objective assessment of vocal fold function is crucial for diagnosing voice disorders.
• Nonstationary phonation presents challenges for traditional voice analysis methods.
• Left-right asymmetries in vocal fold vibration can indicate pathological conditions.

Purpose of the Study:

• To develop and validate a model-based approach for objectively measuring and classifying vocal fold vibration asymmetries.
• To address the challenges posed by nonstationary phonation in voice analysis.
• To provide an intuitive visual assessment and objective quantification of vocal fold instabilities.

Main Methods:

• Real-time high-speed digital videography of vocal fold dynamics during sustained vowels and pitch raises.
• Application of a time-dependent multimass model matched to extracted vocal fold vibrations (dorsal, medial, ventral).
• Block-based optimization using wavelet coefficients to compare vocal fold and model dynamics, accounting for nonstationary vibrations.

Main Results:

• The model-based approach was verified using synthetic data.
• The method was successfully applied to 40 clinical high-speed recordings of normal and pathological voice subjects.
• Generated model parameters enable visual assessment of vocal fold instabilities in an asymmetry diagram.

Conclusions:

• The proposed model-based approach offers an objective method for quantifying vocal fold asymmetries, particularly in nonstationary phonation.
• The asymmetry diagram provides an intuitive tool for assessing vocal fold instabilities.
• This technique has potential applications in the clinical diagnosis and management of voice disorders.