Aerodynamics and Lumped-Masses Combined with Delay Lines for Modeling Vertical and Anterior-Posterior Phase Differences in Pathological Vocal Fold Vibration

Carlo Drioli, Philipp AichingerInterspeech 2019

TL;DR

This paper introduces a novel biomechanical model of vocal fold vibration that explicitly incorporates anterior-posterior (A-P) phase differences, a crucial but often overlooked aspect of vocal fold dynamics. By combining a simplified lumped-element model with delay lines, the authors demonstrate the ability to reproduce complex, pathological vocal fold vibration patterns observed in high-speed video data, offering a more nuanced tool for voice analysis and diagnosis.

What Problem Does This Paper Address?

Understanding the intricate mechanics of vocal fold vibration is fundamental to characterizing voice quality and diagnosing voice disorders. While existing biomechanical models can simulate basic vocal fold oscillations, they often struggle to accurately represent the subtle, yet significant, anterior-posterior (A-P) phase differences in vibration. These A-P phase differences are known to correlate with various voice qualities and pathological conditions, such as hoarseness, breathiness, and diplophonia. Traditional models, like the two-mass model, inherently cannot capture longitudinal phase differences, while more complex multi-mass models can be computationally expensive and difficult to control for specific phase relationships. This paper aims to bridge this gap by developing a model that is both computationally efficient and capable of directly controlling and reproducing these critical A-P phase differences, thereby improving the analysis and simulation of pathological vocal behaviors.

Methods

The authors propose a body-cover biomechanical model of the vocal folds that integrates lumped elements (mass, spring, damper) for simplified vertical dynamics with distributed elements (delay lines) to represent longitudinal propagation.

  • Lumped-Element Core: A simplified single-mass system represents the posterior inferior edge of each vocal fold, capturing basic vertical oscillations driven by glottal pressure. This forms the foundation of the model's dynamics.
  • Vertical Phase Delay: A delay line is incorporated to model the propagation of displacement along the vertical axis of the vocal fold cover, essential for self-sustained oscillations.
  • Anterior-Posterior (A-P) Phase Delay: Crucially, another delay line is introduced to explicitly model the longitudinal propagation of displacement along the A-P axis. This allows for direct control over the phase lag between the posterior and anterior edges of the vocal folds. The model allows for different A-P phase delays for the left and right vocal folds to account for asymmetries.
  • Flow Model: A simplified flow model converts the dynamic glottal area, derived from fold displacements, into airflow, using Bernoulli's law and a discretized representation of the glottis along the A-P axis.
  • Numerical Solution: The resulting system of differential equations is solved numerically in discrete time to estimate glottal flow and fold displacements.

The novelty lies in the explicit inclusion and direct control of the A-P phase delay via delay lines, a feature not easily achieved in prior lumped-element models.

Datasets

The paper utilizes a selection of five recordings from the Laryngeal High-Speed Video Database of Pathological and Non-Pathological Voices described in reference [30]. These recordings were chosen specifically because they exhibit various patterns with A-P phase differences. The authors use these recordings for qualitative validation of their model's ability to reproduce observed vibratory patterns, rather than for quantitative training or benchmark creation.

Key Insights and Ideas

  • A-P Phase Differences are Key: The research highlights that A-P phase differences are not just a minor detail but are intrinsically linked to observable pathological vocal behaviors like "double pulsing" and "zipper-like" motions.
  • Delay Lines as a Powerful Tool: The use of delay lines proves to be an effective and computationally efficient method for modeling wave propagation and introducing controlled phase shifts in biomechanical systems like vocal folds.
  • Bridging Simulation and Observation: The model's ability to qualitatively reproduce complex patterns seen in high-speed videoendoscopy suggests it can serve as a valuable tool for understanding the physical basis of voice disorders and for interpreting clinical observations.
  • Controllability is Crucial: The direct control over A-P phase delays offers a significant advantage over more complex models, allowing researchers to systematically investigate the impact of these differences on vocal fold dynamics and voice quality.

Key Results

The paper demonstrates the model's capability to qualitatively reproduce several observed oscillatory patterns from high-speed video data:

  • Double Pulsing: The model successfully simulated double pulsing with L-R symmetry, attributed to an A-P phase difference of approximately 180 degrees, mirroring observations in sample S1.
  • Zipper-like Motion: A zipper-like posterior-to-anterior opening pattern was reproduced by simulating symmetric A-P phase differences (around 108 degrees), as seen in sample S4.
  • Asymmetric A-P Phase Delay: The model could replicate asymmetric A-P phase delays, where only one vocal fold exhibits a significant phase shift, as observed in sample S5.
  • Complex Vibration Patterns: The model generated complex vibration patterns resulting from a combination of L-R mechanical differences and A-P phase differences, as seen in sample S2.

While quantitative metrics like accuracy are not the primary focus, the qualitative agreement between simulated and observed patterns is presented as a key success.

Limitations and Future Work

The authors acknowledge several limitations and suggest avenues for future research:

  • Closing Pattern Limitation: The current model can reproduce posterior-to-anterior opening patterns but struggles to accurately model zipper-like anterior-to-posterior closing motions. This is an area for further investigation.
  • Cycle Shape Control: While phase differences can be controlled, other aspects like cycle shape are not directly tunable.
  • Physical Realism: The model's A-P phase differences are primarily morphological and do not significantly contribute to the spontaneous dynamics of the model. Future work could incorporate more degrees of freedom in the mechanics and a more realistic flow description, including viscosity and acoustical coupling with the vocal tract.
  • Quantitative Validation: While qualitative validation is presented, future work could focus on quantitative validation and copy-synthesis to achieve precise period synchronization and amplitude matching.

Summary updated

BibTeX
@inproceedings{drioli2019aerodynamics,
  author = {Carlo Drioli and Philipp Aichinger},
  title = {{Aerodynamics and Lumped-Masses Combined with Delay Lines for Modeling Vertical and Anterior-Posterior Phase Differences in Pathological Vocal Fold Vibration}},
  booktitle = {Interspeech},
  year = {2019},
}
Figure