Modeling and imaging of the vocal fold vibration for voice health.

Alba Granados

Research output: Book/ReportPh.D. thesisResearch

358 Downloads (Pure)

Abstract

Identication of abnormalities on the vocal fold by means of dierent diagnostic methods is a key step to determine the cause or causes of a voice disorder, and subsequently give an adequate treatment. To this end, clinical investigations benet from accurate mathematical models for prediction, analysis and inference. This thesis deals with biomechanical models of the vocal fold, specially of the collision, and laryngeal videoendoscopic analysis procedures suitable for the inference of the underlying vocal fold characteristics. The rst part of this research is devoted to frictionless contact modeling during asymmetric vocal fold vibration. The prediction problem is numerically addressed with a self-sustained three-dimensional nite element model of the vocal fold with position-based contact constraints. A novel contact detection mechanism is shown to successfully detect collision in asymmetric oscillations. Optimization procedures for constraint enforcement are studied to improve the accuracy of the predictions as an alternative to classical spring-like contact forces. The second part of this research investigates a non-invasive procedure to quantitatively analyze the two-dimensional vocal fold displacements captured with laryngeal high-speed videoendoscopy. A dense optical ow algorithm is
adapted to the complex nature of the image sequence, and numerical errors are treated to improve the accuracy of the results. Principal components decomposition is applied to extract the underlying modes of vibration, showing dierent characteristics in normal and abnormal phonation. In the last part of this thesis research, the optical ow algorithm for data acquisition as well as the biomechanical model of the vocal fold are used to formulate a nonstationary statistical inverse problem for vocal fold features estimation that accounts for the model uncertainty. An expectation-maximization algorithm for missing data is proposed to nd estimates of the system's unknowns. Due to time limitations no computational results are shown and a purely theoretical discussion is presented.
Original languageEnglish
PublisherTechnical University of Denmark, Department of Electrical Engineering
Number of pages146
Publication statusPublished - 2016

Projects

Cite this

Granados, A. (2016). Modeling and imaging of the vocal fold vibration for voice health. Technical University of Denmark, Department of Electrical Engineering.