Browsing by Author "Blankenstein, Niko"
Now showing 1 - 1 of 1
- Results Per Page
- Sort Options
- Acoustic Centerline Extraction from MRI Data
Perustieteiden korkeakoulu | Bachelor's thesis(2024-02-02) Blankenstein, NikoThis work presents a method for extracting an approximation of an acoustic centerline from the MRI data of a human vocal tract during vowel pronunciation. The approximation is found with the use of Poisson’s equation coupled with Dijkstra’s algorithm. The MRI data set employed in this study comprises a singular male vocal tract pronouncing three different Finnish vowels: [a :], [i :], and [u :]. Extracting an acoustic centerline corresponding to the set poses a challenge because different frequencies of sound are associated with varying acoustic lengths of the vocal tract, influenced by differences in diffraction and refraction. This work aims to provide an approximate representation of the centerline expected to model the lowest Helmholtz resonance frequency of the vocal tract. The resulting centerline can be utilized to define an intersection area function for the vocal tract, applicable in computational acoustic analysis. The method presented in this study can be used to automatically extract the acoustic centerline of an MRI mesh. The finite element method is utilized to solve Poisson’s equation on the mesh by dividing it into finitely sized triangle elements, where the equation becomes more tractable for numerical solution. Dijkstra’s algorithm is applied to the solution for centerline extraction. The lowest Helmholtz resonance frequency of the mesh can be calculated by coupling this extracted centerline with Webster’s equation when spanned correctly. This work does not discuss the correct way to span the centerline.