|
Deriving Articulator Shape Information from Magnetic Resonance Imaging (MRI) Scans| old_uid | 9989 |
|---|
| title | Deriving Articulator Shape Information from Magnetic Resonance Imaging (MRI) Scans |
|---|
| start_date | 2015/10/12 |
|---|
| schedule | 13h30-14h30 |
|---|
| online | no |
|---|
| summary | Analyzing the vocal tract during speech is of great interest in speech science.
In particular, understanding the motion and shape of the major articulators like the tongue is important.
The acquired information can be used in simulations of the vocal tract to perform articulatory speech synthesis or to create animated virtual avatars.
In my talk, I discuss a work-in-progress framework for training a shape space model of an articulator by using MRI scans of the vocal tract.
In particular, we show how MRI scans can by denoised by using a diffusion based method and focus on segmenting them for finding the spatial support of tissue.
Afterwards, we show how template fitting can be used to extract the shape information of an articulator from the found spatial support.
We also show how to reconstruct the hard palate shape in a scan by using information from a different scan.
This is especially useful in cases when its boundary is barely visible because of contacts between the tongue and the hard palate.
Finally, we inspect how a statistical shape space of an articulator can be derived from the extracted shapes. |
|---|
| responsibles | Hueber |
|---|
| |
|