A statistical shape space model of the palate surface trained on 3D MRI scans of the vocal tract

We describe a minimally-supervised method for computing a statistical shape\nspace model of the palate surface. The model is created from a corpus of\nvolumetric magnetic resonance imaging (MRI) scans collected from 12 speakers.\nWe extract a 3D mesh of the palate from each speaker, then train the model\nusing principal component analysis (PCA). The palate model is then tested using\n3D MRI from another corpus and evaluated using a high-resolution optical scan.\nWe find that the error is low even when only a handful of measured coordinates\nare available. In both cases, our approach yields promising results. It can be\napplied to extract the palate shape from MRI data, and could be useful to other\nanalysis modalities, such as electromagnetic articulography (EMA) and\nultrasound tongue imaging (UTI).\n