Gerçek zamanlı konuşma MRI görüntüleri üzerinden şekil ön-modelleri kullanarak ses yolu doku sınır çizgisinin takibi

Human speech production is a process, which is initiating from lungs, phonating in the vocal folds and ultimately passing through the vocal tract comprising several jointly working articulators. Knowledge about the dynamic shape of the vocal tract is the basis of many speech production applications such as, articulatory analysis, modeling and synthesis. Vocal tract air-way tissue boundary segmentation in the mid-sagittal plane is necessary as a pre-step for 3D-reconstruction of the tract shape. This segmentation problem is however challenging due to poor resolution real-time speech MRI, grainy noise and the rapidly varying vocal tract shape. In this thesis we present a robust approach to vocal tract airway-tissue boundary contour tracking by training a model for human vocal tract. We generate a dataset of manually segmented vocal tract and utilize a statistical approach to train a model for the tract. An active contour approach is employed to segment the air-way tissue boundaries of the vocal tract while restricting the curve movement to the trained shape model. Then the contours in subsequent frames are tracked using dense motion estimation methods. Experimental evaluations over the mean square error metric indicate signi cant improvements compared to the state-of-the-art