Item

Abstract

One of Portuguese distinctive marks is the large inventory of nasals, including 5 nasal vowels and many diphthongs. Acoustic and articulatory studies showed nasal vowels having an initial oral part and a short nasal tail, probably related to synchronization between oral and nasal gestures. Previous studies have considered discrete descriptions with EMA-flesh points, limiting our grasp of the whole vocal tract, and preliminary work using real-time MRI (RT-MRI), considered a small framerate (14fps) and a reduced number of speakers, influencing both the time resolution to study an intrinsically dynamic process and the generalization of the outcomes. The recent advances of RT-MRI, with framerates of 50fps, have opened new possibilities for studies that can grasp a finer detail of the dynamics of nasals. However, new challenges need to be tackled to deal with the resulting large amount of data and to foster analyses that go beyond qualitative approaches to tackle a larger number of speakers. Grounded on a new RT-MRI corpus for European Portuguese, this paper explores the capabilities of recent data-driven methods, proposed for this type of RT-MRI data, to analyze dynamic aspects of nasal vowels and coordination. To this end, we consider data for 11 EP speakers and investigate vocal tract configurations, over time, and the coordination of velum and lip aperture in bilabial (oral and nasal) contexts.