The recent advances in real-time magnetic resonance imaging (RT-MRI) for speech studies, providing a considerable increase in time resolution, potentially improve our ability to study the static and dynamic aspects of speech production. To take advantage of the sheer amount of the resulting data, automated methods need to be used to select, process, and analyze the data, and our work has previously tackled these challenges for an European Portuguese corpus acquired at 14 frames per second (fps). Aiming to further explore RT-MRI in the study of the dynamic characteristics of EP sounds, e.g., nasals, we present a novel 50 fps RT-MRI corpus and assess the applicability, in this new context, of our previous proposals for processing and analysis of these data to extract relevant articulatory information. Importantly, at this stage, we were interested in assessing if and to what extent the new data and the proposed methods provide replicability of the articulatory analysis obtained from the previous corpus. Overall, it was possible to automatically process and analyze the 50 fps data and a comparison of automated analysis performed for the same sounds, for both corpora (i.e., 14fps and 50fps), yields similar results, demonstrating the envisaged replicability.
Cite as: Cunha, C., Silva, S., Teixeira, A., Oliveira, C., Martins, P., Joseph, A., Frahm, J. (2018) Exploring Advances in Real-time MRI for Speech Production Studies of European Portuguese. Proc. IberSPEECH 2018, 137-141, doi: 10.21437/IberSPEECH.2018-29
@inproceedings{cunha18_iberspeech,
author={Conceicao Cunha and Samuel Silva and António Teixeira and Catarina Oliveira and Paula Martins and Arun Joseph and Jens Frahm},
title={{Exploring Advances in Real-time MRI for Speech Production Studies of European Portuguese}},
year=2018,
booktitle={Proc. IberSPEECH 2018},
pages={137--141},
doi={10.21437/IberSPEECH.2018-29}
}