Accurate vocal tract filter modeling
The objective of this task is to implement an innovative ‘divide and conquer’ strategy allowing a refined estimation of the vocal tract filter using both harmonic and glottal noise information which is very important for accurate modeling and projection of formant frequencies of linguistically equivalent voiced and whispered phonemes.
In addition to a task report, the outcomes in this task include accurate models of the vocal tract filter that are important for non-invasive glottal source estimation and modeling (Task A.4), as well as algorithms implementing whispered-to-voiced formants conversion, especially in the case of the most representative Portuguese phonemes. One journal paper is expected.
Milestones
M2: Accurate vocal tract filter models and dysphonic-normal mapping rules.
Other expected outcomes include a task report and one journal paper.
Leader: Univ. Aveiro