Perceptually natural voicing implantation and prosodic control
This task aims at the development of a method allowing to convert whispered speech or dysphonic voice, into artificially voiced speech; this conversion should be performed in such a way that the linguistic content of the speech is enhanced, the resulting speech sounds more intelligible and natural than whispered speech or dysphonic voice, and that the speech conveys as much as possible acoustic cues allowing to easily recognize the speaker.
Testing procedures will be designed and conducted in order to assess the effectiveness of the voice conversion process, in terms of listener experience and, as much as possible, patient self-perception.
Milestones
M5: Accurate artificial voicing implantation algorithms.
Other expected outcomes include a task report, one patent application, one PhD dissertation, one journal paper and one conference paper.
Leader: FEUP