Accurate vocal tract filter reconstruction
This task will combine and optimize the tract filter models emerging from task A.3 such as to maximize linguistic correctness, speech naturalness, and speaker-consistent sound signature in the dysphonic voice implanted artificially voiced regions.
Several technical mapping approaches will be tested including statistical projection and signal reconstruction using parametric model substitution. Being specific and dedicated to models for vocal tract filter reconstruction, this task interacts strongly with Task B.2 and will provide Task B.2 with the most effective technical solutions for vocal tract filter reconstruction that maximize the quality of voicing implantation.
Milestones
M4: Accurate vocal tract filter reconstruction algorithms.
Other expected outcomes include a task report and two MSc dissertations.
Leader: Univ. Aveiro