Articulatory Inversion Without Articulatory Training Data

Gopala Anumanchipalli

UC San Francisco, 2015

Tuesday, June 30
12:30 p.m., Conference Room 5A

I will present some ongoing work on estimation from Electromagnetic Midsagittal Articulography (EMA) trajectories from acoustic and contextual data. I will introduce a Random Forest based statistical model for the articulatory inversion problem. I will then present some speaker adaptive extensions for articulatory estimation for target speakers with no available training articulatory data.

Bio:

Gopala Anumanchipalli is a postdoctoral researcher in the Dept. of Neurosurgery at UCSF working with Edward F. Chang on neural mechanisms of speech motor control. He has a PhD from Carnegie Mellon University and the Instituto Superior Tecnico, Lisbon.