INTERSPEECH 2013 14thAnnual Conference of the International Speech Communication Association Lyon, France August 25-29, 2013

Training an Articulatory Synthesizer with Continuous Acoustic Data

Santitham Prom-on (1), Peter Birkholz (2), Yi Xu (3)

(1) King Mongkut's University of Technology Thonburi, Thailand
(2) University Hospital Aachen, Germany
(3) University College London, UK

This paper reports preliminary results of our effort to address the acoustic-to-articulatory inversion problem. We tested an approach that simulates speech production acquisition as a distal learning task, with acoustic signals of natural utterances in the form of MFCC as input, VocalTractLab . a 3D articulatory synthesizer controlled by target approximation models as the learner, and stochastic gradient descent as the training method. The approach was tested on a number of natural utterances, and the results were highly encouraging.

Full Paper

Bibliographic reference.  Prom-on, Santitham / Birkholz, Peter / Xu, Yi (2013): "Training an articulatory synthesizer with continuous acoustic data", In INTERSPEECH-2013, 349-353.