Database minimization of unit selection in visual speech synthesis without loss of naturalness:

Talking heads are generated using the minimized databases, which are clustered by LBG, RSST, and MST. The number of mouth images in the database is reduced by the clustering algorithms from 11594 to 3000.
The target of the database minimization is to enable the talking head for Internet applications and devices with limited storage space.
The software of a talking head with a minimized database is available.




1. Demo of Animations generated by minimized databases

   1. original

   2. LBG-based clustering

   3. RSST-based clustering

   4. MST-based clustering

   5. MST-based clustering without use of probability of the mouth images







2. Contact

Prof. Dr.-Ing. Jörn Ostermann

M.Sc. Kang Liu

Leibniz Universität Hannover

Institut für Informationsverarbeitung

Appelstr. 9A, D-30167 Hannover

Tel. +49(0)511 7625316

Fax +49(0)511 7625333

Email: kang@tnt.uni-hannover.de







TNT Home | LUH | Search | Administrator Updated 27/02/2009 Webpage design M.Sc. Kang Liu