We have contributed this second year to various activities of the InterFace project, especially in machine-to-man part (Facial Animation Tables) and modules for server (Phoneme/Bookmark to FAP converter and Dialogue Manager).
For Feedback based on facial animation, we have worked on facial animation engine based on MPEG-4 Face Animation Tables. We focused our work on different aspects:
- extension of MIRALab facial animation engine to provide automatic FAT generation
- definition of some facial test models
- development of tools to compile deformations designed by an animator into FATs
- study and implementation of different techniques to improve visual quality of facial animation
- development of Java applet for synthesis facial animation based on FATs and integration into different demonstrations. (Sample on
http://www.miralab.unige.ch/ ). We also have provided to partner a common applet form in order to integration face and body synthesis deformation in the same application.
In System Integration, we have worked for the development and the implementation of Dialogue Manager. The Dialogue Manager (DM) serves as the link between the synthesis and analysis parts of the interface platform. As a software module, it takes as input text messages annotated with emotions from the Input Module. The input to the DM is in the form of text annotated with MPEG-4 TTS tags signalling emotions and face actions. The output of the DM is the response text, annotated with face actions and emotions in an SAPI text format using emotional bookmarks. The format of the input and the output has been decided in coordination of all the concerned partners. The DM is based on a freely available/freeware text analysis tool, and is extended to provide support for emotions.
We have also worked to improve the Phoneme/Bookmark to FAP converter. This tool allows translating the data coming from the dialogue Manager and the Text-To-Speech to MPEG-4 FAP streams.
