Generation of Personalized MPEG-4 compliant Talking Heads

This paper studies a new method for three-dimensional (3D) facial model adaptation and its integration into a Text-to-Speech (TTS) system. The TTS System pronounces, in real time, English or Greek speech and simultaneously animates the adapted face model, thus simulating a natural talking face. The 3D facial adaptation requires a set of two orthogonal views of the user’s face with a number of feature points located on both views. Based on the correspondences of the feature points’ positions a generic face model is deformed non-rigidly treating every facial part (i.e. nose, mouth, etc) as a separate entity. A cylindrical texture map is then built from the two image views covering the whole area of the head by exploiting the inherent face symmetry. The result is a complete, textured model of a specific person’s head. The generated 3-D models are then integrated into a talking head system, which consists of two distinct parts: a multi-lingual Text To Speech sub-system and a Facial Animation sub-system based on MPEG-4 Facial Animation Parameters (FAPs). Support for the Greek language has been added to both the TTS and the facial animation sub-system while preserving lip and speech synchronization. Key-Words: -MPEG-4, 3D model based coding, Text to Speech, facial adaptation, talking face