Generation of signed sentences by an avatar from their textual description

The system described is a tool dedicated to learning signs and translating written or verbal messages into signed sentences, in order to make information more accessible to deaf persons. We first focus on all the benefits such a system gives in comparison with video sequences, and then point out some of its original features - namely the generic sign description process, the parametrization of lexical items and the rendering of facial expressions. Subsequently, a synoptic view of its different modules is exposed. Finally, the implementation and achievement of the dedicated graphical user interfaces is presented in a few words, together with some experimental validation results.