Abstract

A human face is an extremely important source of information during human communication. We learn to recognize facial expressions long before we learn to communicate verbally. Facial expressions are the most effective way to communicate our emotions. For years most researchers centred their attention in the automatic recognition of facial expressions and the most important for them was the recognition of the shape of mouth, eyes and eyebrows. Nowadays, researchers are getting more interested in the automatic generation of facial expressions using an input text. This approach can be of a big importance in many disciplines, such as human communication through computer interfaces and games.
Based on this approach, my Master thesis will be focused on building a Non-verbal Emotional Dictionary (NED) where the most important aspect is the relation between from one side an emotional word, and from the other side its corresponding facial expression coded by Action Units (AU) representation and facial expression. This dictionary can then be used as a bridge between an emotional text analyzer and a 3D Talking face system to mimic humane conversations using emotions in a virtual world.