Abstract

Facial expressions play an important role in human communication giving real substance to face-to-face interaction in real life speech. For many years researchers have been involved in automatic recognition and generation of facial expressions.
Motivated by this issue, we investigate the design of a system that is able to recognize the emotional character of the speech, being able to enhance it adding facial expressions that convey its sense. The recognition is automatic which means that it doesn’t need subjective human annotations. The lexical database, Wordnet, and some scripts to access it are used to classify the emotional words in a 2D-space. This space allows us to classify the words according to their degree of pleasure and their degree of activation. Calculating the distances between the words in this space we conclude which facial expression conveys better the emotion of speech.
To develop an automatic system that analyzes human emotions from a text is a difficult task, but the Automatic Smiley Labeling system or ASL has been tested and found to be a viable approach for this purpose based on the lexical meaning of the words.