Facial Expression Recognition Using Decision Trees

Emotion recognition from facial expressions is generally performed in three steps: face detection, features extraction and classification of expressions. The present work focuses on two points: Firstly, a new extraction method is presented based on the geometric approach. This method consists of calculating six distances in order to measure parts of the face that better describe a facial expression. Secondly, an automatic supervised learning method called decision tree is applied on two databases (JAFEE and COHEN), in order to have a facial expressions classifying system with seven possible classes (six basic emotions plus neutral), this system uses as input the six distances previously calculated (using Euclidian, Manhattan or Minkowski distance) for each face. Our results achieved a recognition rate of 89.20% and 90.61% respectively in JAFFE and COHEN database.