A new fuzzy rule-based approach for automatic facial expression recognition

One of the most important methods in human interaction is emotion representation and the most common way is facial expression. In artificial intelligence, automatic facial expression recognition is the basic step in automatic interaction between human and computers. Many different methods have been proposed recently. None of them can do it ideally and completely automatic, however different methods try to improve or solve some parts of it. In this research, in addition to find some new deformation in face when showing a facial expression, a new fuzzy rule based method is proposed to classify features extracted from image sequence of facial expression and detect facial expression. Motion vectors are the features extracted from image sequences and used for classification. They are extracted using optical flow algorithm and classified using Sugeno type fuzzy inference system. Experimental results show that its performance is 81.22%.

Keywords:

automatic facial expression recognition , optical flow , Sugeno type fuzzy inference system