Facial Expression Recognition Using Geometric Normalization and Appearance Representation

Facial expression recognition is a challenging andinteresting problem in computer vision and pattern recognition.Geometric variability in both emotion expression and neutralface is a fundamental challenge in facial expression recognitionproblem. This variability not only directly affects geometricfacial expression recognition methods, but also is a criticalproblem in appearance methods. To overcome this problem, thispaper presents an approach which eliminates geometricvariability in emotion expression; thus, appearance features canbe accurately used for facial expression recognition. Therefore, afixed geometric model is used for geometric normalization offacial images. This model is defined as one of the emotionalexpressions. In addition Local Binary Patterns are utilized torepresent facial appearance features. Experimental results showthat the proposed method is more accurate than the existingworks. Also for facial expression recognition, using geometricexpression models of facial images where they have larger size inmouth/eyes regions, such as Surprise, gives better resultsindicating that mouth and eyes are important regions in emotionexpression.

Keywords:

facial expression recognition , face geometry normalization , Local Binary Patterns (LBP) , Piecewise Linear Warp , Support Vector Machine (SVM)