Facial Expression Recognition System

This thesis describes the problem of facial expression recognition in the field of computer vision. Firstly, the psychological background of a problem is presented. Then, the idea of facial expression recognition system (FERS) is outlined and the requirements of such system are specified. The FER system consists of 3 stages: face detection, feature extraction and expression recognition. Methods proposed in literature are reviewed for each stage of a system. Finally, the design and implementation of my system are explained. The face detection algorithm used in the system is based on work by Viola and Jones [13]. The expressions are described by appearance features obtained from texture encoded with Local Binary Patterns [32]. The Support Vector Machine with RBF kernel function is used for classification. Databases that were used are: The Facial Expressions and Emotion Database [34], which contains spontaneous emotions and CohnKanade Database [35] with posed emotions. The system was trained on two databases separately and achieves accuracy of 71% for spontaneous emotions recognition and 77% for posed actions recognition.