Real time expression recognition using correlation and support vector machine

Facial expressions deliver rich information about human emotion and play an essential role in human communications. This paper presents a design and evaluation of a novel computational model that categorizes facial expressions in real time video for the reason of automating human computer interfaces. It highlights the main system components, methodology for the development of the prototype and some research challenges. The concepts of correlation have been used to detect the face in video sequences and multiclass SVM is used for classification. The method has been evaluated in terms of recognition accuracy using a well known Facial Expression database, Japanese Female Facial Expression database as well as using the database of face images of the authors. The experimental results show the effectiveness of our scheme.