Development of facial expression recognition system

Abstract

Enabling computer systems to recognize facial expressions and infer emotions from them in real time presents a challenging research topic. The recognition of emotional information is a key step towards giving computers the ability to interact more naturally and intelligently with people. One of the potential applications of face detection and facial expression recognition is in human computer interfaces. The system will be used for the interaction between human and humanoid robot head, where the detected expression will be mimicked by the robot head. The problem of facial recognition can be divided into two major areas: detection of the face region and identification of the detected region. Detecting human face in computer vision proves to be very challenging due to the fact that human faces can have different forms and colors, adverse lighting conditions, varying angles or view points, scaling differences and different backgrounds. Attempting recognition on an inaccurate detected face region is hopeless. This thesis describes a face detection framework that is capable of processing input images swiftly while achieving high detection rates. The presented face detection system is developed using the response of Haar-Like features and AdaBoost algorithm. A set of experiments in the domain of face detection is presented in this research. The developed system yields face detection performance comparable to the best existing systems, where its accuracy is up to 98%. The face and facial features detected in the video stream are used as input to a Support Vector Machine classifier, which is used for facial expression recognition. The method was evaluated in terms of recognition accuracy for a variety of interaction and classification scenario, and it was proven that the system is able to detect the four expressions successfully. The person-dependent and person-independent experiments demonstrate the effectiveness of a support vector machine to fully automatic and unobtrusive expression recognition in real time.