Evaluating facial expressions in real time

In this paper an overview of recognizing facial expression is presented using machine learning techniques and open source libraries. There are three different stages to recognize facial expressions. Firstly, we detect faces from non-faces. Secondly, we extract features using a geometric based approach. Finally, we classify various expressions using support vector machines into pre-defined categories. The system is tested on the Cohn-Kanade Extended (CK+) facial expression image dataset which scores a recognition accuracy of 93%.