Facial Emotion Recognition

Abstract

The next generation concept evolving in the field of Artificial Intelligence is how to improve Machine Perception and Social Intelligence by making a smarter system capable of reading and understanding human behavior. Human beings communicate with each other through mutual conversations and they process this information by one of the following medium: speech and vision. Therefore, facial emotion recognition can be considered as a vital and useful visual based tool for building systems which can identify, interpret, process, and sim- ulate human emotions. The traditional approach for performing facial emotion recognition is tracking changes in the facial muscles which are defined as Action Units(AU)[1]. Although Action Units has proven to be a quite successful approach in the process of identification of facial expressions, there are a total of 7000 AUs combinations of different AUs charac- terized to distinguish the emotions which can prove to be really a very extensive and time consuming procedure [2]. In this paper we provide an alternative approach of using feature points instead of action units to develop a faster and efficient recognition method. In this paper we attempt to achieve higher efficiency in recognizing human emotions accurately by using minimal number of geometrical feature points. We first take pre-processed dataset of JAFFE image database as our input, then we perform extraction of the feature points of only two features: Eyes and Eyebrows of human face. In the final stage, we generate some geometrical features on basis of the extracted feature points and classify them by taking as input vectors for the Random Forest Algorithm and are able to achieve 87.4% recognition rate when we run it across for 213 images in overall.