A human emotion recognition system using supervised self-organising maps

Abstract

Emotions constantly guide and modulate our rationality which plays an essential role in how we behave intelligently while interacting with other humans as well as machines. The technique described here provides an effective interface between humans and machines using facial expressions. This technique could be used to allow machines to incorporate an interpretation of human emotions in their principles of rationality, which could result in a more intelligent interaction with humans. In this technique, 15 feature values are calculated from the 18 feature points set on the facial images. It uses clustering based approach and supervised self-organising maps for emotion classification. The novelty of this technique is that it uses a modified form of FACS (Facial Action Coding System) to get 15 facial feature vectors of an image. Five emotions that have been considered are: neutral, anger, happy, sad and surprised. A self-clicked authentic emotion database of web-cam clicked images is used. The technique has been implemented and high efficiency has been confirmed in real-time application.