Analysis of facial features for the use of emotion recognition

The article presents a face image classification system for emotion recognition. In the first step skin recognition, using the elliptical boundary model, is performed. Then, the detection of facial features takes place. Next, an algorithm for extracting geometric and anthropometric features, from the face image is activated. Finally, training and testing classifiers are performed. We achieved averaged classification accuracy 57.7% for 6 different emotions (joy, surprise, sadness, anger, fear and disgust) and average accuracy 95.9% for 2 emotions (joy and surprise).