A 3D Facial Expression Recognition System Based On SVM Classifier Using Distance Based Features

Abstract

In this study, an SVM-based system is proposed for the classification of facial expressions that are represented in 3D. Distance based features are used as a feature vector, which are determined by the distances between the different key points on the image. Study was conducted on a subset (Happy, sadness, surprise) of Bosphorus 3D Face Database. 9 different fiducial points arc used to calculate a total of 5 distance features. SVM classification was performed with K-fold cross validation thus mean classification performance of different training and test clusters were determined. %85 success rate has achieved as a result of the expression analysis performed on the 3D facial scans.