AN ALGORITHM FOR FACIAL EXPRESSION BASED AUTOMATIC DECEPTIVE PAIN DETECTION

Abstract

Facial expressions provide a primary source for detecting pain in humans. In present study an algorithm was designed to distinguish between real and posed pain expressions using a single image as input. Geometrical features of human face were used for pain detection. Different features from the extracted feature set were selected using different feature selection algorithms. Six different classification models were used to access the proposed algorithm. Training and testing was used to evaluate the selected classifiers. Algorithm developed provided accuracy (83%), sensitivity (82.9%) and specificity (78.1%) using all features with Instance-based k nearest neighbour classifier using cross validation.