부산대학교 도서관

Affect is the experience of emotion. Understanding human affect play a vital role in Man-Machine Interaction. In this work, a method to recognize the primary affective states viz. anger, disgust, fear, sad, happy and surprise of an individual from the facial expressions using eNTERFACE database with automatic peak frame selection is presented. Radon transform features extracted from the facial images and K-Nearest Neighbour (KNN) classifier are proposed for the task. Performance of the system is evaluated using the performance metrics accuracy, precision, F-measure and ROC area with five-fold cross validation. An improvement of 4.22% accuracy obtained compared to the existing work reported on the same database adds novelty to the work.