Using neural and distance-based machine learning techniques in order to identify genuine and acted emotions from facial expressions

Abstract

Facial expressions are part of human non-verbal communication. Automatically discriminating between genuine and acted emotion can help psychologists, judges, human-machine interface, and so on. The problems for researchers starts when there are few real emotion facial datasets available, and thus, most of experimentation for evaluation is done by using fake emotions from actors. Thus, this paper explores the problem of classifying emotions from facial expressions as genuine or acted. We propose to extract facial features from images and to classify using k-Means, k-Nearest Neighbor and Neural Network. The best results obtained presented a promising 98.6% of precision for happiness emotion and 92% for sadness emotion.