Do Machines Better Understand Synthetic Facial Expressions than People?

Abstract

[eng] The recognition of facial expressions is a topic frequently addressed since the raising of Artificial Intelligence. From the point of view of the human-computer interaction (HCI), there is also an increasing interest as the emotions conveyed through expressions provide a huge amount of information when dealing with non-verbal communication. Nowadays, neural networks are one of the most used computational learning systems to recognize and analyze emotions. Generally, the main efforts are led to increase the performance of machine learning (ML) models in terms of high accuracy. But, actually, humans are not so good at distinguishing emotions. In this work, we rise the question of whether the validation of such models should rely only on performance measures or if we should also focus on trying to emulate human behavior.We try to give a fair answer performing two experiments using both, human participants and machine learning techniques.