Visual data of facial expressions for automatic pain detection

Facial expressions are complex, progress over time, and are challenging to interpret. The research subject of automated emotion recognition associated with the facial expressions has been a mainstream topic in computer vision focused on image processing and pattern recognition. Numerous databases of facial expressions are available to the research community, and are used as fundamental tools for the evaluation of a wide range of algorithms for face expression recognition. In this paper, we assessed the existing collections of facial expression datasets as a basis for building and evaluating the largely unmapped facet of pain expressions. To accentuate this topic, the study provides the summary of different characteristics of expressions that are relevant and justifiable indicators of pain. A preliminary platform is tested with accuracy rate of 85.66% using the collected FER datasets as testing inputs. Common challenges in face recognition were discussed, as well as the different methods used to address them. Different variants of feature learning techniques were also compared, and why some methods outperforms other existing methods. (C) 2019 Published by Elsevier Inc.