A Study on Facial Expression Recognition in Assessing Teaching Skills: Datasets and Methods

Abstract

Facial expressions recognition (FER) is an important modality in the future assessing teaching skill system (ATS). The fundamental difference with FER is in its low resolution image, unique occlusion of veiled teachers, and varied lighting conditions. However, it is still able to solve the problem of variations in head poses. This study, overall, began with the collection of articles from Scopus, created the FER taxonomy class and mapped the use of datasets. After finding out which datasets were suitable for the image of expression in teaching, a deeper discussion about classifier-classifier (using the same CK+ dataset) was capable of solving problems (low resolution image, occlusion, variation in lighting conditions and head poses). Finally, from the discussion conducted, it is known the potential for modification of algorithms, appropriate datasets and research opportunities for future FER in ATS.