Research Area:  Machine Learning
This Research article studies the Facial Expression Recognition (FER) research area and the evolvement of the techniques employed for FER using the Convolution Neural Network (CNN). Expressions of a human being are a significant aspect of predicting emotions, intentions, social relationships, and health conditions as well. Automated facial Expression Recognition systems have enormous applications like data-driven animation, interactive games, social robotics, neuromarketing, medical treatment, and many other human-interaction systems. In the past few decades, several researchers have come up with many methods to improve accuracy, reduce computational resources, and reduce overfitting problems. To bring significant comparisons between the techniques proposed by researchers, the research work of the last four years is taken into consideration. Based on the survey and analysis, the techniques have been summarized in the table with the issues which require attention while using CNNs for Facial Expression Recognition. This survey study focuses on the challenges faced while using CNNs for Facial Expression Recognition and the solutions to those issues. We observed that Li et al. achieved 98.2% accuracy using CK+ dataset.
Keywords:  
Author(s) Name:  Kshitiza Vasudeva; Saravanan Chandran
Journal name:  
Conferrence name:  International Conference on Communication and Signal Processing (ICCSP)
Publisher name:  IEEE
DOI:  10.1109/ICCSP48568.2020.9182076
Volume Information:  
Paper Link:   https://ieeexplore.ieee.org/abstract/document/9182076