Paper Title
A Facial Expression Recognition System by Analyzing the Deep Textual Features from Convolutional Neural Network Architectures

Abstract
Emotion recognition by Artificial Intelligence is being worked upon and improved by researchers and scientists worldwide to make technology more transparent to human beings. Computers are being trained to recognize and predict the emotions of animate objects, with human beings topping their priority list. However, scanty datasets with evergrowing differences between humans resulting from growth and change in the fashion and makeup industry make emotion recognition a challenging task for computers. Therefore, people in this field are exploring the potentialities of deep learning approaches to overcome these challenges. In this work, a facial expression recognition system has been proposed. The implementation of the proposed system has three components. The first component is image preprocessing, where a facial region is detected from the input image. In the second component, several existing and new deep learning architectures mainly focussed on convolutional neural network architectures have been employed and proposed to perform fea- ture representation followed by classification of facial expressions. In the third component, the classification scores obtained from these different architectures are fused to enhance the performance of the recognition system. The performance of the proposed system has been tested on two benchmark databases: Karolinska Directed Emotional Faces (KDEF) and GENKI- 4k databases. By comparing the state-of-the-art methods regard- ing these databases, it has been observed that the proposed system has achieved outstanding performance than the other competing methods. Keywords - Facial expression, CNN, Fusion, Recognition, Progressive-resizing