Customized CNN with Adam and Nadam Optimizers for Emotion Recognition using Facial Expressions

Abstract

People communicate using one of the communication types of facial expressions to express their emotions. Human feelings are detected through facial expressions to interpret their present state of mood. It stimulates researchers to work in the field of emotion recognition. The design of deep learning models is essential to interpret the human current mind state by capturing the pattern of the facial gesture through their facial expressions. This study proposed a customized Convolutional Neural Network (CNN) with various optimizers Adaptive Moment Estimation (Adam) and Nesterov-accelerated Adaptive Moment Estimation (Nadam) to improve emotion recognition using the dataset FER-2013. The customized proposed model is designed by varying the number of convolution layers, filters, filter sizes, and optimizers. The emotions are recognized using softmax activation in the output layer. The experimental results have proved that the proposed model classified the facial expressions with accuracy of 0.841, 0.826 using Nadam and Adam optimizers respectively.