FACIAL EXPRESSION RECOGNITION USING CONVOLUTIONAL NEURAL NETWORKS [CNN]

Abstract

While humans have traditionally excelled at discerning emotions through facial expressions, achieving the same capability with a computer program has proven challenging. However, recent advancements in computer vision and machine learning have made it feasible to recognize emotions accurately. This paper introduces a novel facial emotion identification technique known as Convolutional Neural Networks for Facial Emotion Recognition (FERC). FERC utilizes a twopart convolutional neural network (CNN) architecture: The paper consists of two main sections. The first section is dedicated to removing the background from the image, while the second section focuses on extracting the facial feature vector. The FERC model utilizes an expressional vector (EV) to identify five different regular facial expressions. With its single-level CNN approach, FERC offers improved accuracy compared to commonly used techniques. Moreover, prior to constructing the EV, a novel background removal strategy is implemented to address various potential challenges, such as distance from the camera. The application of FERC in emotion detection holds promise for a wide range of applications, including student predictive learning and lie detection.