Human Emotion Recognition Based on Facial Expression Using Convolution Neural Network

EMOTION recognition has become an essential aspect of human-computer interaction, encompassing a wide range of applications such as virtual reality, cognitive science, and digital health. Identifying human emotions through facial expressions is challenging due to the intricate and constantly changing nature of facial movements. However, the advancements that have been made in deep learning methods, specifically Convolutional Neural Networks, have shown promising results in this field. This study aims to present a Convolutional Neural Network-based deep learning model for precise identification of emotions from facial expressions. The proposed system underwent training and evaluation using the Multimodal EmotionLines Dataset (MELD) and The Ryerson Audio-Visual Database of Emotional Speech and Song (RAVDESS). The experimental results validate the efficacy of the suggested methodology, attaining a remarkable accuracy rate of 96.3% in the identification of emotions for the MELD dataset and 95.86% accuracy rate in RAVDESS dataset. The system effectively tackles the challenge of distinguishing between fear and surprise emotions, which can exhibit considerable similarity, making differentiation more challenging.