A Novel Facial Expression Recognition System Using Convolutional Neural Network

Abstract

Effective communication requires the use of facial expressions, which are the most natural response in the human heart. Clinical practices and behavioral description are important for human-computer interaction, and facial expression recognition (FER) is also important. There has been a lot of study on FER, with expanding applications such as neuromarketing and avatar animation. Machine learning techniques have a difficult time recognizing facial expressions because people can differ in how they display their emotions. Even a photograph of the same person in one expression can differ in brightness, backdrop, and position. As a result, computer vision's FER challenge remains difficult. In this study, we present a novel FER solution that combines certain image pre-processing processes with common techniques like convolutional networks. Depending on the feature set, our proposed model provides significant results of 91.29% accuracy on FER2013 dataset.