Face Emotion Detection Using Deep Learning

Abstract

Human facial expressions convey abundant information visually instead of vocally. Face expression recognition plays an important role within the world of human-machine interaction. Recognition of facial expression by computer with high recognition accuracy remains a challenging task. This article gives the summary of current Facial Emotion Recognition (FER) stages, techniques, and datasets. FER is usually carried out in three stages involving face detection, feature extraction, and expression classification. In this work, we have used deep learning algorithm to identify the basic human emotions (e.g., anger, fear, neutral, happy, sad, surprise, etc.) on multiple datasets, including FER-2013 (Facial Expression Recognition 2013) and CK+ (Extended Cohn-Kanade). The accuracy of our model, using CNN is 60% for FER 2013 dataset, and for CK+, we achieved significant improvement, highest accuracy was 99.1% and average accuracy was 93% which is better than the one reported.