Leveraging Convolutional Neural Network for Real-Time Facial Emotion Detection

Facial emotion detection is a very relevant research domain, having its applications in a wide diversified field of topics that include statistical analysis, the social sciences, and computer vision. Developing methods for machines that can recognize and interpret human emotions in real-time has been revolutionary for improving human-computer interactions, enhancing security systems, and helping with substantial insights from psychological studies. In this paper, it is proposed to develop a new model of emotion recognition where emotions through facial images will be classified based on seven different emotional features. Taking grayscale photos from a dataset called fer2013, we train a Convolutional Neural Network for the purpose of identifying these emotions. As far as the approach applied in this paper is concerned, the process of emotion identification is basically divided into three major phases: face detection, feature extraction, and emotion classification. We tried different depths and layers of CNNs, achieving as high an accuracy as 89.98%, which has huge potential for deep learning in advancing FER technologies. The contribution with this research is to developing efforts on more intuitive and responsive AI systems that can understand and respond to human emotions in an emotional sense.