Facial Emotion Recognition Using Ensemble Learning

Abstract

Facial emotion recognition (FER) is the task of identifying human emotions from facial expressions. The purpose of this book chapter is to improve accuracy of facial emotion recognition using integrated learning of lightweight networks without increasing the complexity or depth of the network. Compared to single lightweight models, it made a significant improvement. For a solution, the authors proposed an ensemble of mini-Xception models, where each expert is trained for a specific emotion and lets confidence score for the vote. Therefore, the expert model will transform the original multiclass task into binary tasks. The authors target the model to differentiate between a specific emotion and all others, facilitating the learning process. The principal innovation lies in our confidence-based voting mechanism, in which the experts “vote” based on their confidence scores rather than binary decisions.