Emotion Classification: Novel Deep Learning Architectures

Abstract

Emotion Classification is an important task for the machines to understand the emotional changes in human beings. In this research paper, we utilize a combination of Convolutional Neural Networks and Auto-Encoders to extract features for Emotion Classification. We proposed a total of six architectures. Among them, two architectures are trained on JAFFE (Japanese Female Facial Expressions), remaining four architectures are trained with Berlin Database of Emotional Speech. Good classification accuracy is attained with these architectures.