Localization and Classification of Human Emotions using Deep Neural Network

Abstract

Emotions are basic part of human living and they reflect the diverse aspects of human behavior. The way a person responds to certain situations, is very essential to study the emotional development. This paper discusses the difference between the activation of two brain lobes when subject is exposed to  High Arousal Positive Valence (HAPV) emotions like Excited, Happy and  Low Arousal Negative Valence (LANV) emotions like sad, bored. DEAP dataset is used to compare HAPV and LANV. Localization and Dipole fitting of most contributing Independent components is performed. Results suggest that Right lobe shows higher activation level for HAPV state while left lobe shows higher activation level for LANV. These emotions are further classified using deep neural network. Result suggests an average accuracy of 86.5%.