Pattern recognition of emotion with neural network

Abstract

Proposes an emotion model for communication which also transfers personality and character information. The emotion model customizes to individual human communication partners by learning. Learning is achieved by neural networks converting input voice signals to an emotion state. The emotion state decides the response of the partner. The emotion state is divided into four categories: sadness; cheerfulness; happiness; and anger. For example, a loud voice causes the emotion of anger. This paper also discusses the emotion model as network agent between two human communication partners.