Comparing Physiological Feature Selection Methods for Emotion Recognition

Abstract

Human-computer interactions result in psychological effects on human behavior. The analysis of the human behavior can be done using physiological data of a user in intense emotional states. A user may have intense emotions, which could make the user more nervous, sad or aggressive. This paper shows how physiological data can be used to analyze a user’s emotional state and summarizes the findings of using different feature selection and classification techniques to learn the user’s emotional states. The general flow of this approach is to record physiological signals from a person, extract features and feed them to a machine learning algorithm. This algorithm will then predict the user’s emotional state. The outcome will be helpful to analyze and understand how to train the models with the given dataset. Results of this study can be utilized for future research and applications for mitigating the effects of the content on user’s emotions.