Automated Analysis of Emotion Recognition and Personality Prediction using EEG Signals

Abstract

Both human behaviour analysis and a ective computing have been highly signi cant to the study of human-computer interaction (HCI). As a predictor of human behaviour, psychologists and human resources workers have long utilised emotion recognition and personality pro ling. The study of human emotions and personalities in order to design computer models that understand how people respond to various situations is essential for applications in many di erent areas, including entertainment, health care, military, retail, and education. However in literature di erent methods are used to analyze emotions and personality traits but they su er from data reliability issue and masking of physical behavior. The long and tedious process of identifying and estimating personalities is more of a hindrance than a help. Therefore, in order to ful ll this requirement, there must be e orts made to overcome these problems and automate the process of determining the presence of emotions and personality characteristics based on trustworthy data. A new approach is developed in this thesis for recognizing emotions and predicting Myers Briggs Type Indicator based personality traits using brain signals called as electroencephalogram (EEG) signals. In this thesis we investigate Genetic Programming (GP) and Deep Learning (DL) based models to automate the process of recognizing human emotional states and predicting personality. Hybrid GP is suggested in this study, which is e ective in the search for global and optimum solutions for the classi cation of emotions and personality traits. We also, proposed a novel tness function named as Gap score to classify unbalanced human emotion dataset. The reason behind such e cient and reliable performance of GP based frameworks is due to its expressiveness of representing computer program, exibility, and capability of performing evolutionary search. To analyze emotion and personality prediction framework, we developed a novel Long Short Term Memory (LSTM) based DL model to enhance EEG signals-based classi cation performance. The proposed GP and LSTM based frameworks are tested with benchmark datasets and achieved reliable solutions to automate the process of recognizing emotions and predicting personality.