EEG Emotion Classification based on Valence and Arousal using DEAP Dataset

Abstract

Emotion Recognition using Electroencephalography (EEG) signals has emerged as an important area of research due to its wide applications related to mental health monitoring, affective computing and Human Computer Interaction (HCI). However, majority of high performing emotion recognition systems are computationally intensive, lack interpretability making them ineffective for real time application. In this paper, we present a simple yet effective model for EEG emotion classification using an ensemble of basic machine learning classifiers. The model is enhanced with frequency augmentation technique which efficiently classifies the emotional states based on Valence and Arousal while maintaining interpretability and reducing complexity. A comprehensive analysis comparing the effectiveness of   various feature sets, such as Statistical methods, Hjorth parameters, Entropy-based features, Frequency band power, Fractal Dimension has been performed on Database for Emotion Analysis using Physiological Signals (DEAP) dataset. The model was trained using K-Nearest Neighbors (kNN), Random Forest (RF), and CatBoost (CB) classifiers along with their ensemble model utilizing the strength of each algorithm. Results demonstrate high accuracy in classifying Valence (91%) and Arousal (89%) dimensions. We demonstrate through our work that even with basic machine learning classifiers, high accuracy can be achieved for Valence and Arousal labels using the EEG signals.