An Affective Computing Electroencephalogram-based System with Machine Learning Algorithms

Affective Computing plays a crucial role in the Human-Computer Interaction (HCI) field by enabling computers to recognize, interpret, process, and simulate emotions evoked by positive/negative events, objects, or situations. It presents new avenues for applications ranging from personalized user interfaces to mental health monitoring. In this study, an affective computing system for human emotion recognition is developed by measuring the electrical activity of brain, Electroencephalogram (EEG) signals for 15 participants were recorded. Using Machine Learning (ML) algorithms to classify emotional states into three emotion dimensions: valence, arousal, and dominance. Two ML algorithms, Naïve Bayes (NB) and Artificial Neural Network (ANN) are exploited. The system also validated with EEG data from the public DREAMER dataset and compared with recorded data. Results demonstrate the feasibility and effectiveness of the proposed EEG-based affective computing system in accurately identifying and categorizing emotions, leveraging the designed ANN classifier; a maximum accuracy of 95.61% is achieved.