EEG-Based Emotion Recognition Datasets for Virtual Environments: A Survey

Abstract

One of the most important problems in virtual environments (VEs) is the difficulty users face when trying to deal with increasingly complex systems. Thus, giving machines the ability to understand human emotions would make interactions easier and more reliable. By using an EEG device as a biosignal sensor, the human emotional state can be modeled and used to realize a system that can recognize and react to human emotions. This paper provides a systematic review of EEG-based emotion recognition methods, in terms of feature extraction, time domain, frequency domain, and time-frequency domain, with a focus on recent datasets used in studies related to emotion classification using EEG and their investigation, and discusses its challenges. In the field of emotion recognition, two categories of AI-based algorithms, machine learning and deep learning, have gained great popularity. The proposed algorithms and models should be evaluated using data that include emotional ratings or labels. However, some researchers, especially those working in computer science, face challenges in building a professional experimental environment and deriving a scientifically sound experimental user model that requires specialized knowledge in psychology. Thus, many researchers interested in studying emotion recognition models choose to verify their concepts and compare them to related works using specific criteria. Therefore, investigations are presented that aim to provide a basis for future work in modeling human influence to enhance the interaction experience in virtual environments.