Systematic Analysis of EEG Emotion Recognition: Subject-Dependent vs. Independent Approaches

Abstract

Electroencephalography (EEG) signals, which show human brain activity, hold potential beyond medical diagnosis. They are particularly valuable for emotion recognition, while analyzing these EEG signal to distinguish between various emotional states. In this area, research focuses on two tasks including subject-dependent and subject-independent tasks.  While machine learning and deep learning models have been developed to use EEG data for recognizing emotions, achieving high accuracy is still challenging due to the complex and non-stationary nature of these signals. A key challenge is extracting features that encapsulate different information aspects including temporal, frequency, and spatial information. This article presents a systematic review and analysis of the most significant machine and deep learning techniques for the EEG-based emotion recognition task, and outlines the challenges faced by the current models.