EEG based Emotion Recognition: A Tutorial and Review (2203.11279v1)

Abstract: Emotion recognition technology through analyzing the EEG signal is currently an essential concept in Artificial Intelligence and holds great potential in emotional health care, human-computer interaction, multimedia content recommendation, etc. Though there have been several works devoted to reviewing EEG-based emotion recognition, the content of these reviews needs to be updated. In addition, those works are either fragmented in content or only focus on specific techniques adopted in this area but neglect the holistic perspective of the entire technical routes. Hence, in this paper, we review from the perspective of researchers who try to take the first step on this topic. We review the recent representative works in the EEG-based emotion recognition research and provide a tutorial to guide the researchers to start from the beginning. The scientific basis of EEG-based emotion recognition in the psychological and physiological levels is introduced. Further, we categorize these reviewed works into different technical routes and illustrate the theoretical basis and the research motivation, which will help the readers better understand why those techniques are studied and employed. At last, existing challenges and future investigations are also discussed in this paper, which guides the researchers to decide potential future research directions.

EEG-Based Emotion Recognition: A Tutorial and Review

The paper "EEG-Based Emotion Recognition: A Tutorial and Review" provides a comprehensive examination of the current state and methodologies in the field of emotion recognition using Electroencephalography (EEG) signals. As EEG-based emotion recognition technology becomes increasingly relevant in areas such as emotional healthcare, human-computer interaction (HCI), and multimedia recommendations, this paper serves as a significant resource for researchers attempting to navigate the complexities of the subject.

Overview

The paper begins by addressing fundamental concepts essential for understanding EEG-based emotion recognition. It delineates the scientific basis stemming from psychological and physiological observations that link EEG signals to emotional states. The authors detail both discrete and continuous models of emotion, highlighting how different approaches quantitatively define emotional states. Discrete models view emotions as distinct categories, while continuous models employ multi-dimensional space, usually covering axes such as valence and arousal.

EEG in Emotion Recognition

The paper emphasizes the pivotal role of the brain's central nervous system in emotional processing and its observable manifestations through EEG signals. Experimental findings have suggested correlations between specific EEG frequency bands and emotional states, such as the association of alpha band asymmetry with different valence states. The review suggests that while promising results have been achieved, the precise brain activity characteristics and localization of neural sources in emotional processes are still under investigation.

Methodological Advances

In terms of technical methodologies, the authors review both classical and modern approaches:

1. Feature Engineering: Traditional machine learning methods rely heavily on feature engineering, using time-domain, frequency-domain, and non-linear dynamical system features. These methods face challenges due to the noise and complexity of EEG signals but remain fundamental to EEG-based emotion recognition.
2. Deep Learning Approaches: Deep Learning (DL) models, particularly CNNs, RNNs, and hybrid models like CNN-RNN combinations, have been leveraged for their capacity to automatically learn features directly from raw data. These algorithms benefit from their ability to model spatial-temporal dependencies in EEG signals.
3. Graph-Based Methods: With the introduction of Graph Neural Networks (GNNs), researchers seek to better capture the underlying functional connectivity networks of the brain, reflecting the complex relationships across EEG channels.

Challenges and Future Directions

Despite the considerable advancements, several challenges remain, including:

• Domain Shift: EEG data variability across different subjects due to physiological, cultural, or gender factors presents significant hurdles. The review discusses potential solutions such as domain adaptation techniques and transfer learning methodologies.
• Real-time Application: Most current methodologies are computationally intensive, limiting real-time application prospects. Further research is necessary to optimize these models for real-time deployment without sacrificing accuracy.
• Integration with Other Modalities: The integration of EEG with other data sources—such as facial expressions, peripheral physiological signals, or environmental context—could provide richer datasets for more robust emotion recognition systems.

The authors propose investigating biologically inspired models and AutoML approaches to ease up the process of model building, reduce dependency on domain expertise, and tap into large-scale pre-trained models akin to those seen in NLP for generalized EEG application scenarios.

In conclusion, EEG-based emotion recognition holds great promise, but it requires nuanced understanding and improved methodologies to tackle the inherent complexities and achieve real-world applicability. This paper sums up current trends, proposes innovative approaches, and paves a path for future research in this interdisciplinary field.