A Comparative Review of EEG-Based Emotion Recognition Approaches Using Machine and Deep Learning

• Author(s): Tijesunimi Deborah Olashore ; Olutayo K. Boyinbode ; Oladunni A. Daramola ; Mary T. Kinga
• Paper ID: 1708003
• Page: 1070-1084
• Published Date: 29-04-2025
• Published In: Iconic Research And Engineering Journals
• Publisher: IRE Journals
• e-ISSN: 2456-8880
• Volume/Issue: Volume 8 Issue 10 April-2025

Abstract

The field of emotion recognition using electroencephalography (EEG) has witnessed rapid advancement, particularly through the integration of machine learning (ML) and deep learning (DL) techniques. This paper presents a comprehensive comparative review of 25 recent studies, examining diverse methodological approaches to EEG-based emotion recognition. The analysis spans a wide range of models, including traditional Machine Learning classifiers such as Support Vector Machines (SVM), K-Nearest Neighbours (KNN), Naïve Bayes, and Decision Trees, as well as advanced Deep Learning architectures like Convolutional Neural Networks (CNN), Long Short-Term Memory (LSTM) networks, 3D-CNNs, Deep Hybrid Networks, and Graph Convolutional Neural Networks (GCNNs). Reported classification accuracies vary from 83% to over 99%, with DL models consistently outperforming conventional ML algorithms most notably with CNN and LSTM-based models achieving accuracies exceeding 90% in arousal and valence classification tasks. Feature extraction techniques range from traditional statistical measures and clustering algorithms to modern approaches like 3D feature maps, EEG spectrograms, group phase locking values, and electrode-frequency distribution maps (EFDMs). Fusion strategies combining EEG with other modalities (e.g., facial expressions) or hybrid classifiers (e.g., CNN-SVM, CNN-XGBoost) also demonstrate notable performance improvements. Publicly available datasets such as DEAP, DREAMER, SEED and MPED are widely used, as well as custom datasets tailored for this specific purpose of emotion recognition, enabling consistent benchmarking across studies. However, challenges such as subject variability, data imbalance, noise sensitivity, and the need for real-time implementation persist. This review highlights emerging trends in model design and offers critical insights into the strengths and limitations of current EEG-based emotion recognition approaches. The findings serve as a valuable resource for guiding future research toward more robust, scalable, and accurate emotion recognition systems.

Keywords

Electroencephalography, Feature Extraction, Machine Learning, Deep Learning

Citations

IRE Journals:
Tijesunimi Deborah Olashore , Olutayo K. Boyinbode , Oladunni A. Daramola , Mary T. Kinga "A Comparative Review of EEG-Based Emotion Recognition Approaches Using Machine and Deep Learning" Iconic Research And Engineering Journals Volume 8 Issue 10 2025 Page 1070-1084

IEEE:
Tijesunimi Deborah Olashore , Olutayo K. Boyinbode , Oladunni A. Daramola , Mary T. Kinga "A Comparative Review of EEG-Based Emotion Recognition Approaches Using Machine and Deep Learning" Iconic Research And Engineering Journals, 8(10)