Evaluating deep learning models for EEG-based emotion recognition using the SEED-IV dataset

Bahbry, Ali Ahmed Hasan and Eng, Swee Kheng (2026) Evaluating deep learning models for EEG-based emotion recognition using the SEED-IV dataset. Journal of Electrical and Electronic Systems Research (JEESR), 29 (1): 13. pp. 108-112. ISSN e-ISSN : 3030-640X
Identification Number (DOI): 10.24191/jeesr.v29i1.013
Abstract

Emotion recognition using electroencephalography (EEG) has attracted growing interest in affective computing due to its ability to capture neural responses associated with emotional states. However, the variability and complexity of EEG signals pose significant challenges for accurate classification. This study evaluates the performance of three deep learning models—Convolutional Neural Network–Long Short-Term Memory (CNN-LSTM), Temporal Convolutional Network (TCN), and Bidirectional Long Short-Term Memory (BiLSTM)—for EEG-based emotion recognition using the publicly available SEED-IV dataset. Differential Entropy (DE) and Power Spectral Density (PSD) features were extracted from preprocessed EEG signals and used to train and evaluate the models under identical conditions. Experimental results demonstrated that the BiLSTM model delivered superior performance, achieving the highest testing accuracies of 90.38% (DE) and 88.46% (PSD). The TCN model also showed strong generalization and competitive results, securing testing accuracies of 88.03% (DE) and 85.03% (PSD), whereas the CNN-LSTM model yielded the lowest comparative results at 82.85% (DE) and 80.85% (PSD). The findings highlight the effectiveness of temporal deep learning models for capturing dynamic EEG patterns and confirm the suitability of DE and PSD features for emotion recognition tasks.

Item Details
Edit Item
Edit Item
Downloads & Files
[thumbnail of 149159.pdf]
Text
149159.pdf
Download (628kB)
Indexing & Metrics