EPOK: Visual Stimuli-Based Dataset for Emotion Recognition Using EEG Signals
IEEE ACCESS, cilt.14, ss.157887-157910, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 14
- Basım Tarihi: 2026
- Doi Numarası: 10.1109/access.2026.3742760
- Dergi Adı: IEEE ACCESS
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Compendex, INSPEC, Directory of Open Access Journals
- Sayfa Sayıları: ss.157887-157910
- Kocaeli Üniversitesi Adresli: Evet
Özet
EEG-based emotion recognition is a significant research area for developing user-friendly interfaces and improving user experience (UX) in human-computer interaction (HCI). In this study, a novel EEG dataset, named EPOK, was constructed using a low-cost and portable 14-channel Emotiv Epoc X EEG headset. To minimize language-dependent influences during emotion elicitation, visual stimuli were selected from the International Affective Picture System (IAPS) and subsequently labeled using the Self-Assessment Manikin (SAM) questionnaire. Furthermore, to demonstrate the effectiveness of the proposed dataset, a hybrid deep learning model integrating Convolutional Neural Networks (CNN), Bidirectional Gated Recurrent Units (BiGRU), and an Attention mechanism was employed to classify EEG signals into valence and arousal states. Subject-dependent analysis was carried out on the DREAMER and EPOK datasets, whereas subject-independent analysis (LOSO-CV) was performed only on the EPOK dataset. Under the subject-dependent evaluation, classification accuracy reached 99.83% on the EPOK dataset, and 99.62% on the DREAMER dataset. Under the subject-independent evaluation, classification accuracies of 69.75% for valence and 71.07% for arousal were achieved on the EPOK dataset. The results demonstrate that the proposed visual-stimulus-based dataset provides a reliable foundation for EEG-based emotion recognition. These findings highlight the potential of the EPOK dataset as a valuable resource for EEG-based emotion recognition research, particularly due to its visual-only stimulus design, two-session recording protocol, and compatibility with low-cost EEG hardware.