Emotion Recognition Using EEG Signals in Human Brain Waves and Deep Learning Approaches
Issued Date
2025-01-01
Resource Type
Scopus ID
2-s2.0-105004549067
Journal Title
10th International Conference on Digital Arts, Media and Technology, DAMT 2025 and 8th ECTI Northern Section Conference on Electrical, Electronics, Computer and Telecommunications Engineering, NCON 2025
Start Page
363
End Page
367
Rights Holder(s)
SCOPUS
Bibliographic Citation
10th International Conference on Digital Arts, Media and Technology, DAMT 2025 and 8th ECTI Northern Section Conference on Electrical, Electronics, Computer and Telecommunications Engineering, NCON 2025 (2025) , 363-367
Suggested Citation
Mekruksavanich S., Hnoohom N., Phaphan W., Jitpattanakul A. Emotion Recognition Using EEG Signals in Human Brain Waves and Deep Learning Approaches. 10th International Conference on Digital Arts, Media and Technology, DAMT 2025 and 8th ECTI Northern Section Conference on Electrical, Electronics, Computer and Telecommunications Engineering, NCON 2025 (2025) , 363-367. 367. doi:10.1109/ECTIDAMTNCON64748.2025.10962120 Retrieved from: https://repository.li.mahidol.ac.th/handle/123456789/110134
Title
Emotion Recognition Using EEG Signals in Human Brain Waves and Deep Learning Approaches
Author's Affiliation
Corresponding Author(s)
Other Contributor(s)
Abstract
Emotion recognition using EEG signals has gained growing attention in recent years due to its promising applications in human-computer interaction, affective computing, and mental health assessment. This study investigates the performance of various deep learning architectures in classifying emotional states from EEG recordings. We utilize the Emotion EEG dataset as a benchmark and evaluate five prominent deep learning models. The classification task targets three emotion categories: neu-tral, positive, and negative. Experimental results reveal notable differences in model performance. Among the tested models, BiGRU achieves the highest classification accuracy of 97.70% (±0.69%), followed closely by BiLSTM with 97.37% (±0.17%). These bidirectional architectures significantly outperform the CNN model, which attains an accuracy of only 61.59% (±1.22%). The LSTM and GRU models yield intermediate results, nearing 92% accuracy. The findings highlight the effectiveness of bidirectional recurrent neural networks in modeling the temporal characteristics of EEG signals for emotion detection. This research contributes to advancing EEG-based emotion recognition by identifying the most suitable deep learning strategies, thus supporting the development of more intelligent and emotion-aware computing systems.