Name
Abstract | ||
|---|---|---|
Emotion recognition based on EEG has a wide range of applications in the e-healthcare systems, especially it helps diagnose and treat a variety of mental illnesses such as depression. Due to individual differences between subjects and the non-stationary characteristic of EEG, traditional emotion recognition methods are difficult to achieve good performance. In this paper, we propose a dynamically organized graph neural network (DOGNN) for EEG-based emotion recognition. Unlike previous studies that require a fixed graph structure, the proposed DOGNN method can dynamically learn the intrinsic relationship between different electroencephalogram (EEG) channels based on frequency band information, and further, construct graph structure for every subject. We conduct extensive experiments on two open public datasets (SEED and SEED-IV). The experimental evaluation exhibits that the proposed method achieved better performance. In addition, we visualize the topographic maps for different frequency bands learned by the proposed model and the result is consistent with previous neuroscience studies. This demonstrates that our approach is capable of capturing more effective the frequency band and spatial features for EEG emotion recognition. |
Year | DOI | Venue |
|---|---|---|
2022 | 10.1007/978-3-030-98355-0_29 | MULTIMEDIA MODELING, MMM 2022, PT II |
Keywords | DocType | Volume |
EEG, Emotion recognition, Graph neural network, Graph construction | Conference | 13142 |
ISSN | Citations | PageRank |
0302-9743 | 0 | 0.34 |
References | Authors | |
0 | 3 | |