Skip to content

Semi-Supervised Graph Contrastive Learning for Emotion Recognition Based on Electroencephalogram Signals

Published

Venue: Engineering Applications of Artificial Intelligence
Author role: First author
Year: 2025

Research approach

Combined DE and PSD sample-similarity graphs with classification and contrastive objectives under limited labels.

SGCL model architecture — SGCL model architecture
View full-size figure

SGCL embedding comparison — SGCL embedding comparison
View full-size figure

Results

SEED-IV Case 1 accuracy: 84.61 ± 4.20%, with 2.4% labels in subject-wise transductive evaluation.

All publications