Multi-scale graph convolutional EEG emotion recognition method driven by dynamic channel state labels

doi:10.1016/j.bspc.2026.109515

Multi-scale graph convolutional EEG emotion recognition method driven by dynamic channel state labels

2026-01-15

0

PRE

AI

DOI：10.1016/j.bspc.2026.109515

Original

Original request for help

Save

Abstract

En 中文

• Proposes GCPONet to dynamically channel activate label (active/stable/sluggish) for emotion recognition. • Introduces neuroscience-guided brain region coarsening to optimize functional connectivity modeling. • Designs MERNet with multi-scale fusion (local, regional, global) to enhance EEG feature extraction. • Achieves 95.81% accuracy on SEED dataset, outperforming state-of-the-art emotion recognition methods. • Validates cross-subject adaptability with 87.41% accuracy and reduced feature over-smoothing.

Multi-scale graph convolutional EEG emotion recognition method driven by dynamic channel state labels

Abstract

Journal

Organization