PCLC

PCLC: Pattern-Conditioned Lifelong Consolidation for EEG emotion recognition.

PCLC is a novel framework for EEG-based emotion recognition that addresses catastrophic forgetting in continual learning scenarios. By discovering latent neural patterns and conditioning both classification and memory consolidation on these patterns, PCLC achieves robust cross-dataset adaptation from SEED-IV to SEED-V with minimal forgetting.

Datasets

SEED_IV: https://bcmi.sjtu.edu.cn/home/seed/seed-iv.html

SEED_V: https://bcmi.sjtu.edu.cn/home/seed/seed-v.html

The code automatically ⚙️

• Loads SEED-IV & SEED-V datasets
• Preprocesses EEG features (DE extraction, robust scaling)
• Trains PCLC on SEED-IV sessions 1→3
• Evaluates cross-dataset generalization (IV→V)
• Performs continual learning on SEED-V
• Generates all figures (PDF) and statistical and robustness analyses

Citation 📝

If using this code, please cite our incoming paper on PCLC that is currently underreview in an IEEE Transaction journal.