Related Experiment Video
Updated: May 25, 2026
05:51
Exploring the Use of Isolated Expressions and Film Clips to Evaluate Emotion Recognition by People with Traumatic Brain Injury
Published on: May 15, 2016
9.1K
Hybrid transfer learning strategy for cross-subject EEG emotion recognition
Wei Lu1,2, Haiyan Liu3, Hua Ma1
1Henan High-speed Railway Operation and Maintenance Engineering Research Center, Zhengzhou Railway Vocational and Technical College, Zhengzhou, China.
Frontiers in Human Neuroscience
|November 30, 2023
Summary
This study introduces a novel hybrid transfer learning strategy for more accurate cross-subject emotion recognition using electroencephalogram (EEG) signals. The DFF-Net model significantly improves performance by addressing inter-individual differences.
Area of Science:
- Affective computing
- Machine learning
- Neuroscience
Background:
- Emotion recognition is crucial for various applications.
- Deep learning with EEG signals shows promise for emotion recognition.
- Cross-subject emotion recognition faces challenges due to individual differences.
Purpose of the Study:
- To develop an effective method for cross-subject EEG emotion recognition.
- To address the performance drop caused by inter-individual differences in EEG data.
Main Methods:
- Proposed a hybrid transfer learning strategy.
- Designed the Domain Adaptation with a Few-shot Fine-tuning Network (DFF-Net).
- Introduced the Emo-DA module for domain adaptive learning and pre-training.
Main Results:
- DFF-Net achieved 93.37% accuracy on the SEED dataset.
- DFF-Net achieved 82.32% accuracy on the SEED-IV dataset.
- The proposed method surpasses state-of-the-art in cross-subject EEG emotion recognition.
Conclusions:
- The DFF-Net effectively improves cross-subject EEG emotion recognition accuracy.
- The hybrid transfer learning approach successfully mitigates inter-individual differences.
- This work offers a significant advancement in affective computing research.
