Related Experiment Video
Updated: Jan 13, 2026
08:31
Conscious and Non-conscious Representations of Emotional Faces in Asperger's Syndrome
Published on: July 31, 2016
14.4K
Cross-Group EEG Emotion Recognition Based on Phase Space Reconstruction Topology
Xuanpeng Zhu1, Mu Zhu1, Dong Li1
1Tianjin Key Laboratory for Control Theory and Applications in Complicated Systems, School of Electrical Engineering and Automation, Tianjin University of Technology, Tianjin 300384, China.
Entropy (Basel, Switzerland)
|October 28, 2025
Summary
This study introduces novel topological features for electroencephalogram (EEG) emotion recognition, achieving high accuracy in distinguishing emotions for both normal-hearing and hearing-impaired individuals.
Area of Science:
- Neuroscience
- Signal Processing
- Machine Learning
Background:
- Electroencephalogram (EEG) signal nonlinearity and artifacts pose challenges for accurate emotion recognition.
- Extracting meaningful features from complex EEG data requires advanced techniques.
- Understanding emotional states through brain activity is crucial for various applications.
Purpose of the Study:
- To develop a novel method for EEG emotion recognition by extracting topological features.
- To address the challenges of signal nonlinearity and artifacts in EEG analysis.
- To improve the accuracy and robustness of emotion recognition across different subject groups.
Main Methods:
- Applied Local Linear Embedding (LLE) to reduce phase space trajectory dimensionality, preserving local topological structure.
- Constructed 16 novel topological features to describe nonlinear dynamic patterns of emotions at multiple scales.
- Utilized independent feature evaluation and brain topography analysis to select optimal features and electrode channels.
Main Results:
- Achieved subject-dependent average accuracies of 90.33% (3-Class) for normal-hearing and 77.17% (4-Class) for hearing-impaired subjects on SEED and HIED datasets.
- Demonstrated a 77.5% average accuracy for a 6-Class emotion recognition task across different subject groups by integrating topological features.
- Showcased the effectiveness of topological features in capturing emotion-induced nonlinear dynamics in EEG signals.
Conclusions:
- The proposed topological features offer a robust and effective approach for EEG-based emotion recognition.
- The method shows promise for applications involving both normal-hearing and hearing-impaired populations.
- Further integration of topological features with other signal processing techniques can enhance emotion recognition capabilities.
