Related Experiment Video
Updated: Jan 16, 2026
11:28
Concurrent EEG and Functional MRI Recording and Integration Analysis for Dynamic Cortical Activity Imaging
Published on: June 30, 2018
12.2K
Mtfsfn: a multi-view time-frequency-space fusion network for EEG-based emotion recognition
Zhongmin Wang1,2,3, Shengyang Gao1
1School of Computer Science and Technology, Xi'an University of Posts and Telecommunications, Xi'an, 710121 Shaanxi China.
Cognitive Neurodynamics
|September 30, 2025
Summary
This study introduces a novel Multi-view Time-Frequency-Space Fusion Network (MTFSFN) for advanced electroencephalogram (EEG) emotion recognition. The MTFSFN model significantly improves emotion classification accuracy by effectively fusing spatiotemporal features.
Area of Science:
- Neuroscience
- Artificial Intelligence
- Signal Processing
Background:
- Emotion recognition using electroencephalogram (EEG) is a growing field.
- EEG signals possess complex non-stationary and spatially discrete characteristics.
- Extracting discriminative spatiotemporal features from EEG remains a challenge.
Purpose of the Study:
- To propose a novel Multi-view Time-Frequency-Space Fusion Network (MTFSFN) for enhanced EEG-based emotion recognition.
- To address the limitations of current methods in capturing complex EEG signal dynamics.
- To improve the accuracy and robustness of emotion classification.
Main Methods:
- Developed a Multi-view Time-Frequency-Space Fusion Network (MTFSFN).
- Employed a frequency-domain attention mechanism to weight features across different frequency bands.
- Utilized a multi-view Transformer with 2D positional embeddings for spatial information extraction.
- Integrated LSTM to capture dynamic time-frequency-space relationships.
- Validated the model using a subject-independent leave-one-subject-out cross-validation on DEAP, SEED, and SEED-IV datasets.
Main Results:
- Achieved high average accuracies: 78.64% (valence) and 77.42% (arousal) on DEAP.
- Reached 86.91% average accuracy on the SEED dataset.
- Obtained 75.51% average accuracy on the SEED-IV dataset.
- Demonstrated excellent recognition performance of the MTFSFN model.
Conclusions:
- The proposed MTFSFN effectively fuses multi-view time-frequency-space features for superior EEG emotion recognition.
- The model's architecture successfully addresses the challenges posed by EEG signal complexity.
- MTFSFN shows significant potential for real-world emotion recognition applications.
