Convolutional channel modulator for transformer and LSTM networks in EEG-based emotion recognition

Area of Science:

• Neuroscience
• Artificial Intelligence
• Machine Learning

Background:

• Electroencephalogram (EEG) signals are crucial for understanding intrinsic emotions.
• Variations in EEG signals across sessions pose challenges for temporal relationship modeling.
• Accurate emotion recognition from EEG requires robust feature extraction and temporal analysis.

Purpose of the Study:

• To develop a novel model for accurate emotion recognition from EEG signals.
• To address the challenge of session variability in EEG data.
• To enhance temporal sequence modeling for affective computing.

Main Methods:

• Proposed a feature re-weighting mechanism for temporal sequence modeling.
• Introduced the Convolutional Channel Modulator for Transformer and LSTM networks (CCMTL) model.
• Utilized convolution operations for inter-channel correlation extraction and a channel attention map to emphasize important features.
• Integrated sequential temporal modeling for global and sequential context understanding.

Main Results:

• The CCMTL model demonstrated superior performance in emotion recognition tasks.
• Outperformed six state-of-the-art models on public EEG emotion datasets.
• The feature re-weighting mechanism effectively disentangled invariant features for improved temporal modeling.
• Channel attention successfully focused subsequent models on critical affective features.

Conclusions:

• The proposed CCMTL model offers a significant advancement in EEG-based emotion recognition.
• The novel approach effectively handles session variability and enhances temporal dependency learning.
• CCMTL provides a robust framework for analyzing affective states from complex EEG data.