Emotion Recognition from Multiband EEG Signals Using CapsNet.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Signal Processing

Background:

• Emotion recognition using electroencephalograph (EEG) signals is a growing field.
• Conventional methods often overlook the crucial spatial characteristics present in EEG data.
• These spatial features contain significant information for understanding emotional states.

Purpose of the Study:

• To propose a deep learning framework that integrates spatial and frequency characteristics of multi-channel EEG signals.
• To enhance emotion recognition accuracy by leveraging a multiband feature matrix (MFM) and a capsule network (CapsNet).

Main Methods:

• A multiband feature matrix (MFM) was constructed by combining frequency domain, spatial, and frequency band characteristics of multi-channel EEG signals.
• A capsule network (CapsNet) model was employed to process the MFM for emotion state recognition.
• Experiments were conducted using the DEAP dataset.

Main Results:

• The proposed MFM and CapsNet framework demonstrated superior performance compared to existing models on the DEAP dataset.
• Experimental results confirmed the complementary nature of the three integrated characteristics (frequency, spatial, band).
• The capsule network proved effective in mining and utilizing these correlated characteristics.

Conclusions:

• The developed deep learning framework offers a promising approach for accurate emotion recognition from EEG signals.
• Integrating spatial and frequency domain features via MFM enhances the robustness of emotion detection.
• CapsNet is well-suited for capturing complex relationships within EEG data for affective computing.