A Resource-Efficient Multi-Entropy Fusion Method and Its Application for EEG-Based Emotion Recognition.

Area of Science:

• Neuroscience and Biomedical Engineering
• Affective Computing and Human-Computer Interaction

Background:

• Emotion recognition is crucial for mental health and human-computer interaction.
• Electroencephalography (EEG) offers a direct measure of brain activity for emotion analysis.
• Existing methods often face challenges with computational complexity and data dimensionality.

Purpose of the Study:

• To propose a resource-efficient multi-entropy fusion method for classifying emotional states using EEG signals.
• To evaluate the effectiveness of different similarity measures, time windows, and channel input configurations.
• To develop a method suitable for portable, daily-use EEG-based emotion-aware devices.

Main Methods:

• Extracted five brain rhythms (delta, theta, alpha, beta, gamma) from EEG signals using Discrete Wavelet Transform (DWT).
• Acquired multi-entropy features (PSDE, SSE, SE, FE, AE, PE) and fused them into a Brain Rhythm Entropy Matrix (BREM).
• Employed Dynamic Time Warping (DTW), Mutual Information (MI), Spearman Correlation Coefficient (SCC), and Jaccard Similarity Coefficient (JSC) for classification.

Main Results:

• Dynamic Time Warping (DTW) demonstrated the best performance with a 5-second time window.
• Single-channel input outperformed single-region input for classification accuracy.
• The proposed method achieved 84.62% accuracy for arousal and 82.48% for valence classification.

Conclusions:

• The multi-entropy fusion method effectively reduces data dimensionality and computational complexity.
• The method maintains high accuracy (over 80%) even with limited data resources.
• This approach facilitates the development of practical, portable EEG-based emotion-aware devices for everyday applications.