[Research on emotion recognition methods based on multi-modal physiological signal feature fusion]

Area of Science:

• Affective computing and cognitive science.
• Neuroscience and signal processing.

Context:

• Emotion recognition from physiological signals is challenging due to single-modal limitations.
• Accurate emotion detection is vital for human-computer interaction and mental health monitoring.
• Physiological signals like EEG, EMG, and EDA offer objective emotional state indicators.

Purpose:

• To investigate the effectiveness of feature-weighted fusion for integrating multi-modal physiological signals (EEG, EMG, EDA) for emotion classification.
• To compare classification performance using Support Vector Machine (SVM) and Extreme Learning Machine (ELM) with fused signals versus single-modal EEG.

Summary:

• EEG, EMG, and EDA signals were collected from participants experiencing happiness, sadness, and fear.
• A feature-weighted fusion method combined these signals, achieving peak accuracy with weights EEG:0.7, EMG:0.15, EDA:0.15.
• SVM and ELM achieved 80.19% and 82.48% accuracy, respectively, outperforming EEG-alone classification.

Impact:

• Demonstrates significant accuracy improvements in emotion recognition through multi-modal signal fusion.
• Provides a robust methodological framework for utilizing combined physiological data in affective computing.
• Enhances the potential for developing more accurate and reliable emotion classification systems.