Research on negative emotion recognition in EEG signals based on IGC-CNN-CBAM

Negative emotions (such as anxiety, depression, anger, etc.) are critical manifestations of mental health issues, significantly impacting individuals' quality of life and social functioning. In recent years, research on negative emotions has become a hotspot in the field of emotion-based brain-computer interfaces. Traditional machine learning methods like SVM face challenges in handling large datasets, multi-class classification, and interpretability, with substantial room for improvement in classification accuracy. This study proposes an IGC-CNN-CBAM classification algorithm that significantly enhances final classification accuracy. The researcher collected the EEG signals under four emotion-induced materials: low negative, high negative, positive and neutral (the division of the four emotions is determined by the two-dimensional emotion quantitative model), extracted two EEG features (PSD, and differential entropy), and performed feature fusion on these two features(The horizontal concatenation method is adopted). The fused data was then classified using both traditional SVM and the proposed IGC-CNN-CBAM algorithms. Experimental results demonstrated a classification accuracy of 74.5% with SVM, while the IGC-CNN-CBAM algorithm achieved accuracy of 84.4%—a 9.9% significant improvement( paired t-test, P < 0.01). These findings indicate that the IGC-CNN-CBAM algorithm is more effective for EEG-based negative emotion recognition.