DW-FBCSP: EEG emotion recognition algorithm based on scale distance weighted optimization.

Area of Science:

• Neuroscience
• Machine Learning
• Affective Computing

Background:

• Emotion calibration using valence-arousal scales faces limitations in accurately measuring stimulus influence due to simplistic high/low score divisions.
• Existing methods struggle with precise classification and labeling, hindering accurate assessment of emotional responses.

Purpose of the Study:

• To propose a novel Electroencephalography (EEG) emotion recognition algorithm, Distance Weighted Filter Bank Common Spatial Pattern (DW-FBCSP), to overcome limitations in current emotion calibration methods.
• To enhance the accuracy of measuring the influence of emotion stimulation materials on subjects.
• To optimize EEG signal projection matrices for improved emotion classification.

Main Methods:

• Developed the DW-FBCSP algorithm, an extension of Common Spatial Pattern (CSP), incorporating scale distance weighted optimization.
• Optimized classification by considering the distance of scores from an ideal center.
• Fused extracted features with selected features using a Linear Discriminant Analysis (LDA) classifier for emotion recognition.

Main Results:

• Achieved a mean correct rate of 81.14% for valence and 84.45% for arousal using the DEAP dataset.
• Demonstrated superior performance compared to recently published emotion recognition methods.
• The DW-FBCSP algorithm effectively improved the accuracy of EEG-based emotion recognition.

Conclusions:

• The proposed DW-FBCSP algorithm offers a significant advancement in EEG-based emotion recognition.
• The scale distance weighted optimization effectively addresses the classification and labeling defects of traditional methods.
• This approach provides a more accurate and reliable way to measure emotional responses to various stimuli.