Emotion recognition based on the sample entropy of EEG

Area of Science:

• Neuroscience
• Machine Learning
• Affective Computing

Background:

• Emotion recognition from electroencephalography (EEG) signals is crucial for understanding human affective states.
• Developing accurate and generalizable algorithms for emotion classification remains a significant challenge.

Purpose of the Study:

• To present a novel emotion recognition approach based on sample entropy (SampEn) and support vector machine (SVM) classification.
• To evaluate the algorithm's performance in distinguishing emotions with varying arousal levels.

Main Methods:

• Sample entropy (SampEn) was calculated for significant electroencephalography (EEG) channels identified via the Kolmogorov-Smirnov (K-S) test.
• The SampEn features were used to train a support vector machine (SVM)-weight classifier.
• The algorithm was tested on two emotion recognition tasks: positive/negative emotion with high arousal, and emotions with different arousal statuses.

Main Results:

• Key EEG channels for emotion recognition were identified in the prefrontal region (F3, CP5, FP2, FZ, FC2).
• The proposed algorithm achieved accuracies of 80.43% and 79.11% for the two emotion recognition tasks, respectively.
• Leave-one-person-out validation demonstrated the algorithm's reasonable generalization capability.

Conclusions:

• The SampEn-based EEG emotion recognition method shows promising accuracy and generalization.
• The identified prefrontal EEG channels are important for affective state classification.
• This approach offers a viable tool for objective emotion recognition.