Bridging Neuroscience and Machine Learning: A Gender-Based Electroencephalogram Framework for Guilt Emotion...

Area of Science:

• Neuroscience
• Psychology
• Machine Learning

Background:

• Understanding the neural correlates of emotions like guilt is crucial.
• Gender differences in emotional processing are well-documented but require further investigation using neurophysiological data.
• Developing accurate methods for classifying emotional states from brain activity is an active research area.

Purpose of the Study:

• To investigate the relationship between guilt emotion and human electroencephalography (EEG) data.
• To explore gender-specific differences in the expression of guilt and neutral emotions.
• To evaluate the effectiveness of machine learning approaches for classifying emotional states from EEG.

Main Methods:

• EEG data were collected from 16 participants experiencing induced guilt and neutral emotions via pictorial stimuli.
• Data pre-processing involved bandpass filtering and Independent Component Analysis (ICA).
• Support Vector Machine (SVM) classifiers were employed, utilizing both raw EEG data and features extracted via Discrete Wavelet Decomposition (DWT) across various frequency bands (Alpha, Beta, Gamma, Theta, Delta).

Main Results:

• A novel approach feeding raw EEG data to SVM achieved an average accuracy of 83%.
• A feature-based approach using DWT-extracted features (entropy, Hjorth parameters, Band Power) resulted in an average accuracy of 63%.
• Females exhibited more expressive EEG responses and higher feature values compared to males, with higher classification accuracy across most frequency bands.

Conclusions:

• Machine learning, particularly SVM with raw EEG data, can effectively classify emotional states.
• Significant gender differences exist in the neural processing of guilt and neutral emotions, with females showing heightened responses.
• Further research can refine these methods for more nuanced emotional state detection.