Selection of Neural Oscillatory Features for Human Stress Classification with Single Channel EEG Headset

Area of Science:

• Neuroscience
• Biomedical Engineering
• Psychology

Background:

• Psychological stress impacts health and requires objective measurement.
• Electroencephalography (EEG) offers a non-invasive method for brain activity monitoring.
• Accurate stress classification is crucial for timely intervention and mental health management.

Purpose of the Study:

• To develop and validate a method for classifying psychological stress using electroencephalography (EEG).
• To identify significant EEG oscillations associated with stress.
• To improve the efficiency and accuracy of stress classification models.

Main Methods:

• Utilized a correlation-based feature subset selection method to reduce EEG data dimensionality.
• Recruited 28 participants who completed the Perceived Stress Scale-10 (PSS-10) questionnaire.
• Recorded EEG data in a closed-eye condition and labeled it based on PSS-10 scores.

Main Results:

• Identified low beta, high beta, and low gamma EEG oscillations as most significant for stress classification.
• The feature selection method significantly reduced the feature vector length.
• Achieved a classification accuracy of 78.57% using a single-channel wearable EEG device.

Conclusions:

• The proposed method efficiently classifies psychological stress using EEG.
• Low beta, high beta, and low gamma oscillations are key indicators of stress.
• Wearable EEG devices coupled with advanced feature selection offer a promising approach for stress monitoring.