Emotional Stress State Detection Using Genetic Algorithm-Based Feature Selection on EEG Signals.

Area of Science:

• Neuroscience
• Machine Learning
• Signal Processing

Background:

• Electroencephalography (EEG) signals are crucial for indirect brain state measurement and stress analysis.
• Current stress detection algorithms often lack effective feature selection, limiting classifier performance.
• Machine learning techniques are increasingly applied to EEG signal analysis for stress identification.

Purpose of the Study:

• To develop an efficient feature selection method for stress analysis using EEG signals.
• To enhance the performance of a k-nearest neighbor (k-NN) classifier for stress detection.
• To evaluate a genetic algorithm (GA)-based feature selection approach for EEG-based stress identification.

Main Methods:

• Utilized a genetic algorithm (GA) for optimal feature subset selection from EEG data.
• Employed a k-nearest neighbor (k-NN) classifier with GA-selected features for stress classification.
• Extracted features including statistical measures, Hjorth parameters, band power, and frontal alpha asymmetry from 32 EEG channels.

Main Results:

• The proposed GA-based feature selection method achieved 71.76% classification accuracy for stress detection.
• The GA-enhanced k-NN classifier significantly outperformed the principle component analysis (PCA) method (65.3% accuracy).
• The study validated performance using the public Database for Emotion Analysis using Physiological Signals (DEAP) dataset.

Conclusions:

• Genetic algorithm-based feature selection effectively enhances k-NN classifier performance for EEG-based stress analysis.
• The proposed method offers a robust and accurate approach for identifying stress states from electroencephalography signals.
• This technique provides a valuable tool for stress monitoring and mental health research.