Multi-objective optimization for EEG channel selection and accurate intruder detection in an EEG-based subject...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computer Science

Background:

• Electroencephalography (EEG) is a non-invasive neuroimaging technique.
• EEG signal analysis is crucial for brain-computer interfaces and biometric systems.
• Optimal channel selection enhances the efficiency and accuracy of EEG-based applications.

Purpose of the Study:

• To develop and evaluate a four-objective optimization method for selecting optimal electroencephalographic (EEG) channels.
• To enable subject identification and intruder detection systems using minimal EEG channels.
• To improve the accuracy and reliability of EEG-based authentication systems.

Main Methods:

• Feature extraction using empirical mode decomposition (EMD) from EEG sub-bands.
• Support vector machines (SVMs) for classification of subjects and intruders.
• Non-dominated sorting genetic algorithm (NSGA) for multi-objective channel selection.
• Validation on event-related potentials (ERPs) data from 26 subjects and 56 channels.

Main Results:

• A three-channel combination achieved 0.83 accuracy, 1.00 true acceptance rate (TAR), and 1.00 true rejection rate (TRR).
• Seven EEG channels identified by NSGA-III yielded up to 0.98 accuracy for subject identification (TAR: 0.95, TRR: 0.93).
• Validation with random subject subdivisions showed high accuracy (0.97 ± 0.02) using eight channels.

Conclusions:

• The proposed four-objective optimization method effectively selects optimal EEG channels for identification and authentication.
• Minimal EEG channel sets can achieve high performance in subject recognition and intruder detection.
• Results support the potential of EEG-based systems for secure identification and authentication.