EEG electrode selection for person identification thru a genetic-algorithm method.

Area of Science:

• Biometrics
• Neuroscience
• Signal Processing

Background:

• Electroencephalography (EEG) signals offer unique features for biometric identification, surpassing limitations of common methods.
• Current EEG processing for biometrics is time-consuming and labor-intensive, necessitating efficient algorithms.
• Reducing electrode count and segmenting data into frequency bands are key strategies for EEG processing optimization.

Purpose of the Study:

• To propose a genetic algorithm for minimizing the number of electrodes required for EEG-based biometric identification.
• To determine the optimal subset of electrodes for accurate subject identification across different EEG frequency bands.
• To evaluate the efficacy of the proposed method under varying stimuli conditions (eye-open and eye-closed).

Main Methods:

• Utilized a public EEG dataset comprising 109 subjects.
• Applied a genetic algorithm to systematically reduce the number of electrodes from an initial set of 64.
• Analyzed EEG data segmented into distinct frequency sub-bands for both eye-open and eye-closed conditions.

Main Results:

• Accurate subject identification was achieved using approximately 10 out of 64 electrodes.
• Higher EEG frequency bands required fewer electrodes for effective identification compared to lower frequency bands.
• The genetic algorithm demonstrated significant potential in reducing the computational load for EEG biometrics.

Conclusions:

• A reduced electrode set, guided by a genetic algorithm, is sufficient for accurate EEG-based biometric identification.
• The findings suggest that higher frequency EEG bands are more efficient for identification, requiring fewer sensors.
• This approach offers a practical solution to reduce the processing burden associated with EEG biometric systems.