CNN-Based Personal Identification System Using Resting State Electroencephalography.

Area of Science:

• Biometrics
• Neuroscience
• Computer Science

Background:

• Electroencephalography (EEG) signals offer unique biometric traits for secure personal identification due to their inherent difficulty in replication and ease of liveness detection.
• Resting state EEG signals present a practical and convenient protocol for identification compared to other EEG acquisition methods.

Purpose of the Study:

• To propose and evaluate a novel personal identification system leveraging resting state electroencephalography (EEG) signals.
• To enhance the system's robustness and accuracy through data augmentation and advanced neural network architectures.

Main Methods:

• Development of a personal identification system combining data augmentation techniques with a convolutional neural network (CNN).
• Cross-validation was performed on a public EEG database comprising 109 subjects.
• Utilized a minimal configuration of 14 EEG channels and 0.5 seconds of data for evaluation.

Main Results:

• The proposed system achieved an average accuracy of 99.32% and an average equal error rate (EER) of 0.18%.
• The system demonstrated high performance with a short data acquisition time (0.5 seconds) and limited channels (14).

Conclusions:

• The developed EEG-based identification system offers a practical solution with high accuracy and efficiency.
• The system's advantages include short acquisition time, low computational complexity, and suitability for deployment with low-cost EEG sensors.
• This research advances the feasibility of practical, real-world applications for EEG-based personal identification.