Akademik Veri Yönetim Sistemi
Journal of the Faculty of Engineering and Architecture of Gazi University, cilt.41, sa.1, ss.367-382, 2026 (SCI-Expanded, Scopus, TRDizin)
With the growing interest in artificial intelligence in clinical applications, electroencephalography (EEG) has become an increasingly researched technique in biometric-based security systems due to its ability to capture individual-specific patterns and provide higher resistance to spoofing compared to other biometric modalities. In this study, a low-cost approach was targeted by utilizing an EEG dataset acquired from 109 participants using 64 channels within a closed system. The performance of an ensemble machine learning-based single-channel identity recognition approach was investigated using EEG data from a randomly selected limited number of subjects. Using the Gini importance coefficient employed in decision tree-based algorithms, the most discriminative channel was selected from datasets recorded under both eyes-open and eyes-closed conditions. Single-channel machine learning models trained with features extracted from the selected channel were compared with 64-channel models in terms of training time, testing time, and classification performance. In the system consisting of preprocessing, feature extraction, and classification stages, the delta frequency band of the Cp4 channel was used for eyes-closed data, and classification was performed using an ensemble model comprising random forests, multilayer perceptron, gradient boosting, decision trees, and support vector machine algorithms. Cross-validation results achieved 100% classification accuracy, while significant reductions in training and testing times were obtained compared to 64-channel systems. These findings demonstrate that a reliable single-channel EEG-based biometric identification system is feasible even with a reduced number of channels.