The Impact of Segment Length on EEG Based Biometric System

The biometric system is one of the best systems that can be used to replace the password system. Several modalities, such as fingerprints, face recognition, palm prints, and voice recognition, are reported to perform well as biometric modalities. Therefore, the current biometric system is still facing some issues. Artificial biometric data is the main issue in recent biometric systems. The electroencephalogram (EEG) signal is the potential modality to resolve the issue. Many studies have found that the EEG biometric system is able to achieve high accuracy. Most of them require a signal segmentation process to enrich the features and achieve high accuracy. In spite of that, there are only a few studies that focus on the impact of signal segmentation. In this study, 60, 30, 15, 10, 7.5, 6, 5 second lengths of segment were investigated. The power spectral density (PSD) feature from the delta (0.5 – 4 Hz), tetha (4 – 8 Hz), alpha (8 – 14 Hz), beta (14 – 30 Hz), and gamma (30 – 50 Hz) bands were extracted from each segment. It found that there was no linear correction between segment length and accuracy.