Muhammad Afif Hendrawan, Ulla Delfana Rosiani, Arwin Datumaya Wahyudi Sumari
The biometric system is one of the great options for replacing the passworded system. Several biometric modalities, such as fingerprints, image-based recognition, retina, and voice, have been described as performing effectively. Unfortunately, the existing biometric system still faces challenges, such as the tolerance for fabricated data. The electroencephalogram (EEG) signal is a possible resolution modality that is prospective for a biometric system. Numerous investigations have determined that the EEG biometric system can attain high levels of accuracy. However, EEG-based biometric systems also have issues in practical implementation. Most proposed ones require a multi-channel sensor configuration that gives rise the complexity in the data collection and computation. In this study, the optimum channel for biometric systems was investigated. From the experiment, the proposed method could reach an accuracy of 80% by using single-channel sensor's EEG data. This result came from the segmentation of the EEG data into five-second long for each segment. The Power Spectral Density (PSD) feature was then extracted from delta (0.5-4Hz), theta (4-8Hz), alpha (8-14Hz), beta (14-30Hz) and gamma (30-50Hz) bands. The extracted data was then classified by using K-Nearest Neighbours (KNN) and linear Support Vector Machine (SVM) algorithms © 2022 IEEE.
Politeknik Negeri Malang, Department of Information Technology, Malang, Indonesia