Robust electroencephalogram channel set for person authentication

Abstract

In electroencephalogram (EEG) based biometrics, the determination of the right channel set helps improve accuracy and usability, while reducing the required number of electrodes and hence the complexity and cost of the EEG system. In this work we find a reduced set of channels designed to enhance human authentication accuracy regardless of changes in the mental task. The study shows that the resulting eight EEG channels outperform previous state of the art studies. Also the experiments and quantitative comparison are conducted in a database significantly larger (106 subjects) than the ones used previously. The suggested set half total error rate (HTER) is 14.69%.