A User Identification System based on Code-modulated Visual Evoked Potentials with LED Stimulation

Brain-Computer Interfaces are being extended for security purposes considering that electroencephalographic signals can provide unique digital signatures that identify a person. In this paper, we research the use of Code-modulated Visual Evoked Potentials (C-VEPs) to build-up a user identification system. C-VEPs result from visual stimuli modulated by pseudorandom binary sequences. We compare several approaches combining different signal normalizations, detection methods, number of channels, m-sequences, and stimulation time, assessing their impact on user identification accuracy. Using our own LED stimulation framework, the best-case scenario reached an accuracy of 96%, in a 10-user database, using 8 channels of the visual cortex. Using a 17-subject C-VEP public dataset with LCD stimulation, the best-case scenario reached 100% with 32 channels. In both cases the best detection method was the Task-Related Correlation Analysis.