An Independent Component Analysis (ICA) Based Approach for EEG Person Authentication

Exploring brain electrical activity represented by electroencephalogram (EEG) signals for biometric applications has recently attracted increasing research attention since EEG pattern has been shown to be unique for each individual. In this paper, we propose an Independent Component Analysis (ICA) based EEG feature extraction and modeling approach for person authentication. Five dominating Independent Components (DIC) are determined from five brain regions represented by EEG channels, then univariate autoregressive coefficients of DICs are extract as features. Based on AR coefficients of DICs, a Naive Bayes probabilistic model is employed for person authentication purpose. Results from a real EEG motor task study suggest that the proposed ICA-based approach is promising and may open new directions in the emerging EEG biometry area.