Investigating the impacts of brain conditions on EEG-based person identification

Abstract

Person identification using electroencephalogram (EEG) as biometric has been widely used since it is capable of achieving high identification rate. Brain conditions such as epilepsy and alcohol are some of problems that cause brain disorders in EEG signals, and hence they may have impacts on EEG-based person identification systems. However, this issue has not been investigated. In this paper, we perform person identification on two datasets, Australian and Alcoholism EEG, then compare the classification rates between epileptic and non-epileptic groups, and between alcoholic and non-alcoholic groups, to investigate the impacts of such brain conditions on the identification rates. Shannon (SEn), Spectral (SpEn), Approximate entropy (ApEn), Sample (SampEn) and Conditional (CEn) entropy are employed to extract features from these two datasets. Experimental results show that both epilepsy and alcohol actually have different impacts depending on feature extraction method used in the system.