Leave-one-out Authentication ofPersonsUsing

Ithasbeenshownpreviously thatrecognizing persons using 40Hzelectroencephalogram (EEG)oscillations ispossible. Inthemethod, features werecomputed fromthe Visual EvokedPotential (VEP)signals recorded from61 electrodes whilesubjects perceived a picture. Here,two modifications have been proposedto improvethe classification performance: Principal ComponentAnalysis (PCA)toreduce thenoise andbackground EEG effects from theVEP signals andnormalization. Two classifiers were used:Simplified FuzzyARTMAP (SFA),andk-Nearest Neighbor (kNN). Theexperimental results using800VEP signals from20subjects withleave-one-out crossvalidation strategy showedthatPCA andnormalization improved the classification performance forboththeclassifiers. Thebest classification performance of95.25%obtained usingthe improved methodshowsthat40Hz EEG oscillations are suitable foruseasbiometrics.