Correlation Analysis of Features for Fusing in User Verification Using EEG Evoked by Ultrasound

In user verification using electroencephalograms (EEGs) evoked by ultrasound, an error rate of 0% was achieved. However, to achieve this, the classifiers for the number of features multiplied by the number of electrodes must be learned. Therefore, reducing the number of classifiers is crucial and must be achieved. This study confirmed that the random selection of features and electrodes facilitates further reduction in the number of classifiers. Random selection is equivalent to evenly selecting electrodes for each feature and electrode position. Consequently, the effectiveness of even selection was statistically confirmed. Furthermore, even selection resulted in the fusion of uncorrelated features. Thus, four statistical values of an EEG were introduced, and the effectiveness of fusing uncorrelated (independent) features was confirmed.