Assessment of the potential of morphological ECG features for person identification

This study investigates the potential of ECG morphological feature set for person identification. The measurements are done over 145 pairs of ECG recordings from healthy subjects, acquired 5 years apart. Time, amplitude, area and slope descriptors of the QRS-T pattern are analyzed in 4 ECG leads, forming quasi-orthogonal lead system (II&III, V1, V5). The inter-subject variation, the difference of means in 1^st vs. 2^nd recording measurements, as well as the cross-correlation between features are estimated. Thus, 2 area and 4 amplitude descriptors of the QRS complex are highlighted. The population heterogeneity in the space of the selected features is verified via Factor analysis by Principal components extraction method. It confirms the orthogonality of the 6 features (each of them has significant factor loading for a particular factor). The analysis shows that the first 3 factors have eigenvalues higher than 1, both for the measurements in the 1^st and the 2^nd ECG recording and they accumulate respectively 68% and 64 % of the total data variation, which is a sign for their person identification potential.