ECG Biometric using Statistical Feature of EEMD and VMD

Abstract

Electrocardiogram (ECG), as a biometric that has been widely studied, has advantages that are difficult to fake compared to biometrics using physical characteristics. This study simulated an ECG based biometric system with 15 subjects. It used the Butterworth low pass filter (LPF), ensemble empirical mode decomposition (EEMD) or variational mode decomposition (VMD), and statistical features as feature extraction method. The filtered signal will be segmented, and the subsequent five level decomposition using EEMD and VMD. Then, the signal analysis used the statistical feature approach for each intrinsic mode function (IMF) as result of decomposition process. These values become a feature set entered of K-Nearest Neighbor (KNN) as classifier; the highest result of 93% was achieved using VMD and KNN with Manhattan distance.