Integrating EMD attributes for person identification from electrocardiographic signals

Presently biometric applications have been widely employed for security needs. They enable to provide a higher security level concerning physical or logical access. The electrocardiogram (ECG) signal can be exploited as a biological tool for human identification given that it is robust to falsification and more universal. This research introduces a new method applying this ECG particularity. Five fiducial points are extracted locally of each heartbeat using empirical mode decomposition (EMD). Then 14 parameters based on amplitude, temporal and RR intervals between extracted fiducials points are computed. Specifically, Hidden Markov model (HMM) classifier with Bakis model is adopted for human recognition. The approach achieves 98.52% accuracy on a 44 subjects from MIT-BIH Arrhythmia database by combining all features which allows providing a notable improvement in terms of identification.