A Non-Fiducial Noise Robust VMD-based Framework for ECG-based Biometric Recognition

Accurate detection of fiducial points is a challenging task for an electrocardiogram (ECG) signal and failing to achieve the desired features can degrade the performance of the ECG-based biometrics. This paper presents a robust ECG-based biometric framework for user authentication, which does not require the detection of fiducial points. In the proposed method, a data-adaptive variational mode decomposition approach is used for both noise removal and feature extraction from the signal. Five temporal and spectral features extracted from each of the decomposed modes are fed as input to the back propagation neural network based classifier for the classification of genuine and imposter subjects. The performance of the proposed method is evaluated and validated for both clean and noisy ECG signals taken from 43 true subjects and 9 imposters. The proposed method achieves an identification rate of 99.02% and an equal error rate of 0.13 from the identification and verification processes, respectively. The performance results of the proposed method show its strong potential for biometric applications, and thus, it may be employed for security and surveillance purposes.