Deep Neural Network-based Person Identification using ECG Signals

Abstract

In recent times, biometrics is mostly utilized for the authentication or identification of a user for a vast civilian application. Most of the electronic systems have been proposed that employed distinct behavioral or physiological human beings signature for identifying or verifying the user in an automatic manner. Nowadays, Electro Cardio Gram (ECG)-oriented biometric systems are in the exploration stage. The behavior of the ECG signal is distinctive to every person. As ECG is an exclusive physiological signal that is present only in the live people, it is utilized in the new biometric systems for recognizing the people and to counter the fraud as well as the forge attacks. Majority of the traditional techniques limits from the restriction in several points detection in the ECG signal. The contribution of this paper is the enhancement of the novel structure of person identification model by ECG signal. At first, the ECG signal collected from the three benchmark source is subjected for pre-processing, in which the noise is removed by Low Pass Filter (LPF) approach. Further, the Empirical Mode Decomposition (EMD) is adopted for the decomposition of signal. As feature selection is the significant part of classification enhancement, Principle Component Analysis (PCA) is used as the effective feature extraction that takes the most important features from the signal. Finally, the adoption of Deep Neural Network (DNN) is performed as the deep learning model that could identify the exact person from the given ECG signal. The effectiveness of the method is extensively validated on benchmark datasets and retrieves the outcome.