DWT and HRV Based ECG Biometric Identification for Healthy and Cardiac Patients

Abstract

Rising security concerns in the eld of forensics, law enforcement, surveillance, border control, nancial transactions and access control invokes a stringent security solution. While traditional methods such as ID cards, passport, PIN and token are vulnerable to identity theft, fraud and forgery, biometric technology provide a superior solution to curtail these security risks. Similarly traditional biometrics such as ngerprint, iris and face recognition etc. are also vulnerable with the technological enhancement in criminal attacks. Electrocardiograms provide a superior biometric solution because of its uniqueness, permanence and inherent property of liveness of a person. This research advocate the usefulness of ECG as biometric with comparison of major techniques such as ducial and non ducial approaches. However ECG is a continuous signal which can provide some challenges like variability because of some physical or physiological activity. Various heart diseases like Arrhythmia can also introduce variability in normal rhythm of signal, causing di culties in identi cation process. This research cope with the problems caused by variability of ECG signal by using heart rate variability (HRV) artifact correction. This research proposed Discrete wavelet transform (DWT) and HRV based features separately and with the hybridized approach. The methodology involves preprocessing of ECG signal to remove noise and artifacts, feature extraction using ducial, non- ducial and HRV based features, feature selection using best rst search algorithm and classi cation comparison by using K Nearest neighbor (KNN) and Random forests. To emphasize the solution for real world applications and various deployment scenarios, experiments were performed on three publicly available ECG databases containing MITDB (Arrhythmia), NSRDB (Normal sinus rhythm database) and ECGID database. The identi cation accuracy obtained using these databases surpass previous state of the art approaches using only single lead of ECG data. This research advocate for simplest DWT function known as Haar transform while analyzing various families of DWT such as Biorthogonal, Symlets and Coi ets etc. Similarly, multiresolution analysis of Haar wavelet transiiform was also performed to extract more discriminatory information from ECG signals. Experiments were performed on publicly available database containing cardiac anomalies and an accuracy of 95.85% was achieved with false acceptance rate (FAR) of 4.15% and false rejection rate (FRR) of 0.1%. System is also tested on normal population based databases and accuracy of 100% is achieved using NSRDB database and 83.88% for a challenging ECG-ID database. Similarly best results were evaluated using Haar wavelet of 5th level decomposition coe cients providing segmentation window size of cardiac cycle to 95 samples.