Implementation of Kernel Sparse Representation Classifier for ECG Biometric System

Abstract

In this paper, a biometric human recognition system based on Electrocardiography (ECG) signal is proposed. Three processes i.e., pre-processing, feature extraction and classification is discussed. A combination of enhanced start and end point detection namely short time energy (STE) and short time average zero crossing rate (STAZCR) is employed in the pre-processing. Subsequently, an autocorrelation method is applied in feature extraction. For the classification process, the kernel sparse representation classifier (KSRC) is proposed as a classifier to increase the system performance in high dimensional feature space. 79 recorded signals from 79 subjects are used are employed in this study. To validate the performance of the KSRC, several classifiers, i.e. sparse representation classifier (SRC), k nearest neighbor (kNN) and support vector machine (SVM) are compared. An experiment based on different sizes of feature dimensions is conducted. The classification performance for four classifiers are found to be 90.93%, 92.8%, 94.24%, 62.9%, 97.23% and 95.87% for the kNN, SVM (Polynomial and RBF), SRC and KSRC (Polynomial and RBF), respectively. The results reveal that the KSRC is a promising classifier for the ECG biometric system compared to the existing reference classifiers.