Biometric System using PCG Signal Analysis: A New Method of Person Identification

Biometrics are considered as one of the most authentic methods for human identification. In this paper a novel method is introduced which uses heart sound to identify humans. Traditional biometric modalities are not potent against any falsifiability so Phonocardiogram (PCG) signals are a different and new way of identifying individuals. Dataset was collected from 30 individuals using BSL instruments. First, PCG signals were preprocessed to remove noise and other artifacts using Empirical Mode Decomposition (EMD) then eleven features, which gave the best intra-class difference, were selected through extensive experimentation using Support Vector Machine (SVM). A series of tests were performed on data set of 1508 signals each of 10 seconds and a very encouraging accuracy of 95.4% was achieved using Quadratic kernel for SVM.