Ecg Pattern Analysis And Classification For Human Recognition

Abstract

The proliferation of digital technologies has changed our working methods and facilitated newlinethe adoption of innovative human recognition approaches. Biometric systems newlineinvolve the methods and techniques using unique physiological or behavioural modalities newlinefor human recognition. The concerns of physiological and behavioural biometrics systems newlineare to make them secure and protect them from spoof attacks. Therefore, securing newlinethe biometric system from spoof attacks is a challenge. The researchers are exploring newlinestate-of-the-art biometrics that is more secured and possibly free from fraudulent attacks. newlineThe researchers have explored the feasibility of using biosignals as biometrics for secure newlineand robust human recognition in the past. Especially, the electrocardiogram (ECG) newlineis a recording of heart electrical activity representing the repetitive patterns of heartbeats. newlineThis dissertation advocates the use of ECG as biometrics for secured identity proofing. newlineThe ECG has the inherent property of vitality detection that ensures that a biometric sample newlineis being acquired from a legitimate and live individual. Thus, it improves the reliability newlineand robustness of the biometric system that enables reluctance against fraudulent attacks. newlineTherefore, ECG signal as biometrics is sufficiently non-vulnerable to spoof attacks and newlineensures the robustness of the system. newlineThis dissertation work is divided into two parts. In the first part, ECG patterns are newlinestudied and investigated for their feasibility as biometrics for human recognition. We newlinework on a non-fiducial method of ECG biometrics that uses autocorrelation and applying newlinetransformation techniques i.e., discrete cosine transform (DCT), discrete Fourier transform newline(DFT), andWalsh-Hadamard transform (WHT). The reason for employing different newlinetransformation techniques is to measure clear discrimination among the subjects. The newlineeffectiveness of these transformations is evaluated on the dimensionality reduction techniques newlinei.e., principal component analysis and linear discriminant analysis. The method newlinef