This thesis presents the concept of representing finite signals in terms of sequential output pulses called pulse domain to extract Electrocardiogram (ECG) features for biometric authentication systems. Two novel methods based on the pulse domain philosophy namely Pulse Active (PA) and Adaptive Pulse Active (APA) techniques are presented in this thesis. A total of 11 algorithms are derived from these two methods and used to generate novel ECG feature vectors. Six algorithms of the PA technique are named as Pulse Active Bit (PAB), Pulse Active Width (PAW), Pulse Active Area (PAA), Pulse Active Mean (PAM), Pulse Active Ratio (PAR) and Pulse Active Harmonic (PAH). Five APA algorithms are named as Adaptive Pulse Active Bit (APAB), Adaptive Pulse Active Width (APAW), Adaptive Pulse Active Area (APAA), Adaptive Pulse Active Mean (APAM) and Adaptive Pulse Active Harmonic (APAH). The proposed techniques are validated using ECG experimental data from 112 subjects. Simulation results indicate that APAW generates the best biometric performance of all 11 algorithms. Selected ranges of PA and APA parameters are determined in this thesis that generates approximate similar biometric performance. Using this suggested range, these parameters are than used as a personal identification number (PIN) which are a part of the proposed PA-APA ECG based multilevel security biometric authentication system.