Study of acceleration plethysmogram based biometric identification incorporating different time instances

This study investigates the effectiveness of acceleration plethysmogram (APG) to be applied as a biometric identification system in different time instances. Currently, most of the study actively discusses on the ability of photoplethysmogram (PPG) for person identification. To the best of our knowledge, little has been said on related studies on APG signals. A total of 5 PPG signals were collected from a publicly available online repository, which is MIMIC II Waveform Database, version 3, part 3 for two different periods and then undergoes preprocessing using a low pass filter. After that, the signals were segmented and later differentiated to produce APG signals. Lastly, the APG signals were classified using four different types of classifiers, namely, Naïve Bayes, Bayes Network, Multilayer Perceptron (MLP) and Radial Basis Function (RBF). Based on the experimentation results, the accuracy for all classifiers increase when applying APG as a biometric modality of up to 11.72% as compared to PPG signals