Development of an Acceleration Plethysmogram based Cardioid Graph Biometric Identification

Abstract

The increasing identity theft cases are alarming which puts biometric as the alternative solution to combat identity crime. Recently, biosignals are proposed as biometric modalities. Thus, in this study, the development of an Acceleration Plethysmogram (APG) based Cardioid graph biometric identification is presented. A total of 10 Photoplethysmogram (PPG) data from MIMIC II Waveform Database (MIMIC2WDB) with sampling frequency of 125 Hz were obtained. The datasets are later converted to APG signal by the second order differentiation and preprocessed with Butterworth filter. Then, Cardioid based graph of APG signal was generated. Its centroid and Euclidean distance are calculated. Finally, classification is done by applying these extracted features to Multilayer Perceptron (MLP) and Naive Bayes neural networks classifiers. Our experimentation results show that subject recognition is possible by obtaining classification accuracy of 95% for APG based Cardioid graph for both classifiers while only 85% and 70% for PPG signal in MLP and Naive Bayes classifiers. These outcomes indicate that APG based Cardioid graph biometric identification is a feasible solution to overcome identity fraud.