Time-Series Forecasting: Extreme Gradient Boosting Implementation in Smartphone Photoplethysmography Signals for Biometric Authentication Processes

Abstract

Biometric Authentication (BA) is a process where behavioral and physiological inputs are used to determine the identity of individuals. Photoplethysmogram (PPG) is commonly used to provide physiological information of patients, such as heart rate and breathing rate. With technological advances, smartphones can provide PPG information without any external hardware. In this paper, we propose a BA system based on PPG readings. Features were selected by considering possible unique physiological factors during the period when PPG signals are acquired. We adopted the eXtreme Gradient Boosting (XGBoost) algorithm as a classification model. As performance metrics, we considered accuracy, specificity, and equal error rate (EER). Experimental results show that the average training accuracy, specificity, and EER values are 97.36%, 99.94%, and 0.06%, respectively, while the average testing accuracy, specificity, and EER values are 96.38%, 99.57%, and 0.43%, respectively.