A Biometric Authentication Technique Using Smartphone Fingertip Photoplethysmography Signals

Biometric Authentication (BA) is the process by which physiological measurements can be used to identify a specific person. BA has been implemented in various applications such as financial transactions, data privacy, and access control. Photoplethysmography (PPG) signals can be acquired from devices such as smartphones, smartwatches, or web cameras, which allow biometric signals to be readily captured. In this paper, we developed a BA classifier method for smart devices, which first reduces motion and noise artifacts (MNAs) in raw PPG signals, and then identifies subjects using the MNA-reduced PPG signals. Specifically, our BA classifier algorithm adopts an ensemble bagged trees (EBT) classifier with 16 PPG signal features from time and frequency domains. In the BA context, false positive (FP) and false negative (FN) rates are very important measurements to be considered. Acceptable FP and FN values were obtained. Experimental results show that our EBT-based BA algorithm achieved accuracy levels of 98.0% and 95.0%. Also, equal error rates (EER) measurements accomplished acceptable values of 2.42% and 5.90%.