Verifying the Newborns without Infection Risks Using Contactless Palmprints

Abstract

Verification of new-born babies utilizing the biometric characteristics has received an increased attention, especially in applications such as law enforcement, vaccination tracking, and medical services. In this work, we present an introductory study on exploring contactless palmprint biometric for the verification of new-borns. To the best of our knowledge, this is the first work to explore automatic contactless palmprint verification of new-born babies. We have captured a new database of contactless palmprint images from 50 new-born babies in two different sessions. The first session data is captured between 6-8 hours after the birth and the second session data is captured between 28-36 hours after the birth. Extensive experiments are carried out using seven different state-of-the-art palmprint algorithms to benchmark both left and right contactless palmprint characteristics captured from the new-born babies. We further propose a new method based on transfer learning by fine-tuning the pre-trained AlexNet architecture to improve the verification accuracy. Our experiments have demonstrated improved results using proposed scheme and thereby indicate the benefit of the contactless palmprint data to verify the identity of the new-born babies.