MMFV: A Multi-Movement Finger-Video Database for Contactless Fingerprint Recognition

Abstract

Biometric authentication during the COVID-19 and post-pandemic times require a touchless authentication mechanism. While existing studies showcase the use of fingerphoto for touchless authentication, a short video of the finger can provide many good-quality frames. This research presents the first publicly available finger-video dataset, titled Multi-Movement Finger-Video (MMFV) Database. The MMFV dataset has 3792 videos from 336 classes, acquired over two sessions, and spans three different movement types (pitch, yaw, and roll). To establish the baseline performance for the proposed MMFV database, we perform recognition using seven popular fingerprint and deep learning-based algorithms for fingerphoto recognition. The recognition is performed using a fixed, randomly selected frame from all the algorithms. Experimental results showcase that Siamese network-based verification provides the most optimal results across different movements, with observed EER as low as 2.70%.