Fingerprint images of newborns differ significantly from those of adults due to strong deformation, large pores, and thin valleys. As a result, minutiae, which are key features used in traditional fingerprint recognition, cannot be reliably detected in newborn fingerprint images. However, ridge patterns in newborn fingerprints remain observable and resemble those of adult fingerprints. We propose a novel newborn fingerprint recognition method based on fingerprint classification. The proposed method integrates an attention branch network and FingerNet to effectively process these challenging fingerprint images. Through a set of experiments using newborn fingerprint images captured between 2 and 24 h after birth, we demonstrate the effectiveness of the proposed method.