Iris Recognition for Infants

TL;DR

This work evaluates iris recognition for newborns (4–6 weeks) by introducing an infant-focused pipeline that combines a custom NIR iris sensor, an infant-specific segmentation model, and a privacy-preserving synthetic-data generator. It collects 1,920 high-quality iris images from 17 infants, trains a novel segmentation network, and evaluates six iris matchers, including a StyleGAN2-ADA–based synthesis of 1,000 synthetic infant iris images to enable privacy-safe research. The results show substantial performance gains when using infant-specific processing, with an EER of $3\%$ and AUC of $0.99$, suggesting that reliable features can be extracted from infant irises. This approach has potential implications for hospital safety, post-natal health monitoring, and humanitarian contexts where non-invasive, scalable identification is critical. The work also provides privacy-preserving synthetic data and open segmentation tools to stimulate further research in infant biometrics.

Abstract

Non-invasive, efficient, physical token-less, accurate and stable identification methods for newborns may prevent baby swapping at birth, limit baby abductions and improve post-natal health monitoring across geographies, within the context of both the formal (i.e., hospitals) and informal (i.e., humanitarian and fragile settings) health sectors. This paper explores the feasibility of application iris recognition to build biometric identifiers for 4-6 week old infants. We (a) collected near infrared (NIR) iris images from 17 infants using a specially-designed NIR iris sensor; (b) evaluated six iris recognition methods to assess readiness of the state-of-the-art iris recognition to be applied to newborns and infants; (c) proposed a new segmentation model that correctly detects iris texture within infants iris images, and coupled it with several iris texture encoding approaches to offer, to the first of our knowledge, a fully-operational infant iris recognition system; and, (d) trained a StyleGAN-based model to synthesize iris images mimicking samples acquired from infants to deliver to the research community privacy-safe infant iris images. The proposed system, incorporating the specially-designed iris sensor and segmenter, and applied to the collected infant iris samples, achieved Equal Error Rate (EER) of 3\% and Area Under ROC Curve (AUC) of 99\%, compared to EER$\geq$20\% and AUC$\leq$88\% obtained for state of the art adult iris recognition systems. This suggests that it may be feasible to design methods that succesfully extract biometric features from infant irises.

Figures (6)

• Figure 1: Frontal and rear views of the handheld newborn iris scanner custom-designed for this study by an industrial partner.
• Figure 2: Examples of synthetically-generated infant iris images. The trained generative model offers a remarkable realism of the synthetic images and correctly captured intricate details, such as the iris texture, brighter pupil (compared to adult iris images, in which the pupil is usually darker than iris), eyelid retractors used by medical personnel, specular highlights or skin texture.
• Figure 3: Comparison of iris segmentation visualizations across various states: infant, adult, and post-mortem. The visualizations compare the performance of our developed model with state-of-the-art methods. The model effectively segments irises with varying characteristics, including dark, bright, small, and large pupils.
• Figure 4: Distributions of the selected ISO/IEC 29794-6 iris image quality metrics calculated for adult and infant iris samples.
• Figure 5: Distributions of genuine and impostor scores for four iris recognition methods that were able to process infant iris images. Vanilla segmentation models and encoding approaches were used, with default parameters suggested by the original method authors. Selected performance metrics ($d'$ statistic, Equal Error Rate (EER), Failure-to-Match rate and Area Under ROC curve (AUC) are also shown.