Contactless provider Neurodactyl’s fingerprint recognition shows strong accuracy in NIST test
Although it was just founded in 2023, contactless fingerprint biometrics provider Neurodactyl says it has reached a top spot in NIST evaluations. The Tbilisi, Georgia-based startup ranked first for recognition accuracy for border control and visa application datasets in the NIST PFT III evaluation as of May 4, 2023.
The company’s contactless biometric capture technology allows for identification and verification just by holding a fingerprint in front of a mobile phone camera. Neurodactyl says its matching algorithms also recognize scans and photo images and can capture multiple fingerprints at once.
To achieve accuracy with challenging, poor-quality fingerprints, Neurodactyl applied new approaches in deep learning like GANs and diffusion models, the company says in a release.
Neurodactyl says in the announcement that its algorithm demonstrated the lowest missing rate for the most challenging fingerprint pairs in the datasets with flat fingerprints (“US Visit” and “The Port of Entry/BioVisa Application”). Only 0.37 percent of fingerprint templates were not recognized at a false acceptance rate of 0.01, and 0.52 percent at FAR 0.0001. Recognition by two fingers the missing rate is much lower: 0.06 percent and 0.08 percent at FAR 0.01 and 0.0001 respectively.
Besides high recognition accuracy, it also provides extremely fast matching, starting from 100 million matches per second, and a fixed-size template, meaning that biometric templates always have a standard size of 512 bytes. This makes it suitable for large-scale projects that have millions of users, according to the company.
Although contactless fingerprint accuracy has been increasing, recent tests from BixeLab have shown it is still not on par with contact systems, particularly for enrollment. Contactless fingerprint recognition, however, has become accurate enough for suitability to certain use cases.
Neurotechnology claimed a top result in its March 27 PFT III evaluation, with an FNMR of 0.007 at an FMR of 0.0001 on the LASD dataset. Neurodactyl scored a 0.0267 FNMR at 0.0001 FMR on the same dataset.
A dozen biometrics developers have participated in the PFT III evaluation since it launched in 2019.
Last month, the Unique Identification Authority of India (UIDAI) said it will be developing its own touchless fingerprint biometric capture system for the country’s Aadhaar digital ID system.
This post was updated at 9:44am Eastern on May 13, 2023 to clarify that biometric matching technology was tested, rather than capture technology.