ELFT shows improvements in latent fingerprint biometrics accuracy

The latest Evaluation of Latent Fingerprint Technologies (ELFT) by NIST shows further gains in the accuracy and speed of the top-performing latent fingerprint biometrics matching algorithms.
An algorithm submitted by ROC this month scored the lowest Rank-5 search error rate recorded, at 0.0194 for all probes, and 0.0035 for those with EFS data. This is important, the company says, because this is how forensic examiners work with latent fingerprint biometrics in practice.
Innovatrics, Dermalog, Tech5 and HiSign have each had an algorithm evaluated for ELFT since the beginning of March.
For the FBI’s Solved Dataset #1, HiSign currently has the lowest False Non-Identification Rate (FNIR) at a False Positive Identification Rate (FPIR) of 0.01, at 0.0213, followed by Idemia’s algorithm submitted earlier this year at 0.0484, and then the new entry from Innovatrics, at 0.0543.
The results for DoD Dataset #1, both in FNIR at FPIR 0.01 and Rank-5 FNIR, where ROC finished second, were similar.
ROC touts speed, nationality considerations
ROC’s new algorithm completed the comparison with the second-lowest mean mated search duration, at 386 seconds, bested only by the algorithm it submitted in February. While that algorithm delivered significantly improved accuracy compared to the company’s previous submission, the new one reaching an FNIR of 0.0969.
ROC Co-founder and Chief Scientist Dr. Brendan Klare notes the geographic makeup of the top performers in the ELFT.
“With each improvement, law enforcement and intelligence missions move closer to tools that truly serve their needs,” says Klare. “Investigators can now search larger databases faster, widening search parameters and ensuring no potential lead is left behind… All powered by American-built technology that breaks the multi-decade, foreign grip on U.S. government biometric systems.”
Article Topics
accuracy | algorithms | biometric matching | ELFT | fingerprint biometrics | NIST | ROC




Comments