ROC highlights accuracy gains of latest fingerprint algorithm in NIST test

ROC has revealed a 35 percent improvement in the accuracy of its fingerprint biometric algorithm, based on an evaluation by the U.S. National Institute of Standards and Technology.

The “roc+0007” algorithm was evaluated by NIST for the Proprietary Fingerprint Template (PFT) program on February 6. Compared to the roc+0001 algorithm, which ROC says is developed for fast performance, the new algorithm scored error rate reductions in all eight PFT III benchmarks. The improvements range from 10 percent for the Port of Entry (PoE) dataset to 35 percent for the Los Angeles Sheriff’s Department (LASD) database.

The newer algorithm scored the best false non-match rate (FNMR) among global developers for the U.S. Visit #2 dataset at 0.005, with false match rate (FMR) set to 0.0001. While less speed-focussed than the previous algorithm, the new one takes only 701 microseconds per match, according to the company announcement.

The accuracy gain is similar to that recently reported for ROC’s submission to NIST’s ELFT test in February.

ROC also notes that the results are the best among U.S. providers of FBI-compliant biometric systems for law enforcement, and give it the best global combination of speed and accuracy.

“ROC’s debut into PFT showcased an algorithm already top-tier in terms of both accuracy and speed,” says Dr. Joshua Engelsma, principal scientist at ROC. “However, our steady improvements in accuracy over the past year demonstrate that we are deeply committed in our research and engineering efforts to deliver even higher levels of accuracy to our customers. We are confident that the momentum we have in building next-generation fingerprint recognition capabilities will continue to separate us from the competition in the ongoing PFT evaluations.”

The company celebrated “best in the West” accuracy results with its latest submission to NIST’s face biometrics test earlier this month.

Article Topics

accuracy  |  algorithms  |  biometric testing  |  biometrics  |  ELFT  |  fingerprint biometrics  |  NIST  |  ROC