Home ＞ Technology & Service ＞ UB-safe fingerprint authentication engine

Tecnology&Service

Embedded development

OEM/ODM

UB-safe fingerprint authentication engine

UB-safe fingerprint authentication engine

DDS provides UB-safe fingerprint authentication solutions for embedded devices, with a fingerprint matching engine as a core that performs a series of tasks from “image compensation” to “judgment” shown in the following figure.

“Frequency analysis” algorithm

According to the frequency analysis method, the cross section obtained by slicing the fingerprint pattern is viewed as a wave. Fingerprint matching is performed by using a spectral wave series as the characteristic information and obtaining a maximal correlation value. The frequency analysis algorithm is based on an algorithm that is a result of the research carried out by Professor Taizo Umezaki at the Graduate School of Nagoya Institute of Technology. The major characteristic of this algorithm is that it applies the one-dimensional frequency analysis method used in voice recognition to image recognition.

A partial transcription from a paper issued by the Institute of Electrical Engineers of Japan (May 2002), "Evaluation of a Fingerprint Verification Method Based on LPC Analysis" by N. Matsumoto, S. Sato, H. Fujiyoshi, & T. Umezaki.

Features of the frequency analysis method

The frequency analysis method has the following features compared with fingerprint authentication algorithms that are currently in practical use, “Minutia extraction and matching method (Minutia method)” or “Image-matching method (Pattern-matching method).”

Item	Frequency analysis method	Pattern-matching method	Minutia method
Denial of registration	○ No denial in principle	○ No denial in principle	△ Some denied (1 to 3%)
Image operation amount	○ Mainly local operation	△ Global operation also needed	△ High level interpolation processing
Processing time	◎ High-speed and integer operation only	○ Slow	△ Heavy, variable
Accuracy of recognition	◎	△	◎ Susceptible to variability
Influence of faded fingerprints	◎	○	△ Loss of minutia
Privacy protection	◎	△ Matching with fingerprint image	◎
Affinity with sweep sensors	◎	△	△

Past adoption of the algorithm

The fingerprint authentication algorithm based on the frequency analysis method has been embedded into multiple products following technical instructions given by Professor Umezaki, and has actually been used since 1995.

UB-safe product specifications

Matching speed (one-to-one / one-to-ten matching)	ARM7TDMI　66MHz　0.40SEC/1.30SEC VR4181A　131MHz　0.22SEC/0.50SEC M32　216MHz　0.16SEC/0.40SEC PentiumIII　800MHz　0.012SEC/0.04SEC
False acceptance rate (FAR) False rejection rate (FRR)	FAR=0.001%or less、FRR=0.1%or less
Program size (ROM)	Some50KB
Data size (RAM)	Some260KB、Stack 1KB or less
Template size	3KB Threetime registration per finger

Contact us