Fingerprint Authentication

"Technoimagia is a stealth company." (NIST scientist)

The minutiae based fingerprint
matching has being used in the
forensic science for a long time.
There are many available
software based on the method
but the accuracy and the
processing speed are quite
different. Technoimagia's proprietary method uses unique algorithms which have been tested and verified with an in-
house database that has 250,000 fingerprints.

The minutiae method is ideal for security because the fingerprint image is not required for a matching process. In addition, a template of the minutiae information is irreversible; you cannot reverse a template to re-construct an original fingerprint image. We encrypt all templates for additional security.

Diagram below shows steps of fingerprint image acquisition and tempelate construction. The fingerprint image will be deleted from the system once the template is constructed and stored.

	=>		=>
*Finger Scan*		*Fingerprint Image*		*Minutiae Template*

Recently, National Institute of Standards & Technology (NIST) had two major studies on commercially available fingerprint matching algorithms: "Fingerprint Vendor Technology Evaluation 2003 (FpVTE2003)" and "Minutiae Interoperability Exchange Test 2004 (MINEX04)". Technoimagia was one of the finalists (18 companies) in FpVTE2003, and revealed critical factors that need to consider in the negative recognition. In MINEX04, there were 15 companies remained as the finalits. There were 9 vendors which participated in both events. Technoimagia successfully submitted the required templates and SDKs for evaluation. This demonstrated that our propriety extraction and matching methods complies with ANSI INCITS 378-2004 "Fingerprint Minutiae Format for Data Interchange", and the Bio-API specification. It also complies with Biometric Application programming Interface (BAPI) 2.0, which is an extension of Bio-API and adopted by Microsoft.

Technoimagia's minutiae extraction and matching algorithms provide a fingerprint of template of 128 to 256 bytes, and therefore achieve a fast processing speed. Only restriction to the process speed is the clock speed of CPU!

￭ 1:1 verification:

For example, in a preliminary test using 1,000 raw fingerprint images provided by FpVTE2003, our algorithm recorded:

(1) A high end PC (3.0 GHz dual Xenon & 2 GB RAM) took:
0.26 sec for minutiae extraction per finger 4.5 msec per matching pair

(2) A low end PC (1.4 GHz single Pentium 4 & 256 MB RAM) took:
0.50 sec for minutiae extraction per finger 12 msec per matching pair

The actaul test of FpVTE2003 showed that FAST21 took approximately 5 msec per matching pair, and 6.5 days to complete 100 million matching pairs, using the high end PC.

￭ 1:N identification:

We have successfully reduced the processing time of 1:N identification by various methods. FP-RadarBase can search identify a finger from a database of 10,000 fingerprint templates with less than 1 sec even using a PC with a 200 MHz Pentium IV CPU. The fasted algorithm we have developed as FP-RadarBase II can take less 1 sec to search 50,000 fingerprint database using a 200MHz Pentium IV based PC.