Biometric Performance

Factors affecting biometric accuracy

The overall accuracy of a fingerprint recognition system may vary according to a number of factors such as:

• Quality of the sensor,

• Quality of the fingerprint pose,

• Skin condition (dry, wet, scars),

• Population under test.

• etc.

Performance metrics

False non match rate (FNMR) is the proportion of mated comparisons below a threshold set to achieve the false match rate (FMR) specified. FMR is the proportion of impostor comparisons at or above that threshold. Since FMR and FNMR is in inverse proportion to each other, choosing the operational threshold is a trade-off between system security and user convenience.

NIST MINEX III evaluation

The MINEX III was initiated by the National Institute of Standards and Technologies (NIST). It aims at measuring the performance of automated fingerprint recognition technologies using minutiae data only for interoperability among vendors.

Datasets

As stated in MINEX III test plan:

§3.2. Resolution and dimensions

“All images for this test will employ 500 pixels per inch resolution (horizontal and vertical). The dimensions of the fingerprint images will vary from 150-812 pixels in width, and 166-1000 pixels in height.”

§3.3. Sensor and impression types

“All images used for testing in MINEX III […] have been obtained from live-scan sensors (Smiths-Heimann ACCO 1394 and Cross Match 300A). All images tested in MINEX III are plain impression type images.”

Results

The overall ranking can be found here: https://www.nist.gov/itl/iad/image-group/minutiae-interoperability-exchange-minex-iii.

The latest reports of the MINEX III for id3 algorithms can be found here.

The finger minutia detector 3B of this SDK corresponds to the template generator of the id3_13B1 submission. The finger minutia matcher of this SDK corresponds to the template matcher of the id3_13B1 submission.

Internal evaluation

To help the id3 Finger SDK user to choose what data to extract from a fingerprint depending on its use case, we evaluate the various possible fusions on some public datasets.

Datasets

We evaluate internally our algorithms on the FVC datasets (2000, 2002 and 2004). More details can be found at the following URLs:

• http://bias.csr.unibo.it/fvc2000/databases.asp

• http://bias.csr.unibo.it/fvc2002/databases.asp

• http://bias.csr.unibo.it/fvc2004/databases.asp

Results

The following graph presents the various possible combinations of match fusions using id3 Finger SDK.

Matching method	FNMR @ FMR = 1/1,000	FNMR @ FMR = 1/10,000
MINEX only	1.42%	2.01%
MINEX + minutia embeddings	0.98%	1.43%
MINEX + finger embedding	0.05%	0.15%
MINEX + minutia embeddings + finger embedding	0.04%	0.12%

Hint

• In the context of interoperability, use minutiae only.

• In the context of a “small” sensor (under 200x200px at 500dpi), use minutiae and minutia embeddings.

• In the context of a large sensor (above 200x200px at 500dpi), use minutiae, minutia embeddings and finger embedding.