Our own R&D
Since its inception, GEYCE Biometrics has allocated 35% of its annual budget to research in software and biometric specialisation, as well as in development tools that have become unique in the sector:
Optimization of fingerprint quality
Fingerprint enrolment firstly requires a correct analysis of the fingerprint and its quality. Identifying the quality level does not suffice. At GEYCE Biometrics we go beyond this and identify the cause of the problem for the operator, which makes it possible to generate real improvements in fingerprint capture and reduce rejection rates.
This assessment is consistent with NFIQ-2 and the results are very useful so that the operator can guide the user in replicating the process successfully.
This combination of techniques is part of GEYCE Biometricsâ€™ success model aimed at achieving high enrolment quality.
ICAO standard compliance guarantee in face photographs
Front view face photographs affixed to identification documents, such as passports, visas or ID cards in each State, are regulated according to an ICAO standard which establishes certain photograph sizes and characteristics so that subsequent manual or automatic identification systems can carry out recognition.
At GEYCE Biometrics we are applying techniques with neuronal networks to allow us to recognise defective characteristics of the image, such as closed eye detection, facial gestures, open mouth detection or hairs partly covering the face. With these detections we can help the operator or the user to ensure the image complies with the standard.
This solution allows us to design and generate code to build management applications in the various existing platforms: desktop, web, app services.
It is based on the evolution of the lower CASE tool we call JEDICASE to define in a declarative manner how we wish to organise information in a database from the view model.
The code generated creates a framework we call JEDIFC, which is implemented and adapts to each of the operating platforms to ensure the declarative model remains unchanged.