The Council for Scientific and Industrial Research (CSIR) Information Security researchers have developed a structural fingerprint classifier that is able to correctly classify a fingerprint with only partial information.

The researchers also introduced novel fingerprint features – collectively referred to as pseudo-singular points – as feeder to ‘extensible structural fingerprint classifier’.

Professor Fulufhelo Nelwamondo, CSIR Head of Information Security research domain within CSIR Modelling and Digital Science research unit, explains the need for a classification module: “In fingerprint recognition, fingerprint templates normally sit in a database; so when going through an identification process, the system has to sift through thousands, if not millions of templates, making the system slow in yielding results.”

He adds that a classification module essentially breaks down the overall database into smaller, manageable chunks to improve the performance of a fingerprint recognition system.

“Both the extensible fingerprint classifier and the pseudo-singular point detection module will allow the system to be extremely fast and accurate when database search is conducted. This will add to the overall efficiency of the entire fingerprint recognition system,” notes Prof Nelwamondo.

“Such a fast and efficient system is used on a daily basis in large and integrated solutions such as the Department of Home Affairs’ national identification system. However, this breakthrough could contribute to future systems that are even faster because of the system’s ability to match fingerprints using only partial information.”

From this technological breakthrough, CSIR researchers are now rigorously studying the concept of pseudo-singular point detection (P-SPD) and false transition elimination, a concept that also emerged from P-SPD.

The latter involves identification and location approximation of a global fingerprint landmark.

Pseudo-singular points are faux versions of the conventional singular points, yet they are easy to detect and provide almost the same classification accuracy.