A fingerprint classification system and method for extracting the dominant singularity from a fingerprint image.
Inventors: Laurice Phillips, Margaret Bernard, Duc Du Kieu
Fingerprint classification and verification has been an important technique in personal identification and security and has been proven to be a reliable tool in criminal investigation and personal identification. Nonetheless, there is still need for improvement in the technology, particularly in processing and achieving high levels of accuracy with poorer quality images.
A novel approach to fingerprint classification using regular expression machine learning has been developed. This technique takes a digital fingerprint image and applies a new machine learning strategy for extracting the dominant singularity from a fingerprint image. The fingerprint recognition system, firstly, receives as an input a digital fingerprint image and then applies a pre-processing to generate an enhanced and more accurate image. Secondly, feature pattern calculations are performed on the updated image to generate a feature orientation estimation vector. Thirdly, the feature orientation vector is processed using a novel Regular Expression Machine and a classifier prediction model to generate a class label for the digital fingerprint image and identify a fingerprint image. The technique provides an improvement of the existing mainstream techniques and is more robust with poorer quality images.
The core BITREM functionality has been developed on an extensible framework which can be used to develop several Biometric applications, for example:
- Various types of Civilian Registries such as Immigration and Citizen Identification (National Id, Births, Deaths, Marriage, Children, Seniors, Elections) as well as Law Enforcement databases.
- Access Control application/Time Attendance
- Access control on mobile devices
- Faster processing of fingerprint images, especially lower-resolution images.
Patents granted in Trinidad and Tobago and in the United States of America.