Journal of Intelligent and Robotic Systems

, Volume 40, Issue 1, pp 103–112

Fast Robust Fingerprint Feature Extraction and Classification

Authors

  • H. O. Nyongesa
    • Department of Computer ScienceUniversity of Botswana
  • S. Al-Khayatt
    • School of Computing and Management SciencesSheffield Hallam University
  • S. M. Mohamed
    • School of Computing and Management SciencesSheffield Hallam University
  • M. Mahmoud
    • School of Computing and Management SciencesSheffield Hallam University
Article

DOI: 10.1023/B:JINT.0000034344.58449.fd

Cite this article as:
Nyongesa, H.O., Al-Khayatt, S., Mohamed, S.M. et al. Journal of Intelligent and Robotic Systems (2004) 40: 103. doi:10.1023/B:JINT.0000034344.58449.fd

Abstract

Automatic identification of humans based on their fingers is still one of the most reliable identification methods in criminal and forensic applications. Identification by fingerprint involves two processes: fingerprint feature extraction and feature classification. The basic idea of fingerprint feature extraction algorithms proposed is to locate the coarse features of fingerprints called singular-points using directional fields of the fingerprint image. The features are then classified by different types of neural networks. The “five-class” classification problem is addressed on the NIST-4 database of fingerprints. A maximum classification accuracy of 93.75% was achieved and the result shows a performance comparable to previous studies using either coarse features or the finer features called minutiae.

fingerprint classificationdirectional fieldsfingerprint feature extractionneural network classifiers

Copyright information

© Kluwer Academic Publishers 2004