Fingerprint Classification with Combinations of Support Vector Machines

  • Authors:
  • Yuan Yao;Paolo Frasconi;Massimiliano Pontil

  • Affiliations:
  • -;-;-

  • Venue:
  • AVBPA '01 Proceedings of the Third International Conference on Audio- and Video-Based Biometric Person Authentication
  • Year:
  • 2001

Quantified Score

Hi-index 0.01

Visualization

Abstract

We report about some experiments on the fingerprint database NIST-4 using different combinations of Support Vector Machine (SVM) classifiers. Images have been preprocessed using the feature extraction technique as in [10]. Our best classification accuracy is 89.3 percent (with 1.8 percent rejection due to the feature extraction process) and is obtained by an error-correction scheme of SVM classifiers. Our current system does not outperform previously proposed classification methods, but the focus here is on the development of novel algorithmic ideas. In particular, as far as we know, SVM have not been applied before in this area and our preliminary findings clearly suggest that they are an effective and promising approach for fingerprint classification.