A Fuzzy Rule-Based Fingerprint Image Classification

This paper aims to present an improved rule-based fingerprint classification using coherence method and a fuzzy rule-based (or simply fuzzy-based) fingerprint classification. The improved classification using coherence method has been applied to fingerprint images that do not previously produce any singular points while the fuzzy rule-based classification is designed based on the uncertainty in classifying tented arch, left-loop and right-loop fingerprint images. Two common classification schemes have been used, i.e. 5-class and 4-class schemes. The performance measure uses the success rate and it has been found to achieve 88.38% and 88.33% in the 5-class scheme for the improved rule-based and fuzzy rule-based classifications, respectively. On the 4-class scheme, it has been found to be 92.2% and 92.13%, respectively. The study has also considered weighting the success rate based on the natural proportion of the fingerprints. The success rates have been found to be 89.33% and 90.18% for the improved rulebased and fuzzy rule-based classifications, respectively for the 5-class scheme. On the 4-class, it has been found to be 90.36% and 91.25%, respectively.