Noisy fingerprint classification using multilayer perceptron with fuzzy geometrical and textural features

Abstract

A multilayer perceptron is used for the classification of noisy fingerprint patterns. In the first phase the input vector consists of some fuzzy geometrical features. In the second phase, we use some texture-based and directional features. The output vector is defined in terms of five classes, viz., whorl, left loop, right loop, twin loop and plain arch. Perturbation is produced randomly at pixel locations to generate noisy patterns. Cut marks and loss of information in certain random locations are also simulated. The investigation helps to demonstrate the generalization ability of the model in handling distorted fingerprint images.