Noisy fingerprint classification using multilayer perception with, fuzzy georrretrical and textid features

A multilayer perceptrcm is used for the classification of noisy fingerprint patteri%. in the first phase the input vector consists of some fiw gmmekricalt features. In the semnd phasq we use some texture-based and dircctioql fkatures. The ou@ut Gectcr is defined in terms of five Asses, viz., whorl, IelI loop, right loop, twin loop and plain arch. PcrklrQation is produced .mndpnfy at p&q1 hark% to generate noisy @terns. Cut marks and loss of information in certain random Iocations are also simulated. The investigation helps to demoiktrak the generalization ability of the model in fiandling distorted fmgerpritit images.