Moment based wavelet filter design for fingerprint classification

Fingerprints are the most universal, unique and persistent biometrics. This study performs the classification of multiple fingerprint images using Hu Moments based Wavelet designing and Modified VZ Classification Algorithm. The proposed work is divided into two stages. One is the wavelet designing stage and second is the classification stage. The use of moments as filter coefficients is motivated with a discussion on the type of information extracted by a filter and its representation. The basic VZ classification algorithm used has been modified in this work, which is initially developed within a rotationally invariant framework. The probabilistic neural network and support vector machine classifiers are employed for the advancement of VZ algorithm and better classification. The designing and classification stages of the VZ algorithm are described, and the performance on different sets is evaluated. The performance is evaluated on the standard fingerprint verification certification FVC2004 database. The comparative simulation experiments show that the classification produces a good performance as compared to previous results and providing high accuracy in classification rates. Performance of these two classifiers is compared among themselves as well as with the previously reported articles. This concludes that PNN has 99.23% classification rate for classification of ten classes of fingerprint images, which is better than that of SVM.