Fingerprint Classification Based on Spectral Features

Fingerprint Classification Based on Spectral Features

Hossein Pourghassem, Hassan Ghassemian


Fingerprint is one of the most important indexes that can be applied to human verification and identification. In the recent decade, with the population explosion and tremendous increase in fingerprint databases, automation of identification has been unavoidable. Fingerprint classification decreases the time of search for an unknown image in large databases. In this research, translation and rotation invariant features are extracted from spectrum of the fingerprint image, and a new tessellation on the frequency spectrum based on an appropriate definition of frequency concept in the fingerprint application, has been designed. The extracted features obtain not only information from frequency of ridges but also valuable information from direction of ridges in the fingerprint images. Features are classified with Probabilistic Neural Network. FVC2000 and FVC2002 databases are used to assess the proposed algorithm. The proposed algorithm provides an accuracy and speed of classification more than previously reported in the literature. We have obtained an accuracy of 93.4 percent with a rejection ratio 2.8 percent for the seven-class task and for the six-class classification task an accuracy of 95.1 percent with a rejection ratio 2.4 percent is achieved.


fingerprint, classification, probabilistic neural network, feature extraction, spectral features