Classification of Live Scanned Fingerprints using Histogram of Gradient Descriptor

The processing time during fingerprint recognition is a main problem when the fingerprint database is huge. Classifying fingerprints into subcategories is an effective way to restrict the search space into a sub-database. We propose a new fingerprint classification method based on modified Histograms of Oriented Gradients (HOG) descriptor. The way orientation field is computed in HOG descriptor is not adapted to the ridge patterns. We compute the orientation field, which is adapted to the ridge patterns and incorporate in HOG descriptor, enhancing its potential to represent a fingerprint in a robust way. Extreme Learning Machine (ELM) with RBF kernel is used as a classifier. We performed experiments on the noisy fingerprint database FVC-2004, a benchmark database; the proposed method achieved the average accuracy of 98.70, which is better than those of the state-of-the-art fingerprint classification methods.