Fingerprint classification based on Adaboost learning from singularity features

Fingerprint classification is an important indexing scheme to narrow down the search of fingerprint database for efficient large-scale identification. It is still a challenging problem due to the intrinsic class ambiguity and the difficulty for poor quality fingerprints. In this paper, we presents a fingerprint classification algorithm that uses Adaboost learning method to model multiple types of singularity features. Firstly, complex filters are used to detect the singularities. For powerful representation, we compute the complex filter responses of the detected singularities at multiple scales and a feature vector is constructed for each scale that consists of the relative position and direction and the certainties of the singularities. Adaboost learning method is then applied on decision trees to design a classifier for fingerprint classification. Finally, fingerprint class is determined by the ensemble of the classification results at multiple scales. The experimental results and comparisons on NIST-4 database have shown the effectiveness and superiority of the fingerprint classification algorithm.