A Framework for Level-1 and Level-2 Feature Level Fusion

Abstract

The paper presents the feature extraction of level-1 and level-2 fingerprint features. The level-1 features are based on ridges and level-2 features are based on minutiae of the fingerprint, which are used for authentication. The level-1 and level-2 feature level fusion is proposed to enhance the accuracy rate of unibiometric fingerprint authentication system. The performance of this system is evaluated on two publicly available databases and our own database. Fusion scheme is decided depending on the type of information available. Choosing the fusion scheme is important for any multibiometric system as it has significant impact over the performance of the multibiometric system. The identification rate is more in concatenated feature set than for individual feature set. The proposed FFV uses the much richer grey-level information of the fingerprint image. It is also capable of dealing with fingerprints of low quality images from which features cannot be extracted reliably. The results shows that the proposed FFV method outperforms other competing approaches with both EER and DI.