Algorithms on strings, trees, and sequences: computer science and computational biology
Algorithms on strings, trees, and sequences: computer science and computational biology
Guide to Biometrics
Combining Crypto with Biometrics Effectively
IEEE Transactions on Computers
IEEE Transactions on Pattern Analysis and Machine Intelligence
Increasing error tolerance in biometric systems
Proceedings of the 8th International Conference on Advances in Mobile Computing and Multimedia
On the commonality of iris biometrics
Proceedings of the 9th International Conference on Advances in Mobile Computing and Multimedia
Recognising persons by their iris patterns
SINOBIOMETRICS'04 Proceedings of the 5th Chinese conference on Advances in Biometric Person Authentication
Hi-index | 0.00 |
Accuracy is an essential factor to determine how confident a biometric system makes a decision. However, to achieve high accuracy, any biometric systems need to deal with intra-class variability and inter-class similarity problems. The intra-class variability makes exact biometric matching difficult and increases the level of uncertainty, since errors are permitted in matching. Reducing uncertainty in biometric matching is important because it indicates the matcher can reliably decide whether a query biometric template belongs to a registered biometric template or not. In biometric systems where image processing techniques are mainly used, it is impossible to achieve exact biometric matching due to estimation problem, that is exactly estimating the registered image given the query image. However, we approach the estimation problem from a post-processing perspective where derivative information of the registered template is used to transform the query template into another template that can be used to perform an exact match with the registered template. Moreover, large inter-class similarity due to mainly random matches increase false acceptance rate (FAR). To tackle such similarity, the longest common substring (LCS) between the two templates is obtained since it counts for the most significant contiguous matches. We extensively tested our proposed method using iris images from the public and commercial Bath dataset and found that the success rate of the transformation to produce exact matches are 97.5% and 94.34% respectively without falsely accepting an imposter (FAR=0), losing biometric information and multiple scanning.