Feedback based Adaptive matching algorithm for cancellable fingerprint templates

The Cancellable fingerprint generation techniques extract features from the fingerprint, obtain a user key input from the user and transform the features to a non-invertible domain using the user key input and store the transformed features as templates. Non-invertible transformations are many to one transformation which means that distinct biometric feature sets could be transformed in to similar feature sets because of transformation by user key input. The matching algorithms generally match between enrolled and query bit strings and generate a score based on how different the two bit strings are to each other, like hamming distance. This simple approach to matching works well when both bit strings contain raw biometric features such as minutiae locations and angles. But matching cancellable templates using such bit string matching algorithms would not be successful because they don’t take into account the many to one transformations caused by the user key. Specifically, the ability of the matching algorithm to separate between genuine and imposter diminishes and Equal Error Rate increases. A novel feedback based adaptive matching algorithm that matches features instead of bit-strings is presented in this paper. The proposed algorithm matches features in the cancellable templates that are generated by Delaunay triangulations. The algorithm includes a feedback based parameter which enables the matching algorithm to maximize separability and minimize EER by adapting itself to the templates in the population which is demonstrated by applying the algorithm on public database. The separability thus obtained is higher than other such algorithms and EER is comparable with other such algorithms.