Genetic Algorithm Based Optimization of Threshold Parameters in Fingerprint Quality Estimation

In addition to face, fingerprint is one of the most effective biometric features to be used personal authentication for biometric recognition. To make automatic fingerprint recognition system (AFIS) efficient, it is important to have information on the quality and validity of the captured fingerprint images so that it can work independently in a reliable manner. For quality estimation of each fingerprint, feature vector is generated with the help of optimized threshold model in most of the techniques. In order to avoid the local maximization of objective function, we used Genetic algorithm (GA) based approach. However, GA solution is normally used for the solution of string-vector format. Contrary to this fact, we have used GA approach for the solution of value vector format. This algorithm for optimization can be extended to other application whenever there is need of deciding the optimal threshold in situation similar to described above. The new random-variance operator has played crucial role in handling the population evolution in the optimization problem having solution in valued-vector format. Further, new kind of objective function is introduced to include all four classes of quality in objective function. The result shows the usefulness of this approach for optimization problem.