Palmprint Recognition: A Naïve Bayesian Approach

Identification of individuals via palmprint based biometric system is becoming very popular due to its reliability and high performance as it is enriched with several unique and stable features. In this article, we present a novel, convex hull oriented feature based machine learning approach for palmprint recognition. For robust recognition, we proceed as follows: Firstly, for noisy data, a pixelwise Niblack's binarization method is adopted, which adds to the efficiency of the latter stages. Secondly, endpoints are determined for all the lines detected which are then used in the construction of a convex hull. The quantitative, structural and geometrical features of the constructed convex hull are then used to build the palmprint model. For the model creation, we use Naive Bayesian approach based on the independent feature model. The experimental results show that the approach outperforms other approaches.