RBF neural networks for hand-based biometiric recognition

A recognition system based on hand geometry biometrics is reported in this paper. The aim of this development is to improve the reliability of automatic identification systems. The capture of the data, as well as the preprocessing and feature extraction blocks is detailed. Once the features are obtained, they should enter the recognition block, which has been developed using Neural Networks. From the different Neural Networks existing nowadays, the Radial Basis Functions (RBF) ones have been chosen for their shorter training time, and the lack of randomness in the training algorithm. Results are analyzed as a function of the number of vectors of each user taken for the training of the net, obtaining up to 98,1% for only 5 samples of each user.