Fingerprints Recognition Using a Neural Network

Abstract

A fingerprint recognition approach, based on a backpropagation algorithm using a neural network, is presented. The purpose of our work is to show the relevance of the fingerprint method characterization by using the vectorization method. This method reduces dramatically the required amount of data for later application on neural networks having some variations of the backpropagation method for fingerprint recognition. When this step is completed the next step is to establish a comparison between the algorithms in terms of time. The algorithms were implemented by using 'C' language on an IBM PC 's platform. The neural network has four layers. Two of them are hidden. The third one is the input and the last one is the output. In our particular experiment, we used five neurons in the input layer and seven neurons in the output layer. This neural network could be trained to a maximum capacity of 128 fingerprints. However, the method could be easily expanded for the recognition of a larger amount of fingerprints. The performances of the algorithms were checked against the run time for each algorithm using vectors extracted from digitized fingerprint images. The LMS algorithms have several variants and each one of them has their advantages. In this case, the reference for comparison is the computation time, done with the same computer, and for the same application.