Fingerprint authentication system using Back-Propagation with downsampling technique

Fingerprint authentication process plays a crucial role for human identifications. Fingerprint stored in the database are often used to confirm individual identity in cases like security checks, disaster, and medical jurisprudence. However, when dealing with the database consisting of a huge number of the fingerprint, recognizing the correct fingerprint matches using some of the existing methods present a great challenge. Therefore, it is critical to develop automatic fingerprint authentication process with fast execution time and high precision. Here we introduce a novel downsampling pixel preprocessing technique to compress original fingerprint pixel matrix to a unit input vector for Artificial Neural Network for fingerprint authentication system. Furthermore, the proposed downsampling technique compress the original pixel matrix by computing the arithmetic mean of the sum of the pixel values on each input row matrix to generate a unit input vector for ANN. Using back-propagation technique the algorithm trains the system to matches fingerprints samples and relates them to the number provided for each authorized user. The interest of the proposed method is illustrated by its application to 102 fingerprint samples with standard 500 pixels both row and column with 8bits grayscale resolution. The results indicate the proposed Back-Propagation with downsampled pixel precisely recognize 100 fingerprints from 102 tested samples. Moreover, the comparative evaluation shows the proposed method outperform Back-Propagation normal pixel, Perceptron with Downsampled pixel and Perceptron with normal pixel achieving 98.03 % precision with minimum convergence time of 30 seconds and mean square error (MSE) of 0.05%.