Using Hand-Dorsal Images to Reproduce Face Images by Applying Back propagation and Cascade-Forward Neural Networks

Abstract

This paper concentrates on reproducing face images from hand-dorsal images. This idea is adopted to enhance the biometric system outcomes. That is, best identifications can be presented by providing the face images of people as this can lead to directly recognizing the individuals. Non-linear relationships between hand-dorsal images and face images are designed and implemented. The power of Cascade-Forward Neural Network (CFN) and Back Propagation Neural Network (BPN) are employed to reproduce all face details by utilizing a hand-dorsal image. Both networks recorded interesting results in reproducing the details faces. The CFN performance is equal to 2.8571% and the BPN performance is equal to 6.4286%. Furthermore, the Average Correlation (ACORR) for the BPN which achieved 0.9874, this is lower than the ACORR for the CFN obtained to 0.9940. These performances reported that the CFN has significant ability to recognize people according to their face images.