A cascaded deep network architecture for multi modal biometric recognition

Abstract

It is very often to see that the results obtained from a uni-modal biometric system have negative results when the input subject is deteriorated or damaged.  To provide more accurate and high security with biometrics, researchers have integrated the unique feature of more than one biometric system resulting in multi modal system. Face and Iris recognition systems yield higher precision and observed to be robust against tampering because of the unique features that are extracted for classification. In recent times, deep learning has gained a lot of importance in computer vision application especially for image classification with high accuracy and similarity index. Multiple layers of convolution operation are performed on the spatial elements of image, extracting the finer details pertaining to it. This makes the deep network more suitable for solving classification problem. This paper presents a cascaded customized deep network architecture that aims to merge the classification scores of multiple biometrics to attain more robust recognition system that is invariant to illumination. The proposed approach is tested and evaluated with two standard datasets and metrical analysis is compared against state of art methods