An Effective Personal Identification System Using Hand Dorsal Modality and Deep Learning Approach

Abstract

Hand Dorsal identification is a type of biometric technology that has emerged in the past two decades. Because of its safety, accuracy, and efficacy, more and more researchers are participating in the study. In this short paper, A biometric based Hand dorsal identification system is proposed. However, due its potential capability to extract and differentiate between the system user, a CNN based deep learning approach named ALEXNET along with the tied rank normalization is used to extract the discriminant hand dorsal features. In order to perform the classification task, the Support Vector Machine is implemented. Our work is applied to a database known in this field and has produced a very promising result when using the Hong Kong Polytechnique hand dorsal database.