Convolutional neural networks in hand-based recognition system

Abstract

Biometric systems are often discussed today. Authentication systems that work with biometric data (such as fingerprint, iris, hand geometry) have a high level of security. There are many reasons why it is necessary to have a strong authentication system. One of them is the existence of information systems that store sensitive data that needs to be protected. This article is focused on hand-based identification systems. A typical hand-based authentication system performs: data acquisition, feature extraction, classification, and decision. This paper presents the use of a convolutional neural network to identify people based on hand geometry. Convolutional neural networks are used for pattern recognition. When using a convolutional neural network, it is not necessary before classification feature extraction. Experiments were performed on a database of 550 hand images from 114 people, each person provided 5 images. The accuracy of the identification of persons was 94.11%, 3 images of each person were used for training.