Demonstration Of Palm Vein Pattern Biometric Recognition By Machine Learning

Mohd Noh, Zarina and Ramlee, Ridza Azri and Ahmad Radzi, Syafeeza and Ramlee, Ridza Azri (2020) Demonstration Of Palm Vein Pattern Biometric Recognition By Machine Learning. International Journal of Human and Technology Interaction, 4 (1). pp. 11-18. ISSN 2590-3551

Text 5831-16160-1-PB.PDF Download (855kB)

Abstract

This paper aims to demonstrate the extraction of palm vein pattern features by local binary pattern (LBP) and its different recognition rate by two types of classification methods. The first classification method is by K-nearest neighbour (KNN) while the second method is by a support vector machine (SVM). Whilst SVM is optimized for direct classifications between two classes, the KNN is best for multi-class classifications. Based on the biometric recognition framework shared in this paper, both techniques shared comparable performance in terms of the recognition rate. The differences in the recognition rate can only be seen if the LBP features extracted for the classification are different. In general, a higher recognition rate can be achieved for palm vein pattern biometric system if all LBP bins are used for the classification, compared to if only selected features are used for the purpose. The best recognition rate that can be achieved by the three datasets demonstrated in this paper are 60%, 70% and 100% respectively for the CASIA, PolyU and self-dataset. It shows that different input dataset may behave differently even by using the same machine learning approach in its biometric recognition process.

Item Type:	Article
Uncontrolled Keywords:	Biometric recognition, K-nearest neighbour (KNN), Local binary pattern, Support vector machine (SVM), Pattern Biometric Recognition, Machine Learning
Divisions:	Faculty of Electronics and Computer Engineering
Depositing User:	Sabariah Ismail
Date Deposited:	16 Apr 2021 15:46
Last Modified:	16 Apr 2021 15:46
URI:	http://eprints.utem.edu.my/id/eprint/25097
Statistic Details:	View Download Statistic

Actions (login required)

View Item