An efficient feature extraction technique for palm-print recognition system

Abstract

Palm-print recognition has been investigated over many decades. Palm-print recognition has many phases in which region mapping and feature extraction play an important role. In this paper, we propose a region mapping model using Gradient Bidirectional Boundary Tracing (GBBT) method. We also propose a novel scheme of palm-print recognition using skeleton morphological (SM) operations to extract the combined shape, texture and colour features of the gray-scale palm-print image. To this extent we explore the idea of GBBT and SM that comprises of 11 features whose responses on the given palm-print image are processed independently and classified using Multi-SVM (MSVM). The proposed model is compared with KNN and SVM to demonstrate its efficiency. Extensive experiments are carried out on different large-scale publicly available palm-print databases in terms of its sensitivity specificity and accuracy. The experimental results indicate that our approach produces 2.509% and 2.6342% more efficacy than SVM and KNN methods, respectively.