Full and Partial Connected Local Binary Pattern Analysis for Finger Knuckle Classification using Support Vector Machines

Abstract

Hand-based biometric through finger knuckle print has emerged as a more reliable and promising alternative to conventional personal identification solutions. One of the most collaborative keys using the finger knuckle print is its ease to use, widespread public acceptance, and mostly non-intrusive procedure. This paper presents a method to retrieve texture analysis performance by making several orientations for calculating the interconnection between pixels using the fundamental local binary pattern (LBP). The study comprises two fundamental processes: feature extraction and SVM classification. The local Binary Pattern (LBP) algorithm comprises fully, and two orientations of partial LBP analysis are used to extract the feature from the images. We generated three version codes of LBP descriptors to compare the classification accuracy with the SVM classifier. The assessment found that different performance with an average of more than 90% has been achieved for all the suggested orientations.