Palm print recognition using competitive hand valley detection, local binary pattern and probabilistic neural network

Biometric is a method to identify human using the specific feature based on part of human body. Nowadays, biometric is widely used because of the performance to identify human. The palm-print data-set used in this research were manually collected that consist of 110 people with 10 images for each person. Palm-print recognition system consist of : region of interest (ROI) extraction using Competitive Hand Valley Detection (CHVD) method, feature extraction process using Local Binary Pattern (LBP) method and classification process using Probabilistic Neural Network (PNN). In this research we proposed area segmentation technique for LBP that consist of overlap and non-overlap segment, by using segmentation the feature will became more unique and distinctive. In 60 people (50 known and 10 unknown) the best system accuracy achieve was 99.83%, while for 110 people (100 known and 10 unknown) the best accuracy achieve was 91.91%. The best performance above achieve with four overlap segment schemas, 8 check points and the smoothing parameter set at 0.03.