IDENTITY RECOGNITION OPTIMIZATION BASED ON LBP FEATURE EXTRACTION

Abstract

ABSTRACT

 Unimodal systems have limited information that can be used for identity recognition systems. The multimodal system was created to improve the unimodal system. The multimodal system used in this study is the combination of the face and palms at the matching score level. Matching scores is done using the Weighted Sum Rule method. Extract features from each sample using the Local Binary Pattern (LBP) method. Meanwhile, large data dimensions are reduced by using the Principal Component Analysis (PCA) method. The distance between face and palm data is measured using the closest distance, namely the Euclidean Distance method. Benchmark dataset using ORL, FERET and PolyU. Based on testing on each database, an accuracy rate of 98% (ORL and PolyU) and 95% (FERET and PolyU) is obtained. The test results show that the multimodal system using the Hybrid method (PCA and LBP) biometric system runs well and optimally.

 Keywords: Artificial intelegency, recognition, LBP, multimodal