An Approach of Likelihood Ratio Score Fusion for Appearance and Shape based Face Recognition

Abstract

Abstract
This paper deals with an approach of likelihood ratio based score fusion technique where appearance and shape based facial features have been used to enhance the efficiency of existing face recognition system. Active Shape Model (ASM) has been used to extract the appearance and shape based facial features. Two different types of features have been used in this work in such a way that when the appearance based feature contains noise then shape based feature retains its efficiency and the overall performance will be maintained in a satisfied level. Same thing will be happened when the shape based feature degrades with more noise compared with appearance based feature. Hidden Markov Model (HMM) has been applied for learning and testing model for the proposed face recognition system. Likelihood ratio of each appearance and shape modality have been measured using reliability measurement technique and log likelihood ratio scores are fused to evaluate the final face recognition performance. To measure the performance of the proposed system, Olivetti Research Laboratory (ORL) database has been used with addition of various noises at different level. Experimental results and performance analysis show the superiority of the proposed face recognition system.

Keywords: likelihood score fusion, appearance and shape based facial features, active shape model, principal component analysis, hidden markov model