Face recognition

• Main Author: Guo, JinJian.
• Other Authors: Chua Chin Seng
• Format: Final Year Project
• Language: English
• Published: 2009
• Subjects: DRNTU::Engineering::Electrical and electronic engineering::Electronic systems::Biometrics
• Online Access: http://hdl.handle.net/10356/17891
• Institution: Nanyang Technological University

This is a report for the Final Year Project held in the final year of study of the 4-year Electrical and Electronic Engineering course in Nanyang Technological University (NTU), starting on 1st August 2008. The schedule of the project can be found in Appendix A. Face recognition played an important part in daily lifestyle (iPhoto’09 [1]), security systems (Criminal Identification [2]), and biometric purposes etc. The high dependency of face recognition technology in today’s society has hence aroused much interest in researchers to develop a reliable face recognition system. The objective of this project is to develop an efficient and accurate face recognition system based on Principal Component Analysis (PCA), which is one of the earliest approaches to face recognition. The project was divided into two parts. In the first part an analysis is conducted to find the minimum number of eigenimages needed to provide accurate recognition results and subsequently a recognition system will be developed based on part one results and that will be the 2nd part of the project. A program for performing PCA was designed and analyzed with three sets of images. The first set consist of images with all types of variation, the second set consist of frontal images with varying expression and the last set of images consist of variation in face angle positioning, expression and facial detail. Recognition was based on nearest neighbor classifier and two distance measures were evaluated, which are the Euclidean and Manhattan distance measures. Highest recognition rate with respect to the lowest number of eigenimages used will be taken as final result. Result was found to be 30 and it was verified by analyzing it with a set of images with all variation excluding facial detail and by performing image reconstruction. A face recognition system is developed in part two, based on 30 eigenimages. Further results of the analysis will be explained and discuss in detail, together with the possible problems and providing possible solutions to it.