Please use this identifier to cite or link to this item: https://hdl.handle.net/10356/65810
Title: Enhance face recognition accuracy by statistical regularization & dimensionality reduction
Authors: Phoo, Ngon Thin
Keywords: DRNTU::Engineering::Electrical and electronic engineering::Electric power
Issue Date: 2015
Abstract: Face recognition has been a very popular research for several years with the increasing demand for a fast, accurate & robust classification engine. Several face recognition technologies with improved modifications have been developed. However, the computational burden of facial image with high dimensionality and overfitting issues due to inadequate information of real population still remains a demanding research topic. With the study of most powerful statistical analysis methods, the roles of Linear Discriminant Analysis and Principal Analysis in dimensionality reduction are sufficiently understood and investigated. Moreover, regularization methods such as eigen spectrum regularization with constant-addition and Eigen-feature Regularization and Extraction approach are also investigated. By applying the above dimensionality and statistical regularization methods, different versions of Mahalanobis classifiers are implemented using MATLAB to investigate the performance of classification accuracy on facial image databases. Based on the performance result, their corresponding quality and limitations are well compared and discussed in this project.
URI: http://hdl.handle.net/10356/65810
Rights: Nanyang Technological University
Fulltext Permission: restricted
Fulltext Availability: With Fulltext
Appears in Collections:EEE Student Reports (FYP/IA/PA/PI)

Files in This Item:
File Description SizeFormat 
FYP Report - Phoo Ngon Thin.pdf
  Restricted Access
Enhance Face Recognition Accuracy by Phoo Ngon Thin1.66 MBAdobe PDFView/Open

Page view(s) 50

124
checked on Sep 25, 2020

Download(s) 50

8
checked on Sep 25, 2020

Google ScholarTM

Check

Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.