DSpace
 

KNUSTSpace >
Theses / Dissertations >
College of Science >

Please use this identifier to cite or link to this item: http://hdl.handle.net/123456789/9267

Title: Comparative study on face recognition techniques: Principal Component Analysis and Linear Discriminant Analysis
Authors: Peprah, Frank
Issue Date: 17-Oct-2016
Abstract: Face Recognition System employs a variety of feature extraction (projection) techniques which are grouped into Appearance-Based and Feature-Based. In a vast majority of the studies undertaken in the field of Face Recognition special attention is given to the Appearance-Based Methods which represent the dominant and most popular feature extraction technique used. Even though a number of comparative studies exist, researchers have not reached consensus within the scientific community regarding the relative ranking of the efficiency of the appearance-based methods (LDA, PCA etc) for face recognition task. This paper studied two appearance-based methods (LDA, PCA) separately with three (3) distance metrics (similarity measures) such as Euclidean distance, City Block & Cosine to ascertain which projection-metric combination was relatively more efficient in terms of time it takes to recognise a face. The study considered the effect of varying the image data size in a training database on all the projection-metric methods implemented. LDA-Cosine Distance Metric was consequently ascertained to be the most efficient when tested with two separate standard databases (AT & T Face Database and Indian Face Database). It was also concluded that LDA outperformed PCA.
Description: A thesis submitted to The Institute of Distance Learning, KNUST, in partial fulfilment of the requirements for the award of Master of Science Degree in Information Technology, 2015
URI: http://hdl.handle.net/123456789/9267
Appears in Collections:College of Science

Files in This Item:

File Description SizeFormat
FRANK PEPRAH.pdfMain Work3.14 MBAdobe PDFView/Open
FRANK PEPRAH Source Code.pdfSource Code99.77 kBAdobe PDFView/Open

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

 

Valid XHTML 1.0! DSpace Software Copyright © 2002-2010  Duraspace - Feedback