An improved face recognition system based on soft biometric modals

Biometric systems which recognize individuals by facial features or characteristics have become increasingly popular in web applications like Facebook or Twitter. This allows auto-tagging of photographs uploaded into individual’s user account. The most basic block of any biometric recognition system...

Full description

Saved in:
Bibliographic Details
Main Author: Xie, Collin Ding Chang
Other Authors: Teoh Eam Khwang
Format: Final Year Project
Language:English
Published: 2010
Subjects:
Online Access:http://hdl.handle.net/10356/40552
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
id sg-ntu-dr.10356-40552
record_format dspace
spelling sg-ntu-dr.10356-405522023-07-07T17:13:33Z An improved face recognition system based on soft biometric modals Xie, Collin Ding Chang Teoh Eam Khwang School of Electrical and Electronic Engineering DRNTU::Engineering::Electrical and electronic engineering::Control and instrumentation Biometric systems which recognize individuals by facial features or characteristics have become increasingly popular in web applications like Facebook or Twitter. This allows auto-tagging of photographs uploaded into individual’s user account. The most basic block of any biometric recognition system would be the presence of a human trait which exhibits distinctiveness, permanence, universality, collectability, acceptability and resistance to circumvention. Automatic face recognition scans an individual’s face and matches it against a library of known faces. The need to improve the performance of face recognition arises due to its inability to provide perfect recognition by possessing all the properties of a human trait (facial) used in human recognition. This final year project proposes the use of soft biometric traits to complement the identity information obtained from primary biometric identifier like face. Despite the fact that characteristics like skin tone, hair colour, shirt colour and gender are not as reliable and unique, they are still able to provide some information about the user. This report shows some soft biometric traits can be automatically extracted through the use of skin segmentation technique and modifications of the Viola-Jones Face Detection Algorithm. The results from these implementations and gender classification by adaptive boosting of Gabor features are integrated in the decision making process for the primary face recognition system. Weighted majority voting is applied and weighting the decisions of those more qualified experts heavier may further improve the overall performance of the face recognition system. The system uses scores to express the similarity between a pattern and a biometric template. In other words, the higher the fusion score, the higher the similarity is between them. A single threshold which separates the two groups of scores is utilized to identify between the genuine user and the imposters. By using more than one means of biometric identification, the multimodal biometric system is able to retain its high threshold recognition setting. In a comparative study done to substantiate the motivation of this project, it is found out that among the soft biometric traits chosen to enhance the primary face recognition system, improvement is most significant when differentiating siblings of opposite sex. The recognition results achieved an accuracy rate of approximately 80% as compared to 20% without soft biometrics. Such multimodal biometric system based on different biometric identifiers is proven to be more robust to non-universality and noise. Bachelor of Engineering 2010-06-16T07:07:31Z 2010-06-16T07:07:31Z 2010 2010 Final Year Project (FYP) http://hdl.handle.net/10356/40552 en Nanyang Technological University 124 p. application/pdf
institution Nanyang Technological University
building NTU Library
continent Asia
country Singapore
Singapore
content_provider NTU Library
collection DR-NTU
language English
topic DRNTU::Engineering::Electrical and electronic engineering::Control and instrumentation
spellingShingle DRNTU::Engineering::Electrical and electronic engineering::Control and instrumentation
Xie, Collin Ding Chang
An improved face recognition system based on soft biometric modals
description Biometric systems which recognize individuals by facial features or characteristics have become increasingly popular in web applications like Facebook or Twitter. This allows auto-tagging of photographs uploaded into individual’s user account. The most basic block of any biometric recognition system would be the presence of a human trait which exhibits distinctiveness, permanence, universality, collectability, acceptability and resistance to circumvention. Automatic face recognition scans an individual’s face and matches it against a library of known faces. The need to improve the performance of face recognition arises due to its inability to provide perfect recognition by possessing all the properties of a human trait (facial) used in human recognition. This final year project proposes the use of soft biometric traits to complement the identity information obtained from primary biometric identifier like face. Despite the fact that characteristics like skin tone, hair colour, shirt colour and gender are not as reliable and unique, they are still able to provide some information about the user. This report shows some soft biometric traits can be automatically extracted through the use of skin segmentation technique and modifications of the Viola-Jones Face Detection Algorithm. The results from these implementations and gender classification by adaptive boosting of Gabor features are integrated in the decision making process for the primary face recognition system. Weighted majority voting is applied and weighting the decisions of those more qualified experts heavier may further improve the overall performance of the face recognition system. The system uses scores to express the similarity between a pattern and a biometric template. In other words, the higher the fusion score, the higher the similarity is between them. A single threshold which separates the two groups of scores is utilized to identify between the genuine user and the imposters. By using more than one means of biometric identification, the multimodal biometric system is able to retain its high threshold recognition setting. In a comparative study done to substantiate the motivation of this project, it is found out that among the soft biometric traits chosen to enhance the primary face recognition system, improvement is most significant when differentiating siblings of opposite sex. The recognition results achieved an accuracy rate of approximately 80% as compared to 20% without soft biometrics. Such multimodal biometric system based on different biometric identifiers is proven to be more robust to non-universality and noise.
author2 Teoh Eam Khwang
author_facet Teoh Eam Khwang
Xie, Collin Ding Chang
format Final Year Project
author Xie, Collin Ding Chang
author_sort Xie, Collin Ding Chang
title An improved face recognition system based on soft biometric modals
title_short An improved face recognition system based on soft biometric modals
title_full An improved face recognition system based on soft biometric modals
title_fullStr An improved face recognition system based on soft biometric modals
title_full_unstemmed An improved face recognition system based on soft biometric modals
title_sort improved face recognition system based on soft biometric modals
publishDate 2010
url http://hdl.handle.net/10356/40552
_version_ 1772828278803922944