Face Recognition in Mobile Devices

Only available on StudyMode
  • Topic: Facial recognition system, Pattern recognition, Template matching
  • Pages : 86 (15936 words )
  • Download(s) : 39
  • Published : January 27, 2013
Open Document
Text Preview
2010:040 CIV

MASTER'S THESIS

Face Recognition in Mobile Devices

Mattias Junered

Luleå University of Technology
MSc Programmes in Engineering
M edia Technology
D epartment of Computer Science and Electrical Engineering
Division of Signal Processing
2010:040 CIV - ISSN: 1402-1617 - ISRN: LTU-EX--10/040--SE

Face Recognition in Mobile Devices
Mattias Junered
Luleå University of Technology
March 2, 2010

Abstract
Recent technological advancements have made face recognition a very viable identification and verification technique and one reason behind its popularity is the nonintrusive nature of image acquisition. A photo can be acquired easily without the person even being aware of the process. The interest in biometrics by several governments for identifying possible criminals or verifying users for access control is steadily increasing. Other industries are also finding uses for face recognition techniques such as in entertainment systems and for robots that interact with humans.

Mobile phones are constantly improving and the majority are currently equipped with a digital camera. This facilitates taking a large amount of photos every day with a camera phone instead of a stand-alone digital camera. Using face recognition techniques on these images makes it possible to perform so called face tagging to tag images with the names of the photographed persons. This is convenient for sorting photos, creating albums or retrieving images of only a specific person. Having a stand-alone mobile application on the phone that performs these face recognition tasks on recently captured images is an interesting concept. The system can be trained on a set of images containing faces to become capable of automatically recognizing a person from the training set. However, many users have up to hundreds or even thousands of images on their mobile phones and training a system on the phone is prohibitively time-consuming on such devices today. Instead, face recognition can be performed on the client using already trained data transferred from a computer (server). This approach shows promising results and very good success rates. This article covers several methods that can improve results by making the system more robust.

Acknowledgements
This work would not have been possible without the support of Apostolos Georgakis as external supervisor at Ericsson AB. Many thanks to Jiong Sun at the EAB/TVV section for connecting the pieces in the phone application and additional help and support. The author would also like to thank the whole EAB/TV department for their cooperation in creating the internal image database and all the interesting discussions. Finally, thanks to Josef Hallberg internal supervisor, Kåre Synnes examiner at Luleå University of Technology and everyone else who helped out.

1

Contents
1

2

3

4

Introduction
1.1 Background . .
1.2 Chapter outline
1.3 Objective . . .
1.4 Goal . . . . . .

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

4
4
5
5
5

Related work
2.1 Face Recognition Vendor Test .
2.2 Algorithm categorization . . . .
2.2.1 Projection methods . . .
2.2.2 Statistical methods . . .
2.2.3 Graph matching methods
2.2.4 Neural network methods

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

.
.
.
.
.
.

6
6
8
9
9
10
11

Theory
3.1 History . . . . . . . . . . . . . .
3.2 Pre-processing...
tracking img