Face Recognition using Invariant Fourier Wavelet Descriptor

connectionviewAI and Robotics

Nov 17, 2013 (5 years and 4 months ago)


Face Recognition using Invariant Fourier Wavelet Descriptor


This project demonstrates the requirement of biometrics in security systems and advantages of
face recognition over other methods. Because of naturalness, face recognition is more accept
able than most
biometrics. The proposed face recognition system uses invariant Fourier Wavelet descriptor in extracting
features with a very few preprocessing stages. The system developed is invariant to translation and rotation
and gives very high rec
ognition rate. The system is tested on ORL database of 400 images and our own
database of 100 images. A recognition rate of 97.5% to 99% was obtained during the simulation.

Keywords: Biometric, Face recognition, Fourier wavelet descriptor, Recognition rat


Older security mechanisms employed person’s possession (key) or person’s
knowledge (password) as a way to secure their property, but the disadvantage with these
security systems is that mechanical objects like key can be duplicated and
passwords can be hacked, so privacy can be lost. The later systems used both possession
and knowledge (ATM) but this was found to be insecure in some cases. As security is a
major concern in many areas biometrics can be more effective. Among biom
etrics, face
recognition has gained importance due to its appealing characteristics. Face recognition
(FR) is the preferred mode of identity recognition amongst various biometrics since it is
natural, robust and unintrusive. Many algorithms have been propo
sed to implement face
recognition system based upon appearance or geometry or a combination of these two. The
eigenface approach is accepted by many commercially available face recognition systems.
This project emphasizes on “Fourier
Wavelet Descriptor” me
thod which is not only
translation, rotation and scaling invariant but also has multi
resolution matching ability.

In this project, a 2
D based invariant Fourier Wavelet descriptor for face recognition
has been proposed. Though Fourier descriptor has been

a powerful tool for pattern
recognition, frequency information obtained from the Fourier descriptor is global and a local
variation of the shape can affect all the Fourier coefficients. Wavelets are mathematical
functions that split up data into different

frequency components, and then analyze each
component with a resolution matched to its scale. They have advantages over traditional
Fourier methods in analyzing physical situations when the signal contains discontinuities
and sharp spikes. But they are no
t translation invariant. But Fourier descriptor is translation
invariant. Thus combining Fourier descriptor and Wavelet descriptor, a system is proposed
which is not only invariant under translation, rotation and scaling but also has a
multiresolution matc
hing ability.