Combining wavelets with HMM for face recognition

Dissertant

Rasul, Hamid Farhan

Thesis advisor

Abbas, Iyad Ibrahim

University

University of Technology

Faculty

-

Department

Department of Electrical Engineering

University Country

Iraq

Degree

Master

Degree Date

2011

English Abstract

-Automatic human identification, based on facial features is one of the most important research fields.

The face is naturally the most suitable method of identification for security related applications.

This thesis presents an efficient and robust face recognition method based on the Hidden Markov Model and the simplest type “Haar”of the Discrete Wavelet Transform.The Discrete Wavelet Transformhas many advantages,such as thereduction in the data processing and noise elimination.

The facial features are extracted using the 2-level and 3-level of the Discrete Wavelet Transformapplied to the entire face image.

The Hidden Markov Model topology used in the proposed work is one dimensional ergodic, which represent the simplest and robust type of Hidden Markov Model.

A novel method is used to select the training images from the database implemented by choosing the images that havethe odd identifying numbers, and after several trials, some of these images are replaced according to the recognition results.

The proposed work achieves the maximum recognition rate (100%) in both 2-level and 3-level methods using the facial database obtained from Olivetti Research Laboratory in Cambridge University.

The conditionsof obtaining the best results are; fivetraining images for each person, vertical image segmentation method, block of size (5x4) in 2-level method and block of size (5x4) and (4x4) in 3- levelmethod.The 3-level Discrete Wavelet Transform is the most robust method against image noise, such as White Gaussian noise and remains achieving the perfect recognition rate even though the noise variance is 0.01.

Furthermore, the 3- level method reduces the computational complexity significantlyin comparison with the 2-level method.

The time consumed in recognizing one image in 3-level method is 0.24 second, while in 2-level method is 0.8 second.

Main Subjects

Electronic engineering

• APA
• MLA
• AMA

Language

English

Data Type

Arab Theses

Record ID

BIM-305049