![](/images/graphics-bg.png)
Arabic writer identification based on power spectrum and laws' filters methods
Dissertant
Thesis advisor
Comitee Members
Rashid, Abd Allah Abd al-Ali
Haddad, Bassam
Salamah, Walid Khalid
Hammu, Bassam
University
Arab Academy for Financial and Banking Sciences
Faculty
The Faculty of Information Systems and Technology
Department
Computer information systems
University Country
Jordan
Degree
Ph.D.
Degree Date
2006
English Abstract
Many methods have been reported for handwriting-based writer identification.
None of these methods assumes that this technique is possible in Arabic.
In this dissertation, we present two new methods for feature extraction of handwriting texture.
These two methods are based on signal processing techniques : the first is based on Power Spectrum (PS) ; whereas the second is based on Laws' Filtering (LF).
Their effectiveness compared to multiple channels (Gabor) filters and the gray-level co-occurrence matrix (GLCM) is shown.
The Gabor and the GLCM are well known methods proved to yield high performance in writer identification in Latin.
In our experimentations, texture features are extracted for wide range of frequency and orientations.
This is mainly due to the nature of Arabic handwritten which constitutes more spreaders compared to Latin.
The most discriminate features are selected using a proposed model for feature selection based on hybrid Support Vectors Machine (SVM) / and Genetic Algorithm (GA) techniques.
Three classification methods are used for classification, namely; Linear Discriminate Classifier (LDC), Support Vectors Machine (SVM), and the K nearest-neighbors (K-NN) classifier.
Experiments are made using Arabic handwritings from 20 different people and very promising results of 90 % identification rate is achieved.
Main Subjects
Information Technology and Computer Science
Topics
No. of Pages
97
Table of Contents
Table of contents.
Abstract.
Chapter One : introduction.
Chapter Two : normalization.
Chapter Three : feature extraction using texture analysis.
Chapter Four : feature ranking and selection.
Chapter Five : classification.
Chapter Six : experiments and results.
Chapter Seven : conclusions and future work.
References.
American Psychological Association (APA)
al-Dumur, Ayman. (2006). Arabic writer identification based on power spectrum and laws' filters methods. (Doctoral dissertations Theses and Dissertations Master). Arab Academy for Financial and Banking Sciences, Jordan
https://search.emarefa.net/detail/BIM-306829
Modern Language Association (MLA)
al-Dumur, Ayman. Arabic writer identification based on power spectrum and laws' filters methods. (Doctoral dissertations Theses and Dissertations Master). Arab Academy for Financial and Banking Sciences. (2006).
https://search.emarefa.net/detail/BIM-306829
American Medical Association (AMA)
al-Dumur, Ayman. (2006). Arabic writer identification based on power spectrum and laws' filters methods. (Doctoral dissertations Theses and Dissertations Master). Arab Academy for Financial and Banking Sciences, Jordan
https://search.emarefa.net/detail/BIM-306829
Language
English
Data Type
Arab Theses
Record ID
BIM-306829