Arbain, Nur Atikah (2024) Enhanced triangle geometry shape features using multi-stage feature extraction for Arabic handwritten word recognition. Doctoral thesis, Universiti Teknikal Malaysia Melaka.
![]() |
Text (24 Pages)
Enhanced triangle geometry shape features using multi-stage feature extraction for Arabic handwritten word recognition.pdf - Submitted Version Download (800kB) |
![]() |
Text (Full Text)
Enhanced triangle geometry shape features using multi-stage feature extraction for Arabic handwritten word recognition.pdf - Submitted Version Restricted to Registered users only Download (6MB) |
Abstract
Handwritten word is one of the current manuscript studies that has been intensively researched for many years because it is difficult to identify the styles, patterns, and signatures, especially the Arabic handwriting word. The Arabic letters have special unique characteristic such as dotted letters which can be divided into one dot, two dots and three dots (ظ, ي, ش), the writing direction and certain letters in Arabic script exhibit shape variations when connected to adjacent letters within a word. The Arabic word contains more than one character, which makes it quite a challenge due to the difficulties in recognizing the letter connections, the position-dependent letter shape, and the different writing style, which has led many researchers to separate each Arabic letter from the word, including the dots of the Arabic letters for the purpose of recognition. The Arabic dot letters are often considered as noisy features due to their irregular position on word which resulted in their removal during the preprocessing process, which may lead to incorrect data being extracted during feature extraction. The effect of eliminating dots in Arabic recognition is important because dots are used to distinguish Arabic characters that have a similar letter, such as (ب ت, ث). This problem can be addressed by applying a global technique that considers the entire image of the Arabic word for feature extraction. Among the image features extracted, feature-based shape has the potential to recognize Arabic word handwriting based on geometric shapes. The triangle geometry method is applied, where the sides of the length of the triangle can be used as feature-based shape by using the appearances of triangle geometry to represent the Arabic word handwriting. Moreover, the multi-stage feature extraction is employed to extract various features such as angles and points of the triangle geometry shape. This study aims to propose the enhanced triangle geometry shape features for Arabic word handwriting recognition using multi-stage feature extraction. The objectives include to propose the triangle geometry shape features to represent the Arabic handwritten word images, to propose a method named Multi-Stage Feature Extraction of Triangle Geometry Shape Appearance Method (MFeTSA) to extract the triangle geometry shape features, and to improve prior straight-line problem solution by using proposed technique of Triangle Coordinate Point Rearrangement (TCPR). A dataset of Arabic (AHDB) and Iran (Iranshahr) is used for this study. The experimental results are obtained by comparing the classification accuracy results with Support Vector Machine, with promising results with high accuracy for both datasets AHDB (76.3825%) and Iranshahr (63.1563%). Statistical validation is used to check the quality of the proposed method, whether it is better than previous methods that use the same triangle geometry method to extract features from triangle geometry. Independent samples t-test is used to validate the proposed method, where the result based on the significant p-values shows that the proposed method is effective for both datasets of Iranshahr and AHDB.
Item Type: | Thesis (Doctoral) |
---|---|
Uncontrolled Keywords: | Handwritten word recognition, Arabic handwriting, Feature extraction |
Subjects: | T Technology > T Technology (General) T Technology > TA Engineering (General). Civil engineering (General) |
Divisions: | Library > Tesis > FTMK |
Depositing User: | Muhamad Hafeez Zainudin |
Date Deposited: | 17 Mar 2025 11:59 |
Last Modified: | 17 Mar 2025 11:59 |
URI: | http://eprints.utem.edu.my/id/eprint/28559 |
Statistic Details: | View Download Statistic |
Actions (login required)
![]() |
View Item |