Jawi sub-word recognition system using window-based segmentation-free approach / Roslim Mohamad

Mohamad, Roslim (2018) Jawi sub-word recognition system using window-based segmentation-free approach / Roslim Mohamad. In: The Doctoral Research Abstracts. IPSis Biannual Publication, 13 (13). Institute of Graduate Studies, UiTM, Shah Alam.

Abstract

The presence of features such as cursive, diversity of writing styles and sizes of characters in a Jawi text, ligature, and vertical overlapping make the recognition of Jawi handwritten text to be difficult. For the recognition system based on model development using windows, the presence of such features cause the resulting model to be less consistent, where it produces a different sequence of primitive structures of words/ sub-words from the same lexicon. To overcome the inconsistency model problem, a handwritten Jawi text recognition system based on a subword model has been developed. The proposed modeling technique which is known as Selection Segmentation-Free (SSF) separates core and connection structure of a sub-word into a different window. The resulting window will go through a selection process to determine the windows that will be used to represent the sub-word model. In order to increase accuracy and efficiency of the representation feature, two categories of features which are known as primary and secondary features were extracted from each of the selected windows. Primary feature were extracted using Window Code Representation (WCR) technique from main structure. Secondary feature for supporting the primary feature were extracted from dot and main structure. For the experiment purposes, a total of 1200 sub-words of 80 lexicons were used. Each lexicon is randomly selected and divided into three sets. Three experiments to evaluate the performance of SSF, WCR and combination of primary and secondary feature techniques were conducted. The three techniques are combined to represent the proposed system and compared with the comparison system introduced by Remon (2009). Comparison result shows that the recognition rate of proposed system (84.8%) is better than comparison system (79.1%).

Metadata

Item Type: Book Section
Creators:
Creators
Email / ID Num.
Mohamad, Roslim
UNSPECIFIED
Subjects: L Education > LB Theory and practice of education > Educational technology
Q Science > QA Mathematics > Instruments and machines
Divisions: Universiti Teknologi MARA, Shah Alam > Institut Pengajian Siswazah (IPSis) : Institute of Graduate Studies (IGS)
Series Name: IPSis Biannual Publication
Volume: 13
Number: 13
Keywords: Abstract; Abstract of thesis; Newsletter; Research information; Doctoral graduates; IPSis; IGS; UiTM
Date: 2018
URI: https://ir.uitm.edu.my/id/eprint/20699
Edit Item
Edit Item

Download

[thumbnail of Abstract Only]
Preview
Text (Abstract Only)
ABS_ROSLIM MOHAMAD TDRA VOL 13 IGS_18.pdf - Submitted Version

Download (418kB) | Preview

ID Number

20699

Indexing

Statistic

Statistic details