e-Prime: Advances in Electrical Engineering, Electronics and Energy (Sep 2023)
Comparative analysis, classification, and segmentation of the handwritten Gujarati conjuncts depending on the structural properties of the constituent characters
Abstract
This research paper presents a comprehensive analysis, classification, and segmentation of Gujarati conjuncts, with the aim of providing a deeper understanding of the intricate conjuncts in the Gujarati script. The study investigates the coverage area and joining patterns of conjuncts to categorize them based on their distinct structural properties. The coverage area refers to the spatial extent of a conjunct and is classified into three categories: full box, upper half box, and lower half box characters. The joining patterns offer insights into how consonants are connected or merged within a conjunct, including possibilities such as horizontal lines, curves. Accurate segmentation of conjuncts is crucial for retrieving their constituent components. This paper also discusses a segmentation algorithm that considers information from neighboring pixels, as well as the joining patterns and coverage area of conjuncts. The research study incorporates 728 frequently used handwritten conjuncts of the Gujarati script. Experimental analysis is conducted on a substantial dataset of 45,000 conjuncts. The experimental results demonstrate that conjuncts falling into the lower half box category or those connected with a horizontal line or a curve exhibit the highest success rate of over 85%. Furthermore, statistical analysis reveals that the success rate remains consistent and comparable across the various character groups, providing further support for the findings.
