Ligature categorization based Nastaliq Urdu recognition using deep neural networks

• Published in: Computational and mathematical organization theory Vol. 25; no. 2; pp. 184 - 195
• Main Authors: Rafeeq, Muhammad Jawad, ur Rehman, Zia, Khan, Ahmad, Khan, Iftikhar Ahmed, Jadoon, Waqas
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.06.2019; Springer Nature B.V
• Subjects: Acknowledgment; Artificial Intelligence; Artificial neural networks; Business and Management; Classification; Clustering; Dropping out; Learning; Management; Methodology of the Social Sciences; Neural networks; Noise; Operations Research/Decision Theory; Organization theory; Recognition; Regularization; S.i.: Cmkbo; Segmentation; Sociology; Nastaliq; Deep neural network; Ligatures; Categorization; Classification
• ISSN: 1381-298X; 1572-9346
• DOI: 10.1007/s10588-018-9271-y

The cursive nature, Nastaliq writing style and a large number of different ligatures make ligature recognition very difficult in Urdu. In this paper, we present a segmentation-free approach to holistically recognize Urdu ligatures. We first generate a rich dataset which contains 17,010 ligatures with different orientation and different degrees of noise. Secondly, the ligatures are clustered (categorized) in order to reduce the search space and make the learning robust. Finally, we employ a deep neural network with dropout regularization to classify ligatures. The detailed experiments show that a deep neural network with dropout regularization and clustering of ligatures significantly enhances the classification accuracy.