An efficient recognition of Handwritten Kannada Script using Variable Attention-based Coati Integrated Bi Directional Long Short-Term Memory
S. R. Siddanna,Y. C. Kiran
DOI: https://doi.org/10.1007/s11042-024-18999-y
IF: 2.577
2024-03-28
Multimedia Tools and Applications
Abstract:Handwritten Character Recognition (HCR) remains a challenging task in pattern identification, and there is a lack of extensive research on script-independent models. This is caused due to the variation between handwriting, character structure, heterogeneity of scripts, absence of public datasets, and conventional research focus on handcrafted extraction approaches. Besides, Deep Learning (DL) has greater achievement in various fields of pattern recognition, particularly HCR, and provides end-to-end learning. Moreover, several approaches require large amount of data for training, and increased computational cost. In this research work, novel Variable Attention-based Coati integrated Bi directional Long Short-Term Memory (VACBLSTM) is proposed for recognition of handwritten Kannada characters This research work involves four stages. The preprocessing stage uses guided box filtering technique for improving image quality. Next, double segmentation is performed to identify each character individually. Then, features such as statistical, global transformation and geometrical and topological features are extracted. Finally, the variable attention integrated bi-directional long short-term memory is proposed for character recognition. In this, Coati Optimization Algorithm (COA) is employed to maximize the accuracy level of the recognition network. The introduced technique is implemented in python. The implemented method is tested on self-created handwritten dataset and it provides highest accuracy of 97% over other classifiers. The implemented method is suitable for Kannada handwritten character identification.
computer science, information systems, theory & methods,engineering, electrical & electronic, software engineering