Authors :

Mayahi Ibrahim and Jameel Dhannoon

Address :

Department of Computer Science, Wasit University Iraq

Department of Computer Science, Al-Nahrain University Jadriyah, Baghdad

Abstract :

Handwritten character recognition in low-resource scripts is a challenging yet critical task for preserving linguistic diversity and enabling inclusive digital systems. This paper proposes a novel framework that combines Fuzzy Crab Optimization (FCO) with Convolutional Neural Networks (CNNs) to address this challenge effectively. Existing recognition methods often struggle with poor generalization, limited feature extraction, and high computational costs, especially when dealing with small or imbalanced datasets typical of underrepresented scripts. To overcome these limitations, the proposed FCO-CNN framework employs the bio-inspired Fuzzy Crab Optimization algorithm to optimize CNN hyperparameters and enhance feature selection, thereby improving learning efficiency and accuracy. The FCO component mimics crab behavior with fuzzy logic to guide the search process more intelligently, avoiding local minimum and accelerating convergence. The framework is applied to various low-resource scripts such as Santali, Meitei, and Tulu, enabling accurate recognition of handwritten characters even with minimal training data. Experimental results demonstrate that the proposed method outperforms traditional CNN and other metaheuristicoptimized models in terms of recognition accuracy by 98%, robustness by 96.6%, and computational efficiency by 98.1%. This framework offers a promising solution for developing intelligent educational tools, digital archives, and inclusive linguistic technologies.

Keywords :

Fuzzy Crab Optimization, Convolutional Neural Network, Handwritten Character Recognition, Low-Resource Scripts, Feature Optimization, Linguistic Preservation.

DOI :