Optimized Deep Learning Framework for Arabic Handwritten Character Recognition Using Coati Optimization Algorithm

Automatic handwritten character recognition is one of the Artificial Intelligence (AI) applications considered a stimulating field of research and significant in different domains. Numerous studies have concentrated on detecting English handwritten characters, with several addressing Arabic handwriting due to the diverse shapes of characters based on their positions in words. Despite these efforts, there are still opportunities to improve Arabic handwritten letter recognition. Deep learning (DL) presents a promising approach, as it can effectively handle large volumes of unlabeled raw data, which is increasingly available today. Therefore, this article presents an Optimized Deep Learning Framework for Arabic Handwritten Character Recognition using the Coati Optimization Algorithm (ODLF-AHCRCOA) technique. The ODLF-AHCRCOA technique aims to recognize the distinct handwritten characters in Arabic. To accomplish this, the Sobel filter (SF) is utilized for image preprocessing to enhance edge detection. The modified LeN et-5 method is employed for feature extraction to capture high-dimensional representations of character features. Moreover, the long short-term memory (LSTM) classifier is used to classify Arabic handwritten characters. To improve the classification accuracy, the coati optimization algorithm (COA) is implemented for parameter tuning, optimizing the LSTM hyperparameters of the model for enhanced convergence and performance. The simulation study of the ODLF-AHCRCOA method is performed, and the outcomes are examined using varying features. The experimental validation of the ODLF-AHCRCOA method exhibited a superior accuracy value of 98.60% over other approaches.