Automated Handwritten Letter Recognition using Optimized Deep Learning Model

Forensic evidence, bank signature validation, and postal code recognition are just a few of the growing number of applications for handwritten charm recognition technology. The prevalence of numerous mysterious handwritten characters, massively characters, and a lack of academic focus mean that Pashto hand-written character acknowledgement (PHCR) remains largely unresolved despite advances in object recognition technology. For the first time in a controlled setting, we offer an efficient DeepNet model for recognizing Pashto hand-written characters. To begin, a new Pashto handwritten character data set called “Poha” is built, which has 44 characters. Noise reduction and deep fusion image processing methods are used to prepare the text for optimization. The suggested model contains of 11 layers, eighter of which are three of which (FC). When applied to a DeepNet model, Adaptive Elephant Herd Optimization (AEHO) improved the accuracy of the parameters above that of standard deep models. Additionally, the suggested model is compared and contrasted with a number of benchmark Deep Learning (DL) models that have been applied to Poha. With a test accuracy of $97 \%$ for PHCR, the acquired experimental findings demonstrate that the suggested model is superior to existing models. The findings suggest that our approach has potential for use in automated PHCR and handwritten character recognition.