PSILO-XAI-ResDCN: Handwritten Character Recognition Using Optimized Explainable Residual Deep Learning Mechanism

In recent decades, Handwritten character recognition (HCR) has gained significant attention due to its applicability in the historical document digitalization, automated data entry, and so on. The challenges faced by the conventional HCR include variable styles of handwriting, noise sensitivity, and lack of extensive-labeled datasets. In this research, the Paniscus Social Interaction Learning optimization-enabled explainable Artificial Intelligence-based Hybrid ResNet Deep Convolutional Neural Network (PSILO-XAI-ResDCN) is proposed for HCR. The proposed PSILO-XAI-ResDCN model incorporates the PSILO optimization to fine-tune the layer hyperparameters of ResDCN model. Moreover, the ResDCN extracts the deep hierarchical features from the character image, which enhances the model’s flexibility, and thereby offers high learning efficiency. In addition, the introduction of XAI with ResDCN model ensures the transparency and interpretability of the model in decision-making. Experimental results demonstrate that the proposed model achieves superior performance with the higher accuracy, precision, and recall values of 95.59%, 95.63%, and 95.56%, respectively, with the Devanagari character recognition dataset.