A Novel Approach: CNN-RNN and Bi-GRU for Handwritten Character Recognition
Sunitha S. Nair, P. Ranjit Jeba Thangaiah · International Journal of Electronics and Communication Engineering · 2024
Handwritten Character Recognition (HCR) plays a crucial role in converting handwritten content into machinereadable format with applications spanning various sectors such as historical document digitization, postal envelope address reading, form processing, and assisting individuals with disabilities. Recognizing handwritten characters is inherently challenging due to the diversity in writing styles and shapes and the presence of noise in the data. Recent advances in deep learning, particularly the utilization of Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), have greatly enhanced the accuracy and performance of HCR systems. The integration of a CNN-RNN model with Bidirectional Gated Recurrent Units (Bi-GRU) has shown great promise, achieving an impressive accuracy rate of 96.72%. The CNN component excels at capturing spatial features and character structures, while the RNN with Bi-GRU layers effectively models sequential dependencies in handwritten text. As technology continues to advance and more data becomes available, the future holds the promise of even more refined and powerful HCR models. This hybrid approach has the potential to automate processes, enhance data processing, and improve user experiences in a wide range of industries.