Handwritten Gujarati Character Recognition Using Deep Learning Architectures

Abstract

Data management is becoming increasingly difficult in modern society due to the everyday generation of enormous volumes of documents from various industries, including banking, healthcare, education, and business. The accumulation of these physical documents, whether in printed or handwritten form, necessitates substantial storage space and imposes an intricate burden on organisational systems. Moreover, the manual effort required for cataloguing, updating, and retrieving data is laborious and time-consuming. However, the emergence of Optical Character Recognition (OCR) technology has completely changed this procedure by quickly and precisely transferring handwritten or printed text into digital format, making it instantly accessible and manipulable by automated systems. Consequently, OCR technology is pivotal in streamlining document management processes, optimising resource utilisation, and enhancing operational efficiency across diverse sectors. newline newlineOptical Character Recognition (OCR) algorithms are crucial in the translation process, as they analyse the visual patterns and features of individual characters, identifying them through their unique shapes, sizes, and spatial configurations. While Printed Character Recognition focuses on identifying and interpreting characters in printed text, it encounters challenges such as font variations, smudges, or faded ink, which can impede accurate recognition. However, Handwritten Character Recognition (HCR) poses even greater complexities due to diverse writing styles and lack of uniformity in shape, size, and spacing, making deciphering a complex task. Factors like variations in slant, curvature, and stroke thickness further compound the difficulty, along with the need to differentiate between similar characters. Handwritten Character Recognition remains a challenging area of research, influenced by individual writing habits, styles, moods, and environmental factors. newline newlineUnlike printed characters, handwritten ones exhibit significant variations influenced by diverse factors