Handwritten Text Detection System in Sinhala

Abstract

"Computer vision has been expanded to identify both handwritten documents and printed documents to improve human-computer interactions. However, in the case of Asian languages, this is still a major controversy. Sinhala language characters recognition is still a work in progress among Asian languages, as Sri Lanka is the only nation that utilizes Sinhala as its official language. In comparison to other commonly spoken languages, Sinhala characters have circular, intricate forms with connections between characters. There are 60 essential characters in the Sinhala language (non-cursive). Modifiers which can be applied to primary characters further enhance the Sinhala character set. As a result, recognizing Sinhala handwritten characters is a complicated process. Pattern matching and image processing techniques are used in most of the character segmentation and recognition research projects. These methods, on the other hand, are unable to adjust to the changes. This research is aimed at a Convolutional Neural Network-based handwritten Sinhala character recognition system (CNN). CNN differs from other traditional methods to Sinhala handwritten character recognition in that it extracts features automatically throughout the training phase. The objective of this research is to use deep learning approach to develop a process for Sinhala handwritten character recognition. This research project will be able to segment characters in handwritten character and train a CNN model to accurately recognise Sinhala handwritten characters."