Feature extraction based Sinhala Handwriting Recognition 

Sinhala language is used by Sinhalese, the major ethnic group native to Sri Lanka, that comprise the majority of the population (75%) in the country. Converting the written paper documents in Sinhala to electronic format requires an enormous amount of human labour. If the conversion can be automated using handwriting recognition, it would increase efficiency  and a significant cost can be saved. Character recognition is a procedure of converting images of handwritten, typewritten or  printed text into machine encoded code or text. This research is based on offline recognition of Sinhala handwritten text.  Handwritten data can be in structured, unstructured handwriting, constrained or unconstrained formats. Depending on the format segmentation can be a challenging. Poor recognition of all the consonants and vowels used in modern Sinhala writing, misclassification of nearly similar character shapes, lack of Sinhala handwritten text dataset for general applications can be  considered as other technical challenges in identifying Sinhala handwritten text. This work  addresses offline handwriting recognition. Most of the existing recognition techniques lack the recognition of characters with modifier symbols. In this research a new technique is explored  for Sinhala handwriting recognition including the recognition of characters with modifier symbols.  A combination of contour tracing, distance profiles, projection profiles and partial distance profiles are used and word recognition shown to be are improved by a lexeme based  correction mechanism. The proposed method recognizes for 283 different character modifier  combinations with a recognition rate of 79% which is a very high number of combinations compared to the other researches done on Sinhala handwriting recognition.