Reader System for Transliterate Handwritten Bilingual Documents

Abstract

India is a Multistate- Multilingual country. Most of the people in India used their state official language and English is treated as a binding language used for form filling or some official work. So there is a need to create a system which will convert the handwritten bilingual document into digitized form. This paper aims at development of reader system for handwritten bilingual (Marathi-English) documents by recognizing words. This facilitates many applications such as Natural language processing, School, Society, Banking, post office and Library automation. The proposed system is divided into two phases. The first phase focuses on recognition of handwritten bilingual words using two different feature extraction methods including combination of structural and statistical method and Histogram of Oriented Gradient Method. K-Nearest Neighbor classifier is used for recognition. This classifier gives 82.85% recognition accuracy using Histogram of Oriented Gradient method. The dataset containing 4390 words collected from more than 100 writers. The second phase focuses on digitization and transliteration of recognized words and conversion of transliterated text into speech, which is useful in the society for visually impaired people.