A Technique for Character Segmentation in Middle zone of Handwritten Hindi words using Hybrid Approach

Abstract

India is a country where people talk in multilingual and write in multi-script. Devanagari is one of the most popular scripts in India, which is used to write Hindi, Sanskrit, Sindhi, Marathi and Nepali Languages. This research work is performed on Hindi language. A large number of precious and essential documents are available in handwritten form, which needs to be converted into editable form. The existence of Optical Character Recognition (OCR) makes this task easier to convert handwritten text in editable form. Character segmentation is an important phase of OCR, which segment the characters from handwritten words. This enhances the accuracy of OCR system. In this paper a hybrid approach is used to segment the characters that contain single and multiple touching characters within a word. The proposed system is tested on a dataset of various handwritten words written by different writers. The dataset of proposed system contains more than 300 handwritten words in Hindi language. Accuracy of the proposed hybrid system is evaluated to 96% which is better than that of existing techniques.