Offline Recognition of Malayalam Handwritten Text. (2015)

• Record Type: Journal Article
• Authors: Shanjana, C.; James, Ajay
• Is Part Of: Procedia technology. Volume 19(2015)
• Page Start: 772
• Page End: 779
• Publication Date: 2015
• Subjects: Optical Character Recognition -- Line Segmentation -- Word Segmentation -- Character Segmentation -- Horizontal Projection -- Vertical Projection -- Connected Component Analysis -- SVM Classifier ;; Technology -- Congresses; Technology -- Periodicals; Engineering -- Congresses; Engineering -- Periodicals; Engineering; Technology; Conference proceedings; Periodicals; 605
• DOI: 10.1016/j.protcy.2015.02.109
• Languages: English
• ISSNs: 2212-0173

Abstract

Abstract: This paper discusses a method to convert the handwritten Malayalam document into a computer recognized text file format. The scanned image of the handwritten Malayalam text document is given as input in this method and the output is the corresponding Malayalam text document. All the existing systems for Malayalam handwritten text are for recognizing individual characters separately and also they deal only with a subset of Malayalam characters. Here, the handwritten document should be first segmented into individual lines. The segmentation of handwritten documents into lines should address many problems like different handwriting styles, overlapping lines and characters, skewed characters and lines, different font sizes etc. These issues are considered during the segmentation process. These segmented lines are segmented into words and afterwards segmented to characters. Then, the unique features for identifying each character is extracted and given to a classifier. The classifier identifies each character and the corresponding Malayalam character is obtained.