Feature Extraction and Classification Techniques for Handwritten Nandinagari Vowels

Authors : Prathima Guruprasad , Dr. Jharna Majumdar , Guruprasad K S Rao


 In this paper the features of Nandinagari Vowels are extracted and classified to recognize the characters. Data sample is collected from different writers. They were scanned using a flat bed scanner at a resolution of 300 dpi and stored as gray-scale images. These images were preprocessed programmatically using thresholding technique to enhance the quality of image and converted to binary images. A total of Nineteen global scale invariant features are extracted for every Nandinagari Vowel and stored in the database. These features are then used to train and test the input samples using Support Vector Machine. This is the first attempt to recognize Nandinagari Vowels and there is no recognition system available till date. The system has achieved a good recognition accuracy of 94.45% on the handwritten Nandinagari Vowels

Keywords : Nandinagari Vowels, Character Recognition, Pre-Processing, Feature Extraction, Dimension Reduction, Principal Component Analysis, Classification, Support Vector Machines