Devanagari is the basic script for many languages of India, including their National language Hindi. Unlike the Latin script used for the English language, it does not have upper case or lowercase. It has only one case of writing. Moreover each alphabet contains more curves than straight lines. Hence handwritten Devanagari character recognition is a challenging task. To capture different handwritten styles of each alphabet, different approaches have been proposed. In this work, we investigate a simple filtering technique on the features. Support Vector Machine (SVM) was used as classifier. It has been applied on two benchmark Devanagari databases and results show an improvement of as much as 5-10%. This improvement is found to be consistent with different sizes of the database. It was studied on pixel density features and GIST features separately. GIST features were found to be more effective and like having the potency of self-containing filtering.
Dr. Oruganti Venkata Ramana Murthy and Hanmandlu, M., “A Study on the Effect of Outliers in Devanagari Character Recognition”, International Journal of Computer Applications, vol. 32, 2011.