Feature Extraction Techniques Implementation Review and Case Study
Uma Bhati, Krishna Nand Chaturvedi · 2015
In world, around 400 million people using Devanagari language for record. There are a noteworthy change in the exploration identified with the recognition of printed and handwritten Devanagari content in the previous couple of years. The confounded written work style of Devanagari characters with loops and curves create the testing and validation procedure quite difficult. This paper, shows a similar investigation of Devanagari character recognition utilizing 4 feature techniques—Zoning, Projection histograms, Chain code histograms and HOG Gradients. Preprocessing, feature extraction, classification methods valuable for the recognition are talked about in different areas of the paper. In this paper, recognition of Hindi characters is finished by utilizing a three stage system. Initial step is preprocessing, in which binarization of the picture and detachments of characters are performed. The following step is extraction in which either of the existing techniques of feature extraction is used. Third step is testing process. The current paper focuses on the different techniques for the second stage and compares the results of different authors work in the past.