Handwritten Devanagari Similar Character Recognition by Fisher Linear Discriminant and Pairwise Classification

The research works in Handwritten Devanagari Characters are continually evolving into new challenges, which exposed the new sources of further research work like, character normalization, gray-level normalization, a discrimination of the similar characters and many more. This paper discusses the discrimination of the similar characters, which is one of the major sources of classification error. The similar shape character has a very minute difference, which is called critical region and used to discriminate them by human beings. The primary goal of the current work is to identify the critical region of the similar character and use the same to generate additional features in order to minimize the classification errors in the end results. It is also quite challenging to identify the critical region as the characters are written in different handwriting styles and fonts. The paper suggestes the Fisher linear discriminant model to detect the critical region, which is used to extract the additional feature. The experiments work was conducted on the standard database, which has 36[Formula: see text]172 handwritten Devanagari characters and significant improvement has been recorded by the aforesaid technique.