Writer Identification on Multi-Script Handwritten Using Optimum Features

Recognizing the writer of a text that has been handwritten is a very intriguing research problem in the field of document analysis and recognition. This study tables an automatic way of recognizing the writer from handwritten samples. Even though much has been done in previous researches that have presented other various methods, it is still clear that the field has a room for improvement. This particular method uses Optimum Features based writer characterization. Here, each of the samples written is grouped according to their set of features that are acquired from a computed codebook. This proposed codebook is different from the others which segment the samples into graphemes by fragmenting a certain part of the writing known as ending strokes. The proposed technique is employed to a locate and extract the handwriting fragments from ending zone and then grouped the similar fragments to generate a new cluster known as ending cluster. The cluster that comes in handy in the process of coming up with the ending codebook through picking out the center of the same fragment group. The process is finalized by evaluating the proposed method on four datasets of the various languages. This method being proposed had an impressive 97.12% identification rate which is rates the best result on the ICFHR dataset.