Abstract—In recent years, most of the research in handwriting recognition seems to have focused on the problem of on-line cursive script recognition. The cursive nature of Persian script and the existence of different handwriting styles for its alphabet as well as the fact that each character is written in different forms depending on its location in a word, which make segmentation and recognition of Persian words a challenging task. In this paper, we propose a novel segmentation method for online Persian handwriting using some generic features of Persian letters in cursive words. In addition, some easy to implement techniques for extracting those features are presented as well. Our segmentation process is composed of two modules. The first one copes with the preprocessing of input data for which we propose a normalization technique to make distances of consecutive points of the input uniform. By doing so, the input data becomes independent of writing speed and input device. The second module deals with segmentation of a word into its constructing letters. Our results from implementation of the proposed method show a total accuracy of up to 98.625%.
Index Terms—Feature extraction, online cursive script, Persian words, segmentation.
S. Pirnia Naeini and M. Khademi are with a faculty member of the Department of Computer Science, South Tehran Branch, Islamic Azad University, Tehran 11365/4435, Iran (e-mail: email@example.com).
A. Nikookar and Z. Bani are with member of the Young Researchers Club, South Tehran Branch, Islamic Azad University, Tehran 11365/4435, Iran (e-mail: firstname.lastname@example.org; email@example.com).
Cite: Shahriar Pirnia Naeini, Maryam Khademi, Alireza Nikookar, and Zahra Bani, "A Feature-based Approach to Segmentation of Online Persian Cursive Script," International Journal of Modeling and Optimization vol. 2, no. 4, pp. 391-395, 2012.