Recognition of Devanagari Numerals Using Jordan-Elman Network for CAPTCHA Design

Abstract

Devanagari is an Indo-Aryan script adopted by most of the Indian languages. The languages like Hindi, Marathi, Konkani, Haryanvi, Rajasthani, and Bhojpuri are derived from Devanagari. In India, nearly 77% of total population speaks and write in Devanagari based languages and use it for various commercial transactions. Most of the Government schemes are launched on websites. In most part of India, people use language based on Devanagari script. They use Devanagari numerals for various purposes such as for writing amounts in payment slips, for counting and measuring etc. To take benefits of these schemes, user first has to pass the CATCHA test deployed on log-in page of website. CAPTCHA which stands for “Completely Automated Public Turing Test to tell Computers and Humans Apart” is a defensive technique deployed on websites to safeguard the web contents from illegal automated scripts called web bots.

The paper proposed a Jordan-Elman Neural network system that can recognize Devanagari numerals. The proposed system recognizes the noisy numerals with better accuracy as compared to other systems. The Recognized numerals are intended to be use in a CAPTCHA code along with other Devanagari characters.