A novel comprehensive database for Arabic and English off-line handwritten digits recognition

Abstract

The recognition of Arabic handwritten is received at the same interest as other Latin languages. In Optical Character Recognition (OCR), handwriting Arabic recognition is considered as one of the critical and difficult tasks in the various scientific area. The main issues of this matter were due to the lack of public Arabic handwriting databases and the cursive nature of Arabic writing. In this paper, a new benchmark database is built for the Arabic and English off-line handwritten digits Recognition. The original form is divided into three groups: Arabic digits, English digits, and word Arabic digits which written five times by 100 different academic staff and students of university writers. Our database contains 14500 images; divided into two subsets of training and testing to help researchers through evaluating and comparing results obtained from their systems.