Recognition of off-line printed Arabic text using Hidden Markov Models

Abstract

This paper describes a technique for automatic recognition of off-line printed Arabic text using Hidden Markov Models. In this work different sizes of overlapping and non-overlapping hierarchical windows are used to generate 16 features from each vertical sliding strip. Eight different Arabic fonts were used for testing (viz. Arial, Tahoma, Akhbar, Thuluth, Naskh, Simplified Arabic, Andalus, and Traditional Arabic). It was experimentally proven that different fonts have their highest recognition rates at different numbers of states (5 or 7) and codebook sizes (128 or 256).Arabic text is cursive, and each character may have up to 4 different shapes based on its location in a word. This research work considered each shape as a different class resulting in a total of 126 classes (compared to 28 Arabic letters). The achieved average recognition rates were between 98.08% and 99.89% for the eight experimental fonts.The main contributions of this work are the novel hierarchical sliding window technique using only 16 features for each sliding window, considering each shape of Arabic characters as a separate class, bypassing the need for segmenting Arabic text, and its applicability to other languages.

Open-vocabulary recognition of machine-printed Arabic text using hidden Markov models

“…Al-Muhtaseb et al proposed a hierarchical sliding window for printed Arabic text recognition [22]. A window is divided into eight non-overlapping vertical segments.…”

“…These hierarchical windows generate features with greater weight in the center region of the writing line (baseline). These hierarchical windows produced very high recognition rates with synthesized data [22]. Experiments with text line images extracted from scanned documents yielded poor results [23].…”

KHATT: An open Arabic offline handwritten text database

“…It can be used to extract all Arabic characters in their all forms. It was shown in [5] that when the number of samples for some Arabic shapes are not enough in the training data, the recognition rate can be improved by including 50 samples of a minimal text paragraph that covers all shapes of Arabic characters to train the HMM. The second paragraphs in this page were selected randomly from a large corpus that we developed from 46 sources to make the database a statistical representation of the corpus.…”

Arabic Character Recognition

“…The author of PATS dataset [3], the first dataset we used to test our algorithm, uses HMM and a sliding window to segment and recognize Arabic scripts. See also the recent [2].…”