Segmentation-Free Online Arabic Handwriting Recognition

Abstract

Arabic script is naturally cursive and unconstrained and, as a result, an automatic recognition of its handwriting is a challenging problem. The analysis of Arabic script is further complicated in comparison to Latin script due to obligatory dots/stokes that are placed above or below most letters. In this paper, we introduce a new approach that performs online Arabic word recognition on a continuous word-part level, while performing training on the letter level. In addition, we appropriately handle delayed strokes by first detecting them and then integrating them into the word-part body. Our current implementation is based on Hidden Markov Models (HMM) and correctly handles most of the Arabic script recognition difficulties. We have tested our implementation using various dictionaries and multiple writers and have achieved encouraging results for both writer-dependent and writer-independent recognition.

Online versus offline Arabic script classification

“…Some authors have presented a set of features in terms of a couple of clustering handwriting patterns and also combining them [7]. Biadsy et al [8] presented a structural feature set, in which five features were obtained from segmenting of letters, i.e. number of segments: pendown and pen-up the entrance devise as strokes of letter, cross-points: detecting whether the letter has loop or not, shape edges: extracting the critical points of the letter as sharp edge in which, secondary segments: these features are from 2 up to 4 as completed parts for the letter, i.e.…”

HMM and fuzzy logic: A hybrid approach for online Urdu script-based languages’ character recognition

“…Benouareth et al-modeled discrete structure, right to left HMMs with explicit state duration and gamma, gauss and Poisson distribution, is used for modeling the state duration [21]. Rigoll presented HMM-based simultaneous segmentation system for online handwritten mathematical expressions [10] whereas Biadsy et al presented segmentation-free handwritten approach for Arabic [29]. Mitsuru and Akira presented on-line handwriting recognition for Kanji characters by employing sub-stroke HMMs as minimum units and pen motion direction is utilized as features [12].…”

Efficient Generation of Comprehensive Database for Online Arabic Script Recognition

“…As in other scripts, the poor recognition rates results from unsuccessful segmentation shifted the focus to the segmentation free approach. In the holistic approach [10,12,3,7], complete words are processed to be recognized bypassing the character segmentation stage. Obviously, as opposed to the holistic approach, the other two options do not require large databases for training.…”

“…In contrast, very few databases were developed for the Arabic script and fewer became publicly available. Among these, we can mention IFN/ENIT and CEDAR databases, but mostly researchers had developed their own small datasets or large databases that are not available to the public [11,1,5,7]. None of these databases included all the different handwritten words or word-parts in the Arabic language.…”