Combining Slanted-Frame Classifiers for Improved HMM-Based Arabic Handwriting Recognition

Abstract

The problem addressed in this study is the offline recognition of handwritten Arabic city names. The names are assumed to belong to a fixed lexicon of about 1,000 entries. A state-of-the-art classical right-left hidden Markov model (HMM)-based recognizer (reference system) using the sliding window approach is developed. The feature set includes both baseline-independent and baseline-dependent features. The analysis of the errors made by the recognizer shows that the inclination, overlap, and shifted positions of diacritical marks are major sources of errors. In this paper, we propose coping with these problems. Our approach relies on the combination of three homogeneous HMM-based classifiers. All classifiers have the same topology as the reference system and differ only in the orientation of the sliding window. We compare three combination schemes of these classifiers at the decision level. Our reported results on the benchmark IFN/ENIT database of Arabic Tunisian city names give a recognition rate higher than 90 percent accuracy and demonstrate the superiority of the neural network-based combination. Our results also show that the combination of classifiers performs better than a single classifier dealing with slant-corrected images and that the approach is robust for a wide range of orientation angles.

Generative vs. Discriminative Recognition Models for Off-Line Arabic Handwriting

“…In addition, IFN/ENIT information was utilized for testing and training, and maximum recognition rates of 89% were observed. For the detection of offline Arabic handwriting, Al-Hajj et al [ 10 ] suggested an arrangement of three right-to-left HMM classifiers. All classifiers were built upon specific sliding window orientations in order to surmount the significant difficulties of offline handwritten words, including overlaps, inclinations and the shifted positions of diacritics.…”

Comparative Study of HMM and BLSTM Segmentation-Free Approaches for the Recognition of Handwritten Text-Lines

“…Total number of parameters depends on w window size. Each frame consists of a set of features which have been successfully applied to HMM-based latin and arabic handwritten word recognition [9], [10] and recently to text lines recognition [19]. These features are the following: In order to extract features, each window is divided into a fixed number of cells.…”

“…Thus we only test a few state number values. Following previous work [10], we restricted the step parameter to the range [2,4] and the window width to [8,9].…”

“…A sliding window is shift from left to right in order to extract feature sequences, thus avoiding to segment the text-line into words and characters. We use an efficient set of features we have proposed for latin and Arabic scripts [9], [10]. However window parameters must be adequately set according to handwriting image resolution.…”

Dynamic hierarchical Bayesian network for Arabic handwritten word recognition

“…El-Hajj in [11] presented a system for the off-line recognition of handwritten Arabic city names. The system was based on the HMM applied with a set of features including both the independent baseline and the dependent one.…”