Offline Handwritten Arabic Word Recognition Using HMM - a Character Based Approach without Explicit Segmentation
Volker Märgner, Haikal El Abed, Mario Pechwitz · HAL (Le Centre pour la Communication Scientifique Directe) · 2006
This paper presents the IfN's Offline Handwritten Arabic Word Recognition System. The system uses Hidden Markov Models (HMM) for word recognition, and is based on character recognition without explicit segmentation. The first part of this paper deals with databases for word recognition systems, and in particular, the IFN/ENIT -database. The second part gives a short description of the pre-processing, normalisation, and feature extraction methods needed for this system. The final part gives a practical approach to the HMM-Recogniser used in our system and some results are presented.