Online Handwriting Recognition of Arabic Characters Using Fuzzy Rules

Abstract

Summary and Conclusion The objective of this thesis is to discuss the problem of Arabic online handwriting recognition. At first, an online handwriting recognition system for Arabic characters is developed. Later on, a preliminary study is carried on the recognition of handwritten Arabic words (which is cursive by its nature) This study results on developing a word segmentation module to be used in word recognition system. The study also discussed the needed modifications on the other modules of the word recognition system to be compatible with the output from the developed word segmentation module. Most of the described techniques are implemented in the developed programs within this thesis. 6.1 Summary and conclusions The developed system are attempted to overcome the ambiguities of Arabic handwriting with the aid of fuzzy logic. In the character recognition system, at the low level recognition stage, the fuzzy abruptness measure is introduced to make the segmentation output independent of the writer or writing style. The segmentation process is known to be the most difficult part of the whole system because the segmentation has to be robust to cope up with the handwriting changes due to mode, health and different writing styles. In the segmentation module, the spark point correction algorithm is introduced to replace the unwanted points that represents virtual abrupt change in direction (or virtual sharp corners). These spark-points increase the input data distortion and results in more unwanted segments which cause the recognition to fail or at least to take more time which is not acceptable in the online handwriting recognition. In the next level, the feature extraction algorithm is applied to the resulting segments to extract