Decision Trees for Hand-Written Arabic Words Recognition

摘要

In this paper, we present a system based on decision trees for the off-line recognition of handwritten Arabic words. The aim of this work is to design and implement a system for the recognition of Algerian city names (wilayas), based on the symbolic learning decision tree approach. After the acquisition step, images are preprocessed and structural features (sub-words, loops, ascenders, descenders and diacritical dots) are extracted. These features, combined with the corresponding classes are presented as input to a learning process which gives as a result a decision tree that can be used for the classification step in our recognition system. The resulting tree can be expressed more explicitly as a rule base for words classification. These rules are not based on theoretical information, but on training samples. Our experimental recognition results are encouraging and confirm our expectation that the use of structural features and symbolic learning is an interesting issue of wholistic handwritten words recognition.