Recognizing handwritten words has been and still is one of the most challenging problems in Artificial Intelligence (AI). Words are rather complex patterns, having many variations in handwriting style. Despite the considerable progress achieved in recent years, performance of handwriting recognition systems is still far from human's both in terms of accuracy and speed.
    I devoted my M.Sc. thesis to this subject, and providing some of the results here. Since there is a lack of research on the problem of Farsi (Persian) handwritten recognition, the focus of the proposed and surveyed methods was on the Farsi scrip; however, most of them can be applied to Roman scripts with no modifications. To the best of our knowledge, this work is the first to use continuous hidden Markov models with structural features to recognize Farsi handwritten words, and almost all parts of a complete recognition system are addressed. The other contributions of this work are:

• Developing a new machine learning approach based on the naive Bayes classifier for text segmentation.

• Comparing and contrasting four different algorithms for document image binarization.

• Surveying different skew and slant correction algorithms for handwritten words and documents.

• Comparing and contrasting five different skeletonization algorithms with the main focus on preserving text characteristics.

• Extracting a set of structural features from Farsi handwritten words, independent of the baseline location.