A novel segmentation technique for splitting a typed Persian text to sub-words

One common approach in recognition of a Persian text is to segment the text into its component words and then words to single characters. Due to specific characteristics of Persian words, this approach is non-trivial and literature reports relatively low success rate for Persian character segmentation. As an alternative, we segment a Persian text only to its component sub-words; and then, recognize sub-words from a large library of sub-words. In this document, we describe a novel segmentation technique to split a Persian text to its sub-words components with a perfect success rate for the texts and fonts tested.