A programming based handwritten text identification

A handwritten text categorization technique using supervised and unsupervised learning is proposed in this work. The learning system based on neural network is used for alpha numeral identification. The identified characters are subsequently merged to convert the handwritten text into the closest printed form. A word matching algorithm along with a programmed glossary finds out the most appropriate words and forms the printed text thereafter. The printed text is identified by matching the keywords of the text with the programmed glossary of different subjects. Once the text is identified, the inappropriate words in the textual context are corrected to match the respective subject of the text. This further improves the meaningfulness of the identified handwritten text. Holdout method is attempted for performance evaluation.