Machine Learning approaches for Nastaliq style Urdu handwritten recognition: A survey

摘要

In day to day life handwriting is a mean of communication and recording the information regardless of the new technologies. The handwritten information is converted into digital form that is used as an input. Recognition is an attractive area of interest in research in recent years. There are good recognition system available for alphabetical and symbolic languages but for the language based on Arabic script like Urdu no such recognition system exist. The recognition is done either online or offline. For an online recognition system the input is taken through stylus or tablet whereas offline recognition is done by taking images through scanner. With the development in electronic tablets the movements of pen are captured accurately. Recognition of Urdu handwritten text is very complex task due to challenges in Urdu script. In this paper a survey is carried out on different Deep learning and statistical techniques for handwritten text recognition.