A Path Signature Approach to Online Arabic Handwriting Recognition

Abstract

The Arabic script is one that has many properties that come together and result in what is commonly cited as one of the most beautiful scripts. Used by over 400 million people worldwide and with a history spanning over 1800 years, the Arabic script remains one of the most important languages in the world. Using tools from the theory of rough paths, combined with state of the art techniques from deep learning, we develop a recognition methodology for Arabic handwriting. Preliminary results using online Arabic handwritten characters show that the methodology developed can result in a significant decrease in error rate.