A Novel Approach for Online Sindhi Handwritten Word Recognition using Neural Network

Abstract

Online and Offline Handwritten Recognition has become an important part of the field of Pattern Recognition since it is considered as a technological revolution between man-machine interfaces. The handwriting has sustained to be a continuous mode of communication for recording information in day-to-day life. The exciting nature of Handwriting Recognition and Segmentation has attracted the concentration of researchers from academic as well as industry domains. Research for Online and Offline Handwritten Sindhi Word Recognition is at very beginning stage as compared to Latin, Chinese and Arabic languages. Sindhi language is among the ancient languages of the world. This language contains fifty two alphabet characters with different shape, cursive style and position of characters which increases the complexities and difficulties for recognition as compared to other Unicode based languages. This research paper has addressed a novel approach for recognizing Sindhi words using Artificial Neural Network (ANN). Self-Organizing Map (SOM), an ANN algorithm has been used for Sindhi Word Recognition entered on the surface of touch screen device such as Tablet PC or Smart Phone on real time. Unsupervised learning method has been used to train the proposed system that randomly alters the weight of the matrix closer to the input. A dataset consisting of 1200 words has been collected from 60 native writes of Sindhi language. An accuracy rate of 83% has been achieved with recognition time of 20-30 milliseconds.