Skip to Main Content
In this paper we propose an automatic system that recognizes continuous Arabic-Urdu Alphabet scripts through mouse in real- time based on Artificial Neural Network (ANN). The proposed neural network is trained using traditional back-propagation algorithm for self supervised neural network which provides the system with great learning ability and thus has proven highly successful in training for feed-forward neural network. The performance analysis was based upon a set of data consisting of specimens collected from 5 persons; each specimen consisted of 30 basic Arabic-Urdu Alphabets. The system incorporates Neural Networks as its learning and recognition engine. The designed algorithm is not only capable of translating discrete gesture moves, but also continuous gestures through mouse. In this study, we proposed an efficient neural network approach for recognizing Arabic-Urdu scripts drawn by mouse. The proposed approach shows an efficient way for extracting the boundary of the script and specifies the area of the recognition alphabets where it has been drawn in an image and then used ANN to recognize the alphabets. A comprehensive Arabic-Urdu Script Recognition (AUSR) system is designed and tested successfully. The results based on speed and accuracy were analyzed.