Handwritten character recognition using diagonal-based feature extraction

This paper proposes a diagonal feature-based handwritten character recognition, using multi-layered feed forward neural network with back-propagation algorithm. The pixel information along the two diagonals is primarily extracted and used for recognition. The system is trained and tested on handwritten digits, English lower case character set and Tamil character database and their classification accuracy is compared. Finally, the recognition complexity of the fonts is also analyzed by testing the system with all the image pixels and the pixels along the central horizontal and vertical span.