Robotic handwritten Kannada character recognition using neural network

Abstract

Data preparing and the board is basic now a days. In this paper, programmed preparing of structures written in Kannada language is considered. A reasonable pre-preparing procedure is introduced for separating written by hand characters. Essential Component Analysis (PCA) and Histogram of arranged Gradients (HoG) are utilized for highlight extraction. These highlights are sustained to multilayer feed forward back spread neural system for arrangement. Just 57 characters are utilized for acknowledgment. Exhibitions of two highlights are looked at for changed number of classes. Hoard is found to have preferred acknowledgment exactness over PCA as number of classes expanded. This is actualized in Visual Studio 2010 utilizing Open CV library.