Character recognition is one of the most fascinating and challenging researches currently in the area image processing. It has been receiving considerable attention due to its versatile range of real-time application which includes reading aid for the blind, postal automation, processing of cheque and digitization of historical documents. Now a days different methodologies for different language are in widespread use for character recognition. Character recognition from a scanned document page involves difficult task due to the free-flow nature of handwritten. In this study a geometric feature extraction approach is implemented with efficient learning mechanism for training and testing using neural network for Tamil handwritten script. After selective preprocessing steps for constrained inputs, the document is split into paragraph and then segmented to line, word and individual character for further recognition. The geometric features for each character are trained in an Effective Learning Machine (ELM) with almost information. With this information each testing character is analyzed for recognition. This procedure results more than 90% accuracy for individual characters.
展开▼
