Application of 13-point feature of skeleton to neural networks-based character recognition

This paper describes the application of 13-point feature of skeleton for an image-to-character recognition. The image can be a scanned handwritten character or drawn character from any graphic designing tool like Windows Paint Brush. The image is processed through conventional and 13-point feature of skeleton methods to extract the raw data. The extracted data will be used to train two different models of Neural Networks namely Linear Associative Memory model (LAM) and Back-Propagation model (BP). To test the two networks, new sets of data have been used. Based on the results, this paper analyzes the performance of the networks and discusses the problem.