A Simple Feature Extraction Method for Analysis of Hand Written Characters

Abstract

The printed and handwritten alphabets as character recognition by feature extraction are described during this proposed work. A predefined set of 108 samples each containing twenty-six alphabets taken by paper handwritten are taken as set for training. The system portrayed is a good alternative solution for HCR & plays well with detection of handwritten characters. Here, the method experiences binarization and pre-processing for the segmentation stage. Individual characters after segmentation undergo feature extraction part. The artificial neural network is trained using the extracted features. A simple ANN is used as a classifier utilizing the average epochs and MSE for detecting the HCR recognition without a false alarm.