Digit Recognition using Neural Network

Handwritten digit recognition is the process of recognizing and classifying handwritten digits without human interaction. Its application field is very wide, for example Postal code recognition (Automatic sorting of mail by destination ZIP code), Digitizing hand written spreadsheets, tax forms etc. Patterns slightly shifted, distorted and even overwritten can be correctly recognized. Neural network aids in efficient recognition. A Multilayer Neural network trained with “Backpropagation “algorithm is used. Kirsch masks are adopted for extracting feature vectors and a multi layer clustered neural network is used for classifying numerals efficiently. The neural network will be trained with a database consisting of handwritten digits provided by writers of various ages with many different sizes and writing styles. Numerals poorly drawn or cannot be classified are rejected. A very high recognition rate, even above 90% could be obtained while using neural network.