A New Approach of Pattern Matching and Analysis for Handwritten Digits-Using Gradient Descent Back Propagation with Adaptive Learning

Abstract

A neural network is a machine that is designed to model the way in which the brain performs a particular task or function of interest: The network is usually implemented by using electronic components or is simulated in software on a digital computer. “A neural network is a massively parallel distributed processor made up of simple processing units which has a natural propensity for storing experiential knowledge and making it available for use. It resembles the brain in two respects:

1. Knowledge is required by the network from its environment through a learning process.

2. Interneuron connection strengths, known as synaptic weights, are used to store the acquired knowledge”.

   An important contribution of this research lies in the provision of a much needed different person’s handwritten database. This offers practical benefits for pattern matching and analysis of handwritten digits and providing a facilitate training and testing. In this research work i) 100 persons of handwritten digits from 0 to 9 have been collected. Then on that database MLP Classifier is applied. It is found that the result of MLP classifier is giving only 45.50% of accuracy. Therefore new database has been formed by following Way of writing characters.ii) Once again new collected data passes over pre-processing of MLP. In pre-processing different operations are perform such as image acquisition, gray conversion, binary conversion, edge detection, image dilate and image fill. After completion of pre-processing image passes over for normalization and then to feature extraction. Now lastly passes over classifier Gradient descent back propagation with adaptive learning rate. Finally handwritten digit is recognized by 80.10%.