Performance Analysis of BAM and BPN for Character Recognition

The Overall research work contains a number of procedural steps. Such as converting a black and white image to grayscale and various image processing steps, learning a pattern using Back-propagation Neural Network (BPN) and Bidirectional Associative Memory (BAM) recognize a existing or non-existing pattern from the learned pattern using BAM and BPN technique. All the steps of image processing and feature detection steps are performed and coded by the programming language Matlab provided by the Matwork Inc. Primary operation of Matlab is to acquire an image from the storage disk which is stored in a specific format and process the image. One of the most important factors of the research work is Feature Extraction which is implemented by Matlab. Then the processed pattern is stored in a file such that the file can be processed by another procedure. In pattern learning BPN and BAM is used. The parameters of the learning algorithm are varied and observed the nature of the system. From pattern learning, we get a weight set and correlation matrix which is used in recognition. The recognition part, the results from learning algorithm are used to math the unknown input.