Performance Measure of Scanned Tamil Land Documents using Neural Network Approach

Abstract

Recognition is process to find out the noisy or distorted image to make an accurate image. Classification and recognition technique which implemented on scanned Tamil land document. In the pre-processing stage, the given dataset is filtered by using median filter. After that, segmentation process is applied for every word image is splitted into character. Then, feature extraction is done by Gabor wavelet. For post processing stage, classification is process to check out the dataset which using neural network technique like supervised learning method or unsupervised learning method to find out the correct and Incorrect classification measure using confusion matrix. Hence, finally implemented Gabor wavelet technique to find the feature extraction and selection and then classification is done new pattern recognition technique using MATLAB. We also find out performance of plotting function like Training state, Regression, Gradient and validation.
DOI: 10.17762/ijritcc2321-8169.1505192