Recognition of roman characters using geometric and regional features

Abstract

This paper presents the Roman Characters Recognition in clean and noisy environment using geometrical and regional features. The geometrical and regional features are extracted from the standard dataset and are combined to achieve better recognition. The combination of these features is classified using neural network (NN) and random forest (RF) classifiers. In our experiments, we have achieved the recognition accuracy of 100% for some characters. However, the average recognition accuracy of 85.7% has been recorded by using NN and 88% has been recorded by using RF classifier, respectively.