Repossession and Recognition System: Transliteration of Antique Tamil Brahmi Typescript

Tamil is among the ancient languages in the world with a rich literature. Recognition of antique Tamil scripts is difficult and different from the present form of the language. The character recognition of Brahmi script poses a big challenge even today. In this paper, a new technique for extracting the features is proposed, and converting the ancient Tamil script into the present form. Initially, the system is implemented by performing the pre-processing steps. Then the characters are individually separated using the segmentation process. The processed image undergoes a new feature extraction technique, where the system applies a chi-square test to check whether all the zoning feature values of the image are independent or dependent. The characters are recognized from the extracted features using neural networks. NNTool is employed to train the featured image and the data are compared with the database to recognize the Brahmi characters. The feature extraction technique along with the neural network achieved recognition rate accuracy of 91.3% and error rate of 8.7% using the confusion matrix. Our experiment has been simulated using MATLAB.