Recognition of ancient Kannada Epigraphs using fuzzy-based approach

Optical Character Recognition finds application in the field of Epigraphy, which is the study of inscriptions. Epigraphers who read ancient inscriptions are nowadays less in number and almost becoming extinct. There is a lot of scope for digitization of these historical records and automated decipherment of the same. An attempt is made for the recognition of text in ancient Kannada script of two different periods Ashoka and Hoysala. A reconstructed epigraph image is taken as input, then characters are segmented using Nearest Neighbor clustering algorithm. Next Statistical features such as Mean, Variance, Standard Deviation, Kurtosis, Skewness, Homogeneity, Contrast, Correlation, Energy, and Coarseness are extracted, and stored as training data set and for comparison at the later stage of testing. Mamdani Fuzzy Classifier is used in classification of characters. Finally the classified characters of ancient times are displayed in modern Kannada form. Proposed system successfully recognizes ancient text and maps into equivalent modern character and observed that the recognition rate for Brahmi script is appreciable when compared to Hoysala script.