Automatic Generation and Recognition of Telugu Script Characters

Abstract

This paper deals with the computer generation of characters of various fonts and sizes for the Telugu alphabet. It also deals with the problem of automatic recognition of Telugu printed scripts. Telugu and many other Indian languages scripts have the problem of combined character formation that makes a few hundred shapes to be generated or recognized, although the number of vowel and consonant characters in isolation does not exceed hundred. In this paper, 23 basic and 24 primitive shapes are carefully chosen which when properly concatenated can generate the whole character vocabulary. To generate these shapes, curve drawing based on Bezier and Spline techniques are used. A new font is proposed where the characters are concatenation of straight lines. A grammar to generate the combined characters is also proposed. Next, the recognition of characters is done by shape matching using directed curve tracing method. A concept of T-tuple is introduced for the tracing problem that act on the skeleton of the character to recognize initially the basics and primitives and then combine the information to recognize the characters.