A brief description of statistical and syntactic pattern matching techniques is presented with an emphasis on statistical techniques. The characteristics of the Telugu script are described. A subset of 16 characters, which are both easy and hard to recognize, is selected for the dictionary of standard characters. A weighted linear difference polynomial of features is used to recognize Telugu characters. The features were Fourier descriptors of projection profiles and cross sections taken in various directions. Algorithms for obtaining the projection profiles cross sections and adaptive learning method are presented. The system was trained and tested with a set of 8 nano-written samples of each of 16 different Telugu characters. More than 90% of the 123 samples were correctly recognized by the system. Results of numerous trials examining the different features and classification techniques are discussed.

Library of Congress Subject Headings

Optical pattern recognition; Pattern recognition systems; Telugu language--Data processing

Publication Date


Document Type


Student Type


Degree Name

Computer Science (MS)

Department, Program, or Center

Computer Science (GCCIS)


John A. Biles

Advisor/Committee Member

Larry Coon

Advisor/Committee Member

Peter Anderson


Physical copy available from RIT's Wallace Library at TA1650.M36 1987


RIT – Main Campus