Fisher Linear Discriminant Analysis based Technique Useful for Efficient Character Recognition

This paper describes the character recognition process from printed documents containing Kannada and English text. Kannada is the fifth most popular language in India and English is the most popular language in the world. Kannada is the language spoken by more than 60 million people of South India and English is the second official language at various government organizations through out India. The proposed character recognizer is based on the Fisher linear discriminant (FLD) analysis. It is usually performed to investigate differences among multivariate classes, to determine which attributes discriminate the classes, and to determine the most parsimonious way to distinguish among classes. The proposed system is tested on various fonts, degraded characters, noisy characters of Kannada and English. The overall accuracy of the proposed system is 96.1%.