A Bilingual Machine-Interface OCR for Printed Kannada and English Text Employing Wavelet Features

An Optical Character Recognition (OCR) system is one of the important research areas in the field of Human- machine interface. This paper presents a bilingual OCR system for printed Kannada and English text. Gabor filter based features are used for separating the Kannada and English words from the bilingual document. Wavelets that have been progressively used in pattern recognition are used in the system to extract the features for classifying both the Kannada and English characters. Multilayer feed forward Neural classifiers known for their good generalization and approximation property have been effectively used in the system for the classification. An overall recognition rate of 90.5% is obtained at character level.