A highly effective system for printed Thai character recognition by word prediction method

Abstract

This paper proposes a novel approach: word prediction and verification technique for enhancing the recognition performance for printed Thai characters. Instead of attempting to recognize individual characters, the system, named as OCR-WP, is a technique employed to identify whole words by using contextual analysis in order to predict those probable words and verify the right one by template matching. A series of experimental results shows that the OCR-WP could yield a better performance in terms of speed improvement, while still maintaining the quality of error-free accuracy. © ICITA 2009.