Analysis of Telugu Palm Leaf Characters Using Multi- Level Recognition Approach

Palm leaf character recognition is an area which is at the nascent stage of research. Although character recognition is a well-known application of pattern recognition, lot of work is still to be exploited in handwritten character recognition. The recognition accuracy as per the literature survey for handwritten English characters is very low and for Indian languages it is just started. Research has been started for Indian languages like Bangla, Hindi, Telugu, Tamil, Devangari, etc., but still at the starting stage. Palm leaf character recognition is an open area of research and is also very important since these palm leaves contain huge amount of information related to astronomy, astrology, architecture, law, medicine and music. In the present work, an additional feature called depth of indentation at important pixel points like the starting point, curves, joints, loops and end points is considered which is directly proportional to the pressure applied by the scriber on the palm leaf. This depth of indentation is considered in the Z-direction measuring in microns. In the proposed work, multistage recognition approach is used to improve the recognition accuracy up to 92.8%.