Abstract

The process of recognizing scanned documents or machine printed documents using automated tools are used in different real life domains. Designing a method with cent percent accuracy of character recognition is a challenging and unachievable task. Presence of noise, distinct styles of font under real time environment makes character recognition more difficult. In this paper we describe recognition of handwritten basic characters of OlChiki script, used by more than 10 million tribal people in India mostly from Assam, Bengal, Bihar, Odisha and Jharkhand. There are 30 basic characters and 10 numeral digits in OlChiki and we have used a dataset of 10000 handwritten isolated character samples written by 50 persons. Samples in this dataset are composed of one stroke. Curvelet and Geometry based feature extraction has been used for comparison of performance. Strokes are recognized dynamically by using KNN and SVM classifier together. We have received an encouraging recognition result of 87% accuracy.