A high reliability classifier using decision trees and AdaBoost for recognizing handwritten Bangla numerals

It is rather hard to achieve high recognition reliability using a single set of features and a single classifier for off-line handwritten numeral recognition systems. In this paper, we present a two-stage classifier for recognizing handwritten Bangla numerals. The first stage classifier is a decision tree based on ID3 algorithm, and the second one is a series of decision trees combined by Weight-Restricting-Based AdaBoost algorithm (WRB AdaBoost). Two sets of features are employed in the different stages. The first set is Open and Closed Cavity (OCC) features, and the other is a combination of OCC features and Ending and Crossing Point (ECP) features. Experiments on numeral images obtained from real Bangladesh envelopes show that the proposed recognition method is capable of achieving high recognition reliability.