A Cascaded Scheme for Recognition of Handprinted Numerals.
Ujjwal Bhattacharya, Tanmoy Kanti Das, B.B. Chaudhuri · 2002
This paper proposes a novel off-line handprinted Bangla (a major Indian script) numeral recognition scheme using a multistage classifier system comprising multilayer perceptron (MLP) neural networks. In this scheme we consider multiresolution features based on wavelet transforms. We start from certain coarse resolution level of wavelet representation and if rejection occurs at this level of the classifier, the input pattern is passed to a larger MLP network corresponding to the next higher resolution level. For simplicity and efficiency we considered only three coarse-to-fine resolution levels in the present work. The system was trained and tested on a database of 9000 samples of handprinted Bangla (a major Indian script) numerals. For improved generalization and to avoid overtraining, the whole available data set had been divided into three subsets -- training set, validation set and test set. We achieved 94.96% and 93.025% correct recognition rates on training and test sets respectively. The proposed recognition scheme is robust with respect to various writing styles and sizes as well as presence of considerable noise. Moreover, the present scheme is sufficiently fast for its real-life applications.