ANN based Approach for Recognition of Handwritten Bengali Numerals by Analyzing Two Different Feature Extraction Methods

Mahua Nandy (Pal)

Department of CSE, MCKV Institute of Engineering, Howrah, India.


Character recognition is an emerging area of research in the field of image processing and pattern recognition because of different application potentials such as recognition of characters in video index or in vehicle license plate, postal automation, bank cheque processing etc. Recognition of handwritten Bengali characters has received much attention already. There are some effective character recognition approaches like artificial neural network, support vector machines, learning vector quantization. But it has been noticed that the performance of a character recognition system depends greatly on the recognition feature used. In this paper, a zone segmentation based feature extraction method and a zone distance metric based feature extraction method for handwritten Bengali numerals are analyzed and subsequent recognition through application of feed forward back propagation neural network classifiers are proposed. This system is tested on handwritten Bengali numerals and an overall accuracy of 96.49% and 98.04% respectively, by using two different feature extraction methods.

     Back to TOC