Digit recognition using multiple classifiers

Abstract

The aim of this paper is to describe the combining of several classifiers to the recognition of printed digits using a novel approach to describe the digits by hybrid feature extraction. The study has been conducted using three different features computed from cavities, zonal extraction and retinal representation along with nine different classifiers, K-Nearest Neighbor - KNN - with different distance measure, Support Vector Machine - SVM -, decision tree, linear discriminant analysis - LDA -. Classifier combination is considered by Majority Voting method. Experimental tests carried on the multi-font and multi-size printed digits dataset.