New approach on structural feature extraction for character recognition

Abstract

In this paper, a new structural feature extraction technique was proposed. The features from character images are extracted through a new approach which based on a structural feature model called feature template. The feature template was created by feature vector extracted from a character image corpus then clustered into some N classes. These N classes of feature vectors will be used as template for generalized structural feature extraction. The structural features here are defined automatically through K-Mean clustering, unlike traditional techniques that was manually defined. Through a very simple template matching for character recognition application, the technique show a very promising result with average 95% accuracy. Keywords: feature extraction, structural feature, statistical feature, pattern recognition, character recognition.