Gaussian mixture model based handwritten numeral character recognition

Automated character recognition is a wide field and current area of research in image processing and pattern recognition. It has its applications in optical character recognition, handwritten character recognition, postal code readers, car number plate identification and even in biometrics for identification of persons on basis of their handwritings. In this paper, we present an automated system for identification and classification of handwritten numeral characters. Our system consists of three stages i.e. preprocessing, feature extraction and classification. We propose intensity, shape and geometric based features for accurate representation of each numeral character. The system applies a Gaussian Mixture Model using expectation maximization for classification of input characters. In order to check the accuracy of proposed system, we use United States Postal Service (USPS) database and the results show the validity of proposed system.