Unconstrained handwritten digit OCR using projection profile and neural network approach

Abstract

The recognition accuracy of an Optical Character Recognition (OCR) system mainly depends on the selection of feature extraction technique and the classification algorithm. This paper focuses on the recognition of handwritten digits using projection profile features. Vertical, Horizontal, Left Diagonal and Right Diagonal directions are the four different orientations that are used for abstracting features from each handwritten digit. A feed forward neural network is proposed for recognition of digits. 750 digit samples are collected from 15 writers; each writer contributed each of the 10 digits 5 times. Thus a local database containing 750 digit samples is used for training and testing of the proposed OCR system. Preprocessing of handwritten digits is also done before their classification. The combination of proposed feature extraction method along with back-propagation neural network classifier is found to be very effective as it yields excellent recognition accuracy. © 2012 Springer-Verlag GmbH Berlin Heidelberg.