Intelligent Character Recognition (ICR): a case study in ballot digit recognition

ICR is an emerging technology for the recognition of hand writing (i.e. turning images of hand written characters into machine readable char-acters). The reliability of ICR is continuously improving. There are many commercial ICR engines available currently and most of them are capable of recognizing free hand writing, hand printed digits (or alphanumerics) and check boxes. This paper reports on a feasibility study to determine the accuracy of widely available commercial ICR engines for hand printed digits and check box recognition in ballot papers. The study concludes that current ICR engines are sufficiently robust and achieve very high levels of accuracy (i.e. 95-99%). However none of the current ICR engines achieved 100% accuracy level. Our research suggests that a combination of ICR engines and improved "voting" strategies could increase the accuracy rate close to 100%.