The need for automating the processing of an ever-increasing volume of documents combined with the availability of fast, inexpensive computers has resulted in an explosive growth in documentunderstanding algorithm development. High-volume document-processing applications (such as those required by census and revenue collection organizations) are dominated by a variety of forms. The regions of interest in forms are typically a number of handprinted fields, which must be recognized by a computer and converted to an ASCII code. A number of very successful algorithms have been developed by various researchers (Martin and Pittman (1). LeCun et al. (2). Anderson and Gaborski(3)). and the problem of recognizing isolated, handprinted characters is widely regarded as solved. The potential market for document processing systems makes it desirable to implement character recognition algorithms in low-cost, mass-produceable integrated circuits. This paper describes the hardware implementation of an algorithm which has excellent recognition accuracy and is simple in design. An automatic layout method. performance on a standard data set, SPICE simulation results and quantization issues are discussed.

Date of creation, presentation, or exhibit



Proceedings of the Third IEE International Conference on Artificial Neural Networks (2003) Appears in: Proceedings of the Third IEE International Conference on Artificial Neural Networks, 1993, pp. 36-40. Article is also available online at: ISBN: 0-85296-573-7Note: imported from RIT’s Digital Media Library running on DSpace to RIT Scholar Works in February 2014.

Document Type

Conference Proceeding

Department, Program, or Center

Chester F. Carlson Center for Imaging Science (COS)


RIT – Main Campus