Bank check reading: Recognizing the courtesy amount
We developed a check reading system which recognizes both the legal amount and the courtesy amount on bank checks. It addresses the problem of French, omni-scriptor, cursive handwriting recognition, and is designed to meet industrial requirements, such as high processing speed, robustness, and extremely low error rates.
Our system is based on several key ideas: (1) hierarchical organization; starting out with pixel images, the system elaborates intermediate representations, such as strokes, letters, and words, which are grouped to form objects in higher levels. (2) The objects of any hierarchical level are described in terms of soft decisions (probabilities). Hard decisions are only taken in the final amount recognition process. (3) Use of prior information when available. (4) Wherever possible, our system makes use of several complementary algorithms to accomplish a given task.
This paper deals particularly with the recognition of the courtesy (numeral) amount. Results obtained on a large data base of French bank checks are presented.
KeywordsRecognition Performance Character Recognition Candidate List Character Candidate Numeral Amount
Unable to display preview. Download preview PDF.
- 1.Baret O., Gorsky N., Simon J.-C., A system for recognition of handwritten literal amounts of checks. Proc. of the Workshop Document Analysis Systems, Kaiserslautern, 1994.Google Scholar
- 2.Gilloux M., Leroux M., Recognition of cursive script amounts on postal cheques. JET POST'93, Proc. of the 1st European Conf. on Postal Technologies, Nantes, pp. 705–712, 1993.Google Scholar
- 3.Dimauro G., Grattagliano M.R., Impedovo S., Pirlo G., A system for bankcheck processing. Proc.of the second ICDAR, Tsucuba, pp. 454–459, 1993.Google Scholar
- 4.Moreau J.V., A new system for automatic reading of postal checks. In: From Pixels to Features III. Frontiers in Handwriting Recognition, S. Impedovo and J.-C.Simon, eds., North-Holland, 1992.Google Scholar
- 5.Simon J.-C, Baret O., Gorsky N., Reconnaissance d'ecriture manuscrite. C. R. Acad. Sci Paris, t. 318, Serie II, pp. 745–752, 1994.Google Scholar
- 6.Simon J.-C., Off-line cursive word recognition. Proc. of the IEEE, Vol. 80, No.7, pp. 1150–1161, 1992.Google Scholar
- 7.Kimura F., Shridar M., Chen. Z., Improvements of a lexicon directed algorithm for recognition of unconstrained handwritten words. Proc. of the second ICDAR, Tsucuba, pp. 18–22, 1993.Google Scholar
- 8.Bridle J.S., Probabilistic Interpretation of Feedforward Classification Network Outputs with Relationships to Statistical Pattern Recognition. In Neurocomputing: Algorithms, Architectures and Applications, F. Fogelman-Soulie and J. Herault (eds.), NATO ASI Series, Springer, 1990.Google Scholar
- 9.Knerr S., Personnaz L., Dreyfus G., Handwritten digit recognition by neural networks with single-layer training. IEEE Transactions on Neural Networks, Vol. 3, No. 6, 1992.Google Scholar
- 10.Price D., Knerr S., Personnaz L., Dreyfus G., Pairwise neural network classifiers with probabilistic outputs. Proc. of Neural Information Processing Systems7, Denver, 1994.Google Scholar