A System for the Automated Reading of Check Amounts - Some Key Ideas

  • Guido Kaufmann
  • Horst Bunke
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 1655)


In this paper we present a reading system for amounts extracted from bank and postal checks. We focus our description on the legal amount recognition and the combination of the recognition results for the legal and the courtesy amounts. For these tasks we developed and applied several new techniques. In our work we deal with German check amounts. The automated reading of German legal amounts is a great challenge since the literal words the legal amount is constructed of are completely connected. Our system was tested on a database with real checks from the Swiss postal services.


Hide Markov Model Recognition Rate Rejection Rate Digit String Handwritten Word 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


  1. 1.
    R. Bippus: 1-Dimensional and Pseudo 2-Dimensional HMMs for the Recognition of German Literal Amounts, Proc. of the 4th Int. Conference on Document Analysis and Recognition, Ulm, Germany (1997) 487–490Google Scholar
  2. 2.
    C. Y. Suen, L. Lam, D. Guillevic, N. W. Strathy, M. Cheriet, J. N. Said and R. Fan: Bank Check Processing, Int. Journal of Imaging Systems and Technology, Vol. 7 (1996) 392–403CrossRefGoogle Scholar
  3. 3.
    S. Impedovo, P. S. P. Wang, H. Bunke (Eds.): Automatic Bankcheck Processing, Special Issue of Int. Journal of Pattern Recognition and Artificial Intelligence, Vol. 11, No. 4 and 5 (1997)Google Scholar
  4. 4.
    T. Caesar, J. M. Gloger, E. Mandler: Preprocessing and Feature Extraction for a Handwriting Recognition System, Proc. of the 2nd Int. Conf. on Document Analysis and Recognition, Tsukuba, Japan (1993) 408–411Google Scholar
  5. 5.
    H. Bunke, M. Roth, E. G. Schukat-Talamazzini: Off-line Cursive Handwriting Recognition Using Hidden Markov Models, Pattern Recognition, Vol. 28, No. 9 (1995) 1399–1413CrossRefGoogle Scholar
  6. 6.
    G. Dimauro, S. Impedovo, G. Pirlo and A. Salzo: Automatic Bankcheck Processing: A New Engineered System, in Reference 4, 465–504Google Scholar
  7. 7.
    S. Knerr, V. Anisimov, O. Baret, N. Gorski, D. Price and J. C. Simon: The A2iA Recognition System for Handwritten Checks, Proc. of Int. Workshop on Document Analysis Systems, Malvern, Pennsylvania (1996) 431–494Google Scholar
  8. 8.
    J. Makhoul and R. Schwartz and C. LaPre and C. Raphael and I. Bazzi, “ Language-Independent and Segmentation-Free Techniques for Optical Character Recognition, Proc. of Int. Workshop on Document Analysis Systems, Malvern, Pennsylvania USA (1996) 99–114Google Scholar
  9. 9.
    G. Kaufmann, H. Bunke, T. M. Ha: Recognition of Cursively Handwritten Words Using a Combined Normalization / Perturbation Approach, Proc. of the 5th Int. Workshop On Frontiers in Handwriting, University of Essex, England (1996) 17–22Google Scholar
  10. 10.
    L. R. Rabiner, B. H. Juang: An Introduction to Hidden Markov Models, IEEE ASSP Magazine (1986) 4–16Google Scholar
  11. 11.
    Thien M. Ha, M. Zimmermann, H. Bunke: Off-line Handwritten Numeral String Recognition by Combining Segmentation-based and Segmentation-free Methods, Pattern Recognition, Vol. 31, No. 3 (1997) 257–272CrossRefGoogle Scholar
  12. 12.
    G. Kaufmann, H. Bunke: Amount Translation and Error Localization in Check Processing Using Syntax-Directed Translation, Proc. of Int. Conf. on Pattern Recognition, Brisbane, Australia (1998) 1530–1534Google Scholar
  13. 13.
    G. Kaufmann, H. Bunke: Error Localization and Correction in Check Processing, Proc. of Int. Conf. on Frontiers in Handwriting Recognition VI, Taeduck, Korea (1998) 77–87Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 1999

Authors and Affiliations

  • Guido Kaufmann
    • 1
  • Horst Bunke
    • 1
  1. 1.University of BernInstitut für Informatik und angewandte Mathematik Neubrückstrasse 10BernSwitzerland

Personalised recommendations