International Journal of Document Analysis and Recognition (IJDAR)

, Volume 8, Issue 1, pp 27–34

An MLP-orthogonal Gaussian mixture model hybrid model for Chinese bank check printed numeral recognition

Authors

    • Pattern Recognition Research CenterHarbin Institute of Technology
  • X. L. Tang
    • Pattern Recognition Research CenterHarbin Institute of Technology
  • Peng Liu
    • Pattern Recognition Research CenterHarbin Institute of Technology
Article

DOI: 10.1007/s10032-005-0143-x

Cite this article as:
Zhu, H., Tang, X.L. & Liu, P. IJDAR (2006) 8: 27. doi:10.1007/s10032-005-0143-x

Abstract.

A hybrid model based on the combination of an orthogonal Gaussian mixture model (OGMM) and a multilayer perceptron (MLP) is proposed in this paper that is to be used for Chinese bank check machine printed numeral recognition. The combination of MLP with OGMM produces a hybrid model with high recognition accuracy as well as an excellent outlier rejection ability. Experimental results show that the proposed model can satisfy the requirements of Chinese bank check printed numeral recognition where high recognition accuracy, high processing speed, and high reliability are needed.

Keywords:

Orthogonal Gaussian mixture modelMultilayer perceptronMultiple classifier systemsChinese bank check recognitionOutlier rejection

Copyright information

© Springer-Verlag 2005