A novel pair-wise recognition scheme for handwritten characters in the framework of a multi-expert configuration
A novel pair-wise recognition scheme for the recognition of handwritten characters is presented. The recognition scheme is based on using multiple neural networks to process the handwritten characters grouped in pairs. An intelligent combination scheme to combine the decisions of these individually formed and trained neural networks is developed and an overall decision tree for the identification of separate classes is realised. The whole concept is implemented and tested in the context of the classification of handwritten numerals and a substantial performance enhancement is gained.
Indexing termsNeural networks decision combination handwritten character recognition multiple expert classifiers.
Unable to display preview. Download preview PDF.
- 1.Kittler, J. and Hatel, M.: Improving recognition rates by classifier combination. Fifth International Workshop on Frontiers of Handwriting Recognition, Sep. 2-5, University of Essex, UK., (1996) 81–102.Google Scholar
- 3.Waterhouse, S. R. and Robinson, A. J.: Classification using hierarchical mixtures of experts. Proc. IEEE Workshop on Neural Networks for Signal Processing IV. (1994) 177–186.Google Scholar
- 4.Rahman, A. F. R. and Fairhurst, M. C.: A new approach to handwritten character recognition using multiple experts. Fifth International Workshop on Frontiers of Handwriting Recognition, Sep. 2-5, University of Essex, UK., (1996) 283–286.Google Scholar
- 5.Rahman, A. F. R. and Fairhurst, M. C.: Recognition of handwritten characters with a multi-expert system. IEE Workshop on Handwriting Analysis and Recognition-A European Perspective, May 23, London, (1996).Google Scholar
- 6.Fairhurst, M. C. and Rahman, A. F. R.: A Generalised approach to the recognition of structurally similar handwritten characters. IEE Proc. on Vision, Image and Signal Processing, 144(1) (1997) 15–22.Google Scholar
- 7.Rumelhart, D. E., Hinton, G. E. and Williams, R. J.: Learning internal representations by error propagation. in Parallel Distributed Processing 1, Rumelhart, D. E. and McClelland, J. L., eds., MIT Press, Cambridge, MA, (1986) 318–362.Google Scholar