Kernel PCA for HMM-Based Cursive Handwriting Recognition
In this paper, we propose Kernel Principal Component Analysis as a feature selection method for offline cursive handwriting recognition based on Hidden Markov Models. In contrast to formerly used feature selection methods, namely standard Principal Component Analysis and Independent Component Analysis, nonlinearity is achieved by making use of a radial basis function kernel. In an experimental study we demonstrate that the proposed nonlinear method has a great potential to improve cursive handwriting recognition systems and is able to significantly outperform linear feature selection methods. We consider two diverse datasets of isolated handwritten words for the experimental evaluation, the first consisting of modern English words, and the second consisting of medieval Middle High German words.
Unable to display preview. Download preview PDF.
- 3.Vinciarelli, A., Bengio, S.: Off-line cursive word recognition using continuous density HMMs trained with PCA and ICA features. In: 16th Int. Conf. on Pattern Recognition, vol. 3, pp. 81–84 (2002)Google Scholar
- 6.Schölkopf, B., Smola, A., Müller, K.R.: Kernel principal component analysis. In: Advances in Kernel Method – Support Vector Learning, pp. 327–352. MIT Press, Cambridge (1999)Google Scholar
- 9.Shi, D., Ong, Y.S., Tan, E.C.: Handwritten chinese character recognition using kernel active handwriting model. In: Proc. IEEE Int. Conf. on Systems, Man and Cybernetics, vol. 1, pp. 251–255. IEEE, Los Alamitos (2003)Google Scholar
- 10.Takiguchi, T., Ariki, Y.: Robust feature extraction using kernel PCA. In: Proc. IEEE Int. Conf. on Acoustics, Speech, and Signal Processing, pp. 509–512. IEEE, Los Alamitos (2006)Google Scholar
- 12.Wüthrich, M., Liwicki, M., Fischer, A., Indermühle, E., Bunke, H., Viehhauser, G., Stolz, M.: Lanugage model integration for the recognition of handwritten medieval documents. Accepted for publication in Proc. IEEE Int. Conf. on Document Analysis and Recognition (2009)Google Scholar