Character Recognition by Adaptive Statistical Similarity

Thomas Breuel

In: International Conference for Document Analysis and Recognition (ICDAR). International Conference on Document Analysis and Recognition (ICDAR) IEEE Computer Society 2003.


Handwriting recognition and OCR systems need to cope with a wide variety of writing styles and fonts, many of them possibly not previously encountered during training. This paper describes a notion of Bayesian sta- tistical similarity and demonstrates how it can be applied to rapid adaptation to new styles. The ability to general- ize across different problem instances is illustrated in the Gaussian case, and the use of statistical similarity Gaus- sian case is shown to be related to adaptive metric classi- fication methods. The relationship to prior approaches to multitask learning, as well as variable or adaptive metric classification, and hierarchical Bayesian methods, are dis- cussed. Experimental results on character recognition from the NIST3 database are presented.

CharRecAdapStatSim.pdf (pdf, 214 KB )

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence