An Adaptive Approach to Offline Handwritten Word Recognition
IEEE Transactions on Pattern Analysis and Machine Intelligence
Confidence Measures for an Address Reading System
ICDAR '03 Proceedings of the Seventh International Conference on Document Analysis and Recognition - Volume 1
| Hi-index | 0.00 |
While current Postal Address Readers have been adapted and fine tuned once and off--line in a central development lab, the concept of next generation readers presented in this paper will contain an inbuilt learning capability, which can be referred as learning from letters. This will be performed on different levels of reading and comprehension, enabling the system to adapt itself to slow changes of mail mix and writing conventions. This concept prevents the system from read quality degradation over time according to increasing discrepancies of internal and exterior world.Preconditions of an adaptive reading system are discussed and an outline of such a system is given. The system's adaptive components are explained, particularly for words and characters. Information is taken automatically from the documents themselves, in this case from the envelopes of mail pieces. The system presented here not only keeps track with the slow changes of input, but also optimizes itself within a given frame, say a nation wide application, to deviating site specific conditions, thus guaranteeing an optimal solution at each site, and not only at an average of all sites. Components have been publicly funded by the Project READ. The system presented here is planned as a succeeding project READ II funded by German Federal Ministry of Education and Research.