Practical algorithms in C++
Experimental evaluation of expert fusion strategies
Pattern Recognition Letters - Special issue on pattern recognition in practice VI
StrCombo: combination of string recognizers
Pattern Recognition Letters - In memory of Professor E.S. Gelsema
Proceedings of the First International Workshop on Multiple Classifier Systems
MCS '00 Proceedings of the First International Workshop on Multiple Classifier Systems
Proceedings of the Second International Workshop on Multiple Classifier Systems
MCS '01 Proceedings of the Second International Workshop on Multiple Classifier Systems
Proceedings of the Third International Workshop on Multiple Classifier Systems
MCS '02 Proceedings of the Third International Workshop on Multiple Classifier Systems
Combination of Multiple Classifiers for Handwritten Word Recognition
IWFHR '02 Proceedings of the Eighth International Workshop on Frontiers in Handwriting Recognition (IWFHR'02)
Hi-index | 0.00 |
Frequently changing applications require combination systems with flexible architectures: depending on the recognition problem at hand, different sets of recognizers may be selected or different algorithms for ranking alternatives may be preferred. This paper presents an adaptive combination framework for OCR result strings and describes methods of finding suitable parametrizations automatically. Given an image of a text-line, string results are obtained from geometrical decomposition followed by character recognition. The combination strategy described tries to improve both steps by synchronizing input strings according to geometric criteria before applying classical voting algorithms (like Majority Vote or Borda Count) on the character level. The best string candidate is determined by an incomplete graph search. Quantitative results showing the difference between various voting strategies are presented for a two-recognizer system.