METHOD OF COMBINING MULTIPLE EXPERTS FOR THE RECOGNITION OF UNCONSTRAINED HANDWRITTEN NUMERALS

For pattern recognition, when a single classifier cannot provide a dec ision which is 100 percent correct, multiple classifiers should be abl e to achieve higher accuracy. This is because group decisions are gene rally better than any individual's. Based on this concept, a method ca lled the ''Behavior-Knowledge Space Method'' was developed, which can aggregate the decisions obtained from individual classifiers and deriv e the best final decisions from the statistical point of view. Experim ents on 46,451 samples of unconstrained handwritten numerals have show n that this method achieves very promising performances and outperform s voting, Bayesian, and Dempster-Shafer approaches.