Multiple experts recognition system based on neural network

For numeral recognition, when a single classifier cannot provide a decision which is 100 percent correct, multiple classifier should be able to achieve higher accuracy. This is because group decisions are generally better than any individual´s. In this paper, as evidence, the differences between a ANN classifier and a traditional classifier are discussed. Based on this concept combination methods were developed, which can aggregate the decisions obtained from individual, derive the best final decisions. But different combination methods lead to different performance: accuracy, efficiency and so on. A ANN combining algorithm is developed. The authors analyze it and a voting algorithm within experiments. First experiments on 10000 samples of handwritten numerals have powerfully shown that the results of two different individual classifiers with same features are disparate in performance. Second experiment have discussed two disparate combination methods in contrast