基于Sugeno模糊积分神经网络分类器融合方法在手写数字识别中的应用

Neural Network Multiple Classifier Fusion Method to Handwritten Numeral Recognition Based on Sugeno Fuzzy Integral

神经网络是模式识别中一种常见的分类器。针对同一个分类问题,构建多个分类器并把多个分类器进行融合可以提高分类系统的分类正确率、改善系统的稳健性。首先介绍了Sugeno模糊积分及Sugeno模糊积分神经网络分类器融合方法的一般原理,而后将其应用于手写数字识别,通过实际的案例验证了该融合方法的有效性和可行性。

The Neural Network is a common kind of classifier of pattern recognition.The accuracy and robustness of classification system can be improved through fusion of multiple classifiers.This paper,firstly,introduces Sugeno fuzzy integral and the theory of the multiple classifiers fusion method which is based on Sugeno fuzzy integral.Then the method is used to solve the handwritten numeral recognition problems.