多分类器组合及其应用

Classifier Combination and Its Application

1 引言传统的模式识别系统通常只使用样本的某种单一特征描述和特定的一个分类器来进行分类。这种系统对于类别数较大、输入样本带噪声的问题很难获得好的分类效果。近来发现不同的特征描述、不同的分类器在分类性能上存在着彼此互补的现象,因此同时使用多种特征描述和多个分类器可能提高分类精确性。目前,多分类器组合的研究吸引了学者们广泛的注意。

Classifier combination has been successfully applied in many fields. From the point view of component classifier construction and combination,we review classifier combination in this paper. Four kinds of combination rules were discussed and compared theoretically. How to integrate component classifier construction and combination is the future work of classifier combination.