特征选择方法中的信号分析方法研究

特征选择是模式识别领域中最重要的环节，也是最根本的论题。论文从随机信号的傅立叶分析中自相关函数与谱密度函数之间的对应关系出发，提出了一种基于自相关函数的特征选择方法，并以实验方式进行了有效性验证。其研究意义还在于将这一特征选择方法与人工智能中的归纳学习方法相结合，其归纳性能比传统的熵最小化准则更为优越。

Method of FeatureSelection Using Signal Analysis

Feature selection is one of the most important issues in pattern recognition. From the viewpoint of signal analyses that there is a correlation between the signal's auto correlation function and spectrum density, a new kind of method for feature selection is presented in this paper. The validity of this method is verified through experiments. An important implication of the research work is that it finds a joint between feature selection and inductive learning using decision tree, and the result of that combination shows that it has higher performance than ID3 whose inductive strategy is the rule of minimal entropy.