一种新的线性判别函数的加权方法

A New Weighted Way of Linear Discriminant Function

线性判别函数法是计算机模式分类的一种基本方法。该文以平均频率作为线性判别函数加权方法的一个特例,将N维的模式识别问题简化为一维问题,并且给出了两种确定分界点的方法。使用平均频率作为线性判别函数具有物理意义明确、计算量小的特点,可以用于以信号的功率谱作为特征矢量的模式识别方法之中。

The linear discriminant function is a basic method of the computer pattern classification. In this paper, average frequency is regarded as a peculiar weighted way of the linear discriminant function, and the classification of the N-dimension space is converted to the one dimension space. And two dividing point methods are put forward. The average frequency has clear physical sense and few computational complexity. It can be applied to the applications where the eigenvector is the power spectra.