机构地区: 广东工业大学信息工程学院
出 处: 《计算机仿真》 2008年第7期192-195,共4页
摘 要: 针对现有的基于Fisher准则的线性特征提取方法存在的不足,提出了一种新的改进的Fisher特征提取方法。通过重新定义类内散度矩阵与类间散度矩阵,削弱了边缘样本与边缘类别的影响,提高了准则模型的准确性,进而提高了判别矢量的特征提取能力。同时,也给出了一种实用的求解具有统计不相关的最优判别矢量集的方法,实验结果表明,算法得到的最优判别矢量具有更好的特征提取能力。 In order to overcome the drawback of the previous feature extraction methods based on the Fisher criterion,an improved Fisher discriminant analysis is developed in this paper.The new method weakens the influence of outlier samples and classes through redefining the between-class scatter and the within-class scatter,and accordingly also enhances the feature extraction ability of optimal discriminant vectors.A feasible solution of uncorrelated discriminant features to this new approach is also suggested.Experimental results show that the new optimal discriminant vectors have more powerful ability of feature extraction in terms of rate of classification.
领 域: [自动化与计算机技术] [自动化与计算机技术]