一种新的基于DCT变换的线性判别分析  

A Novel Linear Discriminant Analysis Based on DCT

作　　者：黄国宏[1] 刘刚[2] 

机构地区：[1]广东工业大学信息工程学院,广东广州510006 [2]上海电力学院电力与自动化工程学院,上海200090

出　　处：《计算机技术与发展》2008年第6期97-100,共4页Computer Technology and Development

摘　　要：对于高维复杂模式识别问题,传统的线性判别分析通常首先采用PCA变换来降低模式的维数,然后再求取最优判别矢量集。然而PCA变换是以判别信息的损失为代价的,故无法保证所提取的特征是最优的。DCT变换具有"能量聚集特性"和变换的保距特性,文中正是基于此特性,提出一种新的基于DCT变换的线性判别分析方法,同时,也给出了一种在该模型下的最优判别矢量集的直接求解方法。实验表明,文中算法具有计算速度快、识别率高的优点。For the issue of high- dimensional complex pattern recognition, classical linear discriminant analysis usually use PCA transformation to reduce the dimensionality of the patterns, and then solve the optimal discriminant vector set. However, the process of reducing the dimensionality is at a cost of losing the discriminant information. DCT transformation has the properties of ＂energy compaction＂ and distance preserving. Based on the properties, takes the DCT as the preprocessing method of data dimensionality reduction. At the same time, a direct algorithm to solve the optimal discriminant vector set is also proposed. The experimental result shows that the present method is superior to the existing methods in terms of correct classification rate.

关 键 词：DCT变换 主元分析 特征提取 线性判别分析 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]