【作者】 孙大瑞； 吴乐南；

【Author】 SUN Da-rui WU Le-nan (Department of Radio Engineering,Southeast University,Nanjing,210096,China)

【机构】 东南大学无线电工程系；

【摘要】 线性辨别分析(LDA)特征空间的坐标轴是非正交的,并且基于LDA的分类器的性能容易受训练集变化的影响。Okada提出了一种优化正交系统,但运算量大,且增加了特征空间坐标轴的数目,影响分类器速度。本文提出一种新的正交分量辨别分析(OCDA),没有增加坐标轴的数目,并且稳定性和识别率都优于LDA。更多还原

【Abstract】 The LDA space is not orthonormai, and its performance is often affected by the variance of train set. Okata has put forward an optimal orthonormai system,but increases the number of coordinate axes of feature space and needs a long time. A new orthonormai component discriminant analysis (OCDA) is proposed in this paper, which does not increase the number of coordinate axes. The experiment on Yale and ORL databases shows that OCDA is reliable and has a superior recognition rate over LDA.更多还原