利用一个小样本人脸样本库创建数据库ar_test.mat中的训练样本数据和待测样本数据将生产的数据库文件.mat利用FDDL程序进行测试，分类；在程序最后先利用最短距离测试提取特征的效果本文利用SVD分解，并提取了每一幅图像的前3列特征向量（前三个主成分），共计3*200个特征向量，构成一个训练样本列元素最后的简单分类测试效果在90 ，区分度较好，能够适用于其他分类器实验..后续中，选取了第一主成分，发现识别率达到100 ；选取了前33个，反而低于了90 ，不知如何解释？-Using a small sample of face image database to create the training sample data in the database ar_test.mat and sample data will produce the database file.Mat using FDDL program to test it, classification　in the program finally to the effect of the short distance test feature extraction using SVD decomposition, and extracted from each image in the first 3 columns the feature vector (the first three principal components), a total of 3*200 feature vector, simple classification test results constitute a training sample column elements finally in 90 , discrimination is good, can be applied to other classification experiment. Later, the first principal component, found that the recognition rate is up to 100　 the first 33, but less than 90 , explain how do?