摘要
最临近支持向量机Proximal SVM(PSVM)是一种有效的、简单的和快速的近似支持向量机方法,识别效果和标准支持向量机相当,相比之下有较少处理时间.虽然有此优点,它的有效性仅仅是针对维数不高、大样本的数据集,而对于上千维甚至上万维的、小样本的人脸数据库情况没有人给出实验结果.文章把PSVM稍做改变,对四个公开的人脸库进行分类.同时采用几种典型的泛化线性鉴别分析(GLDA)方法,对人脸图像预处理.从识别率和所用的处理时间两方面以及用最近邻及最近特征线分类器进行对比,得出具有较好识别效果和处理时间的方法.
Proximal SVM classifier is an approximate SVM algorithm which is useful, simple and very fast. It has comparable test set correctness and faster computational time to that of standard SVM classifiers. And it is verified very useful only using in low-dimensional dataset and big sample set, but nobody gives the usage in high-dimension and Small Sample Size sample sets. This paper changes the PSVM algorithm to adapt for four open face dataset. And several generalized LDA algorithms are used in pre-processing for face image before it is classified. In this paper we mainly discuss the comparison performance of three classifiers, that is, Nearest-Neighbor classifier, Nearest Feature Line classifier and Linear PSVM classifier. And we find an algorithm that can bring a good result on face database from recognition rate and training time.
出处
《韩山师范学院学报》
2007年第3期29-36,共8页
Journal of Hanshan Normal University