摘要
归纳出了基于Gabor小波和支持向量机(SVM)的人脸识别算法在实际应用中所遇到的三个问题,即主元分析(PCA)降维过程中累积贡献率的选择,多项式核函数阶数的选择以及决策函数的确定。给出了累积贡献率和多项式核函数的阶数选择规则,提出了支持向量机和最大值(MAX)相结合的分类决策方法。最后,利用AT&T人脸库进行仿真比较研究。
This paper induces that there are three problems (the selection of the cumtdative contribution rate in the principle component analysis, the choice of the order of polynomial kernel function, and the determination of the decision function). Some rules which are used to select the eumtdative contribution rate and the order of polynomial kernel function are given. The decision method of mtdti-elassifieation based on the combination of SVM and maximum is proposed. At last, the simtdation of the comparative study is performed by using the AT&T face database.
出处
《计算机与现代化》
2007年第4期20-23,共4页
Computer and Modernization
基金
河南省自然科学基金资助项目(0523020600)
河南省高校杰出科研人才创新工程项目(2005KYCX012)