一种基于布尔核SVM的人脸识别策略

A Recognizing Strategy for the Problem of Face Based on Boolean Kernel Function SVM

针对人脸识别问题,给出基于布尔核的SVM识别策略,该策略首先应用K-L变换对人脸图像进行特征参数的提取,然后将提取出的特征进行0-1化处理,用于构造基于布尔核的SVM。在标准人脸库ORL上的试验结果表明,基于布尔核函数的SVM在分类准确率上明显高于传统PCA算法,同时,也优于线性SVM。

For the problem of face recognition, a recognizing strategy based on Boolean kernel function SVM is proposed. Firstly, Karhunen-Loeve transform is employed to get the representation basis of face image set ; secondly, the extracted characteristics is translated into 0-1 format; thirdly, SVM based Boolean kernel function are used to classify. The face recognition experiment with ORL face databases shows that the proposed method led to significantly better classification accuracy compared with traditional PCA method and Linear SVM.