基于小波包和支持向量机的人脸识别

提出了用小波包变换、聚类分析和缩放支持向量机进行人脸识别的方法。首先，用小波包对图象进行2层分解，提取每个子频带的能量组成向量为该图象的特征；其次，对待识别图象进行聚类分析，以减少进入支持向量机的样本数；然后，在特征向量输入支持向量机之前先进行缩放处理，以减少运算量和提高识别准确率；最后，用支持向量集合距离度量进行人脸识别。实验表明：采用本文的方法，识别的正确率可达98％。

Human Face Recognition Based on Wavelet Package and SVM

In this paper a human face recognition method using wavelet package, clustering analysis and scaling SVM is presented. First, we use wavelet package to decompose the image for two times, and the feature vector is composite of every sub-band energy. Second, in order to reduce the samples of SVM, the clustering analysis is necessary. Then feature vectors are scaled before they are input to SVM, in order to reduce computation and improve veracity. At last, we use distance measure and SVM to recognize face images. Experiment show that recognition is correct to 98%.