一种高性能的人脸识别方法

在基于人脸图像的身份认证系统中，最关键的技术就是如何提取人脸图像的高质量特征以及如何进行分类识别，该文就提出了一种快速、准确的人脸图像识别方法。该方法利用基于核函数的学习算法，进行人脸图像的特征提取和分类。首先，该方法分别利用核主分量分析以及核Fisher算法提取人脸图像的特征，然后对这些特征进行合理的组合以构成组合特征向量，再利用支持向量机进行识别。实验结果显示，所提出的高性能人脸识别方法的识别率高，即使对于轻度光照不均匀的人脸图像、人脸姿势的有限变化图像，也能获得较高的识别率；同时，该方法的训练速度和识别速度也非常快，完全满足人脸识别系统实时性要求。

Higher performance method of face recognition

The keyproblems are how to extracting higher quality features of face images and how to classify the face images in the human identification system based on face images. A fast and higherperforce method of face recognition is presented. The features of face images are extracted and are classified by using a new kind of kernel-based learning algorithms. Kernel principal component analysis and kernel fisher discriminant analysis are adopted firstly to extract features of human faces. After the features of faces are acquired, the features in reason are assembled. Then the method performs recognition by using support vector machines which are trained by a incremental support vector machines learning algorithm. The experimental results display the method certainly has higher correct ratio of recognition even the face images have a little asymmetry of light and different poses of face images. The method also has fast speed of train and recognition. It can satisfy needs of real-time face recognition.