基于奇异值特征和统计模型的人像识别算法

人像识别是模式识别领域中的一个前沿课题。目前多数研究者采用人脸的一维和二维几何特征来完成识别任务。人脸的几何特征抽取以及这些特性的有效性都面临着很多问题，至今人像识别的研究仍然处于较低的水平。作者证明了图象矩阵的奇异值特征矢量具备了代数上和几何上的不变性以及稳定性，提出用它作为识别人脸的代数特征。本文的人像识别算法是基于奇异值特征矢量建立Ｓａｍｍｏｎ最佳鉴别平面上的正态Ｂａｙｅｓ分类模型。在本文的

Human Facial Image Recognition Algorithm Based on Singular Value Features and Statistical Model

Human facial image recognition is a frontier topic of pattern recognition.The state of the art in this research field still lies in low level up till now. In respect of facial image feature extraction,most researchers have adopted 1-D and 2-D geometric features of facial image for image recognition.However,the effectiveness of these features to recognition is open to question.We proved that Singular Value feature vector has some im portant properties of algebraic and geometric invariance and insensitiveness to noise Therefore,we use SV features to recognize human face by a normal pattern Bayes classifier based on Sammon's optimal discriminant plane in our experiment,the statistical recognition model.which is constructed by using 9 photos,can recognize these photos.It is interesting to note that our model can recognize photos of one person taken in different period