基于主元分析和Fuzzy　ART模型的人脸识别算法

Face Recognition Based on Principal Component Analysis and Fuzzy ART Neural Model

本文从子空间的理论出发，先用一定数量的人脸样本构造优化的“人脸空间”，“人脸空间”能更好的描述人脸图象矢量的分布。把人脸图象在子空间进行投影得到人脸的特征向量。然后在FuzzyART模型的基础上设计神经网络分类器。用40组共计400张人脸图象对系统进行测试，实验结果表明识别率在95％以上，系统具有良好的识别能力和鲁棒性。

A new face recognition algorithm based on Principal Components Analysis and Fuzzy ART neural model ispresented. The method collstructs optimal 'face space' from a number of typica1 face images, which can describe thedistribution of face image better. The face feature vector is obtained from the projection of face image on the constructed'face space'. Then a neural network recognizer is designed and implemented on the basis of the Fuzzy ART model. Testingthe system with 4Oo face images from 40 groups, the correct recognition rate is over 95%, which demonstrates the goodpartition capability and robustness of the recognition system.