基于最大值-均值的改进的频谱脸算法  

An Improved Spectroface Recognition Algorithm Based on Max-Mean Judgment

作　　者：张强[1] 王黎[1] 孙飞[1] 孙平[2] 郑南宁[1] 

机构地区：[1]西安交通大学人工智能与机器人研究所,陕西西安710049 [2]西安交通大学信息与系统科学研究所,陕西西安710049

出　　处：《计算机应用研究》2003年第9期74-76,共3页Application Research of Computers

基　　金：国家创新研究群体科学基金资助项目(60024301)

摘　　要：给出了一种基于最大值-均值的改进的频谱脸的人脸识别算法。为了提高识别性能,要求每个用户在注册时使用多张图像,这样在识别阶段就面临一个如何对待人脸库中每一个用户的多个不同的频谱特征的问题。为了解决这一问题,采用了一种称之为最大值-均值的判决方法。同时根据实验结果,选择了其它的小波来进行图像分解。实验表明,这种改进后的频谱脸算法的识别性能要优于原来的频谱脸算法。This paper proposes a technique called improved Spectroface algorithm based on maxmean judgment for face recognition.In order to get a better performance,it alters the original spectrum face algorithm with more than one enrolled image for each person at the training stage.Thus,a new problem that how to deal with multispectrum features of one person at the recognition stage occurs.To solve this problem,a new approach called maxmean judgment is introduced.The selected wavelet used in the improved algorithm is also altered according to our experiments.From experimental results,this improved face recognition algorithm has a better performance than the original one.

关 键 词：人脸识别 频谱脸 最大值-均值判决法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]