摘要
借鉴了传统的信号频谱滤波原理,根据最小均方误差原则,在特征空间法模式识别中提出了特征空间维纳滤波算法,它充分利用""先验知识,为模式识别系统构造一个线性滤波器;理论和实验结果表明,维纳滤波使总偏差达到最小,实验结果还表明它对识别性能有一定改善。
Traditional signal frequency filtering theory is used for reference in this paper. According to the least mean square (LMS) deviation rule, an algorithm named 'eigenspace Wiener filtering' that constructs a linear filter for the recognition system taking full advantage of prior knowledge is presented in pattern recognition using eigenspace method. The algorithm as well as experiment shows that it results in the least deviation; and the experiment tells that it improves the recognition performance in a certain extent.
出处
《计算机工程》
CAS
CSCD
北大核心
2001年第10期32-33,143,共3页
Computer Engineering
基金
广东省自然科学基金项目
高等学校骨干(980603)
老师资助计划项目()[2000]65
关键词
模式识别
人脸识别
特征空间
特征滤波器
Pattern recognition
Face recognition
Eigenspace
Feature filter
