摘要
提出并设计了模糊ART神经网络的结构、学习规则和识别算法。为了把该算法应用于人脸识别,定义了相似函数和匹配搜索方法,通过向量柱状图提取人脸特征,并用模糊ART神经网络对向量柱状图生成的特征向量进行识别。仿真实验结果表明,对于快速学习和非快速学习,不同的人具有不同的识别率,各有不同的警戒参数值可以使神经网络到达在线最大识别率82.25%和86%。
Structure, learning rule and recognition algorithm of fuzzy ART is described and designed. In order to apply the algorithm to face recognition, the vector histogram, choice and match function is presented. Simulation experiment shows that the vector histogram can extract face features. For fast and non-fast learning, there is different recognition rate for different person. The maximum online recognition rate for fast and non-fast learning is 82.25% and 86% when the fuzzy ART network parameters are selected properly.
出处
《计算机工程与设计》
CSCD
北大核心
2008年第18期4786-4788,4793,共4页
Computer Engineering and Design
关键词
神经网络
模糊ART
人脸识别
向量柱状图
识别算法
neural network
fuzzy ART
face recognition
vector histogram
recognizing algorithm
