摘要
本文采用对称主元分析神经网络,用去冗余和权值正交相结合的方法对人脸进行特征提取和识别,具有所用特征数据量少,对测试图像进行特征提取运算量较小等特点.从而较好实现了大量人脸样本的存储和人脸的快速识别.模拟结果表明,在特征数量较少(20个)的情况下,图像仍可以较好地恢复和识别.
In this paper, PCA Neural Networks combined with deflation method and weight orthogonalization is used to extract thd features of face images. With this method, large amount of face images can be memorized and quickly recognized because of fewer features of face images and less complexity of computation for test image feature extraction. Simulation shows that the face images can be recovered and correctly recognized with only 20 features.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
1996年第1期52-58,共7页
Pattern Recognition and Artificial Intelligence
基金
国家攀登计划认知科学(神经网络)重大关键项目
江苏省自然科学基金
关键词
主元分析
神经网络
人脸识别
特征提取
Lateral Connection Networks, Deflation Method, Principal Component Analysis (PCA).
