基于主成分分析的超声人脸识别算法研究  

Ultrasonic face recognition algorithm based on principal component analysis

作　　者：王婧瑶[1] 许勇[1] 曹本希[2] 杨军[2] 

机构地区：[1]中国科学院声学研究所通信声学实验室,北京100190 [2]中国科学院声学研究所噪声与振动重点实验室,北京100190

出　　处：《计算机工程与设计》2013年第8期2867-2871,共5页Computer Engineering and Design

基　　金：国家自然科学基金项目(11074279)

摘　　要：针对超声波探测人脸识别系统中多通道探测模式,从数据融合的角度对特征进行了优化,研究了基于主成分分析(principal components analysis,PCA)的数据降维和人脸特征提取算法。利用该算法对100人的自由表情样本进行特征提取,在保证识别率超过80%前提下,可显著降低特征向量的维数达80%以上,提高系统速度85%以上。实验结果表明,PCA算法能有效降低特征数据的维数,提高运算速度。Aimed at multi-channel detection mode in ultrasonic detection face recognition system,the facial features are optimized from the perspective of data fusion.The data dimensionality reduction algorithm and feature extraction method are studied,which are based on principal component analysis.The effectiveness of the algorithm is proved.On the premise of ensuring the recognition rate over 80%,this algorithm,which is used to extract the features of 100 subjects＇ free expression,can significantly reduce the number of dimensions of the feature vectors up to 80% and improve the speed of system more than 85%.The experimental results show that the PCA algorithm can effectively reduce the dimension of features and improve the operation speed.

关 键 词：主成分分析 一维距离像 K近邻分类器 感知器算法 超声人脸识别 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]