基于基空间转换的姿态人耳识别-《郑州大学学报(工学版)》

文章信息/Info

作者:: 王瑜; 穆志纯; 封筠.; 北京科技大学,信息工程学院,北京,100083, 北京科技大学,信息工程学院,北京,100083, 北京科技大学,信息工程学院,北京,100083

Keywords:: base space conversion; Human ear recognition; principal element analysis; nuclear principal element analysis; Attitude conversion matrix; Support vector machines

摘要:: 人耳识别是近年来新兴起的一种生物特征识别技术,然而姿态问题一直是其难点问题之一,针对此问题提出了一种基于基空间转换的新方法.首先,利用主元分析和核主元分析方法得到姿态人耳图像和正侧面人耳图像的基空间,通过计算两种基空间之间的线性转换关系求出姿态转换矩阵,然后将待测的姿态人耳图像特征集利用基空间姿态转换矩阵转变成正侧面人耳图像特征集,最后用支持向量机进行分类识别.实验结果表明,该方法与没有经过姿态转换的方法相比,识别率显著提高.

Abstract:: Human ear recognition is a biometric recognition technology that has emerged in recent years, but the pose problem has always been one of its difficult problems, and a new method based on base-space transformation is proposed to solve this problem. Firstly, the base space of the pose human ear image and the positive side human ear image is obtained by using the principal element analysis and nuclear principal element analysis methods, and the pose conversion matrix is obtained by calculating the linear conversion relationship between the two base spaces, and then the pose human ear image feature set to be measured is transformed into the positive side human ear image feature set by using the base space pose conversion matrix, and finally the support vector machine is used for classification and recognition. Experimental results show that the recognition rate of this method is significantly improved compared with the method without attitude transformation.