基于二维主分量分析的人耳身份识别研究

有效的特征提取方法是解决人耳身份识别任务的关键之一。以主分量分析(PCA)为代表的线性子空间方法在特征提取工作中得到了广泛应用。为了更有效地提取人耳图像特征并减少运算量，将基于二维图像矩阵的2D PCA方法应用于人耳身份识别。针对三个USTB人耳图像库，采用最近邻分类器，研究了选用不同的特征维数、贡献率，及不同的相似性测度时，2D PCA方法与传统的PCA方法的识别性能。交叉验证的实验结果表明:2D PCA方法较PCA方法获得了更短的训练时间和更高的识别率，说明基于图像矩阵的2D PCA方法是一种效率更高

Ear Recognition Based on 2D Principal Component Analysis

Feature extraction is one of the essential techniques to solve the problems of ear recognition effectively. The principal component analysis (PCA) method, as a typical linear subspace method, is applied extensively to feature extraction. The intensity of calculation can be reduced significantly and features can be extracted more effectively if 2D PCA method based on 2D image matrix is utilized for ear recognition. With the data from USTB human ear database 1, 2, and 3, the recognition performance of 2D PCA and PCA are compared with different feature dimensions, contribution rates and similarity measures when nearest neighbor classifier is adopted. The cross validation experimental results showed 2D PCA method can obtain higher recognition rate with less training time than PCA method. 2D PCA method based on two dimensional image matrix is effective and robust in ear recognition.