Abstract: Tensor Principal Component Analysis（TPCA） is a new Principal Component Analysis（PCA） method，which can solve the problem when image dimension is reduced by conventional Principal Component Analysis.Wavelet transform has good time-frequency analysis features and plays a dimension reduction role.According to the two advantages of the above algorithms，a new ear recognition algorithm based on Tensor Principal Component Analysis is proposed.Wavelet transform is firstly used and obtains four sub-band images.Tensor Principal Component Analysis is used to extract the feature of “LL” low frequency sub-band images.Support Vector Machine（SVM） method is used to identify.Experimental results show that the method compared with conventional Principal Component Analysis improves the recognition rate and shortens the identification time.On the USTB ear database test，the recognition rate of the proposed algorithm is 6% higher than that of the conventional Principal Component Analysis（PCA） algorithm，and the recognition time of the proposed algorithm is 35.23% of the conventional Principal Component Analysis（PCA） algorithm.

Key words: Tensor Principal Component Analysis（TPCA）, wavelet transform, ear recognition, Support Vector Machine（SVM）

摘要： 张量主成分分析是一种新的主元分析方法，可以解决传统PCA方法对图像进行降维时出现的问题。小波变换具有良好的时频分析特性，同时还能起到降维的作用。综合利用这两个方法的优点，提出了一种基于张量PCA的人耳识别新方法。该方法对人耳图像采用小波变换做预处理得到4个子带图像，对其中“LL”低频子带图像用张量PCA进行特征提取，用支持向量机的方法进行识别。实验结果表明，利用此方法与传统主成分分析识别相比，提高了识别率，缩短了识别时间。在USTB人耳库上实验，该方法的识别率比传统PCA方法提高了6%，识别时间为传统PCA方法的35.23%。

关键词: 张量主成分分析, 小波变换, 人耳识别, 支持向量机