关键词: 鉴别矢量, 多模态, 核主元分析, 广义主元分析, 典型相关分析

Abstract: Ear and face recognition have occupied important places in biometrics. Drastic pose change, however, has always been an obstacle for their practical application in true life. Therefore this paper proposes a novel method based on strengthened authentication vectors to solve pose problem of ear or face recognition. At the same time, the strengthened ear and face authentication vectors using this method are fused by serial strategy, parallel strategy（General Principal Component Analysis, GPCA） and Canonical Correlation Analysis（CCA） method, and are identified by the nearest neighbor method. Experimental results show that it is remarkably effective for single modal biometrics such as the posed ear or posed face to strengthen authentication vectors, and multimodal biometrics method outperforms conspicuously single modal one.

Key words: authentication vector, multimodal, Kernel Principal Component Analysis（KPCA）, General Principal Component Analysis（GPCA）, Canonical Correlation Analysis（CCA）