針對人耳圖像自身的特點，并通過對現有生物識別技術的研究，本文嘗試采用了一種
基于核函數的Fisher 判別分析算法對人耳進行識別。該算法不僅可以有效地提取人耳特征，獲得較高的識別率；而且還可以解決因為光照和人耳旋轉角度等因素帶來的非線性問題。實驗表明: 采用基于徑向基核函數的Fisher 判別分析算法對人耳圖像進行識別，其識別率最
高，為98.701%。
關鍵詞： 人耳識別，特征提取，Fisher 判別分析算法，核函數
Application of Kernel Function Based Fisher Discriminant Analysis Algorithm in
Ear Recognition LIU Ying, MU Zhi-chun, YUAN Li (School of Information Engineering, University of Science and Technology Beijing, Beijing 100083) Abstract: According to the characteristic of ear image and the study of the existing biometricsrecognition technology, we try to use a Fisher discriminant analysis algorithm based on kernel function for ear recognition in this paper. The algorithm can not only extract ear feature effectively and achieve a high recognition rate, but also resolve the nonlinear issue caused by lighting variation and pose variation etc. The experiment shows that it achieves the highest recognition rate of 98.701% when using Fisher discriminant analysis algorithm based on RBF kernel function to recognize the ear image.
Key words: ear recognition，feature extraction，FDA algorithm，kernel function