基于小波变换和LDA/FKT及SVM的人耳识别

Human ear recognition based on wavelet transform,LDA/FKT and SVM

人耳识别技术是生物特征识别和人工智能领域的一个重要分支。针对人耳图像自身的特点并通过对现有方法的研究,本文提出了一种新的人耳识别方法,即先对人耳图像进行二维的离散小波分解,然后使用LDA/FKT算法对小波分解后得到的低频信息进行降维,进而获得图像的特征向量,最后采用支持向量机作为分类器对样本向量进行判别。实验证明,本文提出的方法不仅较好地解决了人耳识别中的小样本问题,而且还取得了比传统的PCA+LDA方法更高的识别率,是一种有效的人耳识别方法。

Human ear recognition is one of the important branches of biometrics and artificial intelligence. Considering human ear image characteristics and through research on methods in existence, the author proposes a new method in this paper, within which the low frequency sub-images are obtained by utilizing two-dimensional wavelet transform and the features are extracted by applying LDA/FKT to the sub-images and finally SVM is used as the classifier to make decision. Experimental results demonstrate that the new method can overcome the small sample size problem and also perform better than classical PCA＋LDA method in accuracy, so it is a an effective human ear recognition method.