基于小波分解和鉴别共同矢量的人耳识别

Ear Recognition Based on Wavelet Decomposition and Discriminative Common Vector

针对高维、小样本的情况下使用Fisher线形鉴别分析进行特征提取存在的病态奇异问题,提出一种新的特征提取方法,即先对人耳样本图像进行二维离散小波分解,再利用DCV算法对小波分解后的低频信息分量作进一步的降维处理。不仅克服了小样本问题,也解决了直接使用DCV算法对人耳图像降维所引起的计算量大和计算速度过慢的问题。实验证明,该方法具有较好的识别率,是一种有效的特征提取算法。

LDA is widely used for linear dimension reduction. However, LDA has some limitations that one of the scatter matrices is required to be nonsingular. Discriminative Common Vector（DCV） is one of the most successful methods which overcomes the problem caused by the singularity of the scatter matrices. But when DCV is directly used to reduce the dimension of the ear images, the computational expense of training is still relatively large. A new method is proposed in this paper that the low frequency subimages are obtained by utilizing two-dimensional wavelet transform and the features are extracted by applying DCV to the subimages. The experimental results show that the proposed method achieves better performance than Fisher face method and the computation burden is also reduced greatly.