基于二维Fisher线性判别的掌纹识别方法

在Fisher线性判别(FLD)中，类内离散矩阵总是奇异的。为了解决矩阵的奇异性问题，应用一种新的二维Fisher线性判别(2DFLD)直接进行矩阵投影。对于PolyU掌纹图像库，分别用PCA, PCA+FLD和2DFLD提取特征掌纹子空间，将待识别图像投影到低维子空间上，用余弦距离进行掌纹匹配。实验结果表明，与PCA相比，PCA+FLD的识别率最多提高1.18%。2DFLD识别率最高达到99.34%，比PCA+FLD提高7.61%，特征提取仅耗时0.047 s。

Palmprint Recognition Based on Two-dimensional Fisher Linear Discriminant

In the FLD-based recognition, the within-class scatter matrix is always singular. To overcome the above problem, a new way is to directly project the image matrix based on Two-Dimensional FLD(2DFLD). In PolyU palmprint database, this paper applies PCA, PCA+FLD and 2DFLD to extract the palmprint feature subspace. The images to be recognized are projected on small dimension subspace. A classifier to palmprint match based on cosine distance is used. Experimental results show that the recognition rate of PCA+FLD is about 1.18% higher than that of PCA. Compared with PCA+FLD, this method is able to yield recognition rate as high as 99.34%, with accuracy enhanced by 7.61%, while the feature extraction time is only 0.032 s.