基于独立分量分析的一种改进的掌纹识别方法

An Improved Method of Palmprint Recognition Based on Independent Component Analysis

独立分量分析方法(ICA)用于掌纹识别不仅可以得到图像的高阶统计信息,而且使得各分量尽可能独立.但计算量大,分类特征不明显.本文在ICA基础上提出一种改进的新方法,首先用小波变换进行降维处理,使得在保证了图像信息特征的最小损失下大大减少计算量;再用ICA方法得到独立基向量;最后在独立基向量张成的子空间用Fisher线性鉴别(FLD)方法进行特征提取,使得图像有更好的分类信息.实验结果表明,改进后的方法不仅识别率得到提高,而且缩短了识别时间.

Independent Component Analysis is undertaken for palmprint recognition not only to get the image of the higher-order statistics information, but also to make each component as independent as possible. However, it needs a large quantity of calculation, and the classification feature is not obvious. An improved method is developed, based on ICA. First of all, wavelet transformtion was conducted to reduce dimension, which can guarantee minimum loss of image information features, and at the same time, reduce sharply the computional cost. Then ICA method was em- ployed to get the independent basis vectors. Finally, Fisher Linear Discriminant （FLD） method was used for feature extraction in the subspace fermed by the independent basis vectors to make the image have better classification information. Experimental results show that the improved method not only improves recognition rates, but also reduces recognition time.