Abstract: Aiming at the problem that the Independent Component Analysis(ICA) can not be achieved if the observed variables are gaussian distribution, this paper presents ICA based on a set of uncorrelated discriminant features. Compared with whitening transform which accomplished by classical Principal component analysis, a set of uncorrelated discriminant features extracts features with less relativity, it is more useful for ICA. Experiments with ORL face image database have been performed and results demonstrate the effectiveness of this approach.

Key words: Independent Component Analysis(ICA), feature extraction, face recognition

摘要： 针对随机数据在高斯分布条件下，独立成分分析在本质上不可能实现的问题，提出在统计不相关矢量集上进行独立成分分析。与一般白化变换后的数据相比，基于统计不相关矢量集的数据具有更好的不相关性，在估计独立分量时性能更优越。在ORL人脸数据库上进行的实验结果证明了该算法的有效性。

关键词: 独立成分分析, 特征提取, 人脸识别

CLC Number: 

• TP391.41