摘要
ICA作为一种传统的人脸识别方法,虽然识别效果较好,但却没有考虑类别信息。为了将类别信息融入ICA方法中,尝试利用FLD和ICA相结合的方法对人脸进行识别处理,即在使用ICA方法获得训练模式的统计独立基向量的基础上,对基向量张成的子空间使用FLD方法。利用几个人脸数据库对该方法进行了实验。实验结果表明,使用上述方法进行人脸识别,其效果优于传统的PCA方法、FLD方法和ICA方法。
Although the traditional ICA method has good performance in face recognition, it does not consider class information. In order to incorporate class specific information into ICA, attempts to combine ICA with FLD. After getting statistically independent basis vectors by applying ICA to all training patterns, apply FLD to the subspace that the basis vectors span. Experimental results using several facial databases show that this method has better performance than PCA only based methods as well as other representative methods such as FLD and ICA methods.
出处
《计算机应用研究》
CSCD
北大核心
2005年第8期255-257,共3页
Application Research of Computers
关键词
主成分分析
独立成分分析
Fisher线性辨别分析
人脸识别
Principal Component Analysis
Independent Component Analysis
Fisher Linear Discriminant Analysis
Face Recognition