基于独立成分分析和模糊支持向量机的人脸识别方法

成分分析(PCA)只能从2阶上去消除数据的相关性，传统支持向量机在解决多类问题时会出现分类的盲区问题，针对这两种情况，首先采用独立成分分析(ICA)方法解决了高阶上的数据相关性问题；同时在传统支持向量机中引入模糊隶属度函数，用模糊支持向量机解决了传统支持向量机在多类数据识别中的盲区问题。通过实验证明了该方法在人脸识别率上取得了显著提高。

Principal Component Analysis (PCA) only removes the second order correlation of data, and traditional support vector machine (SVM) has unclassifiable region on multi-class problem. In order to overcome these problems, Independent Component Analysis (ICA) is adopted to solve data correlation on high order and Fuzzy Support Vector Machine (FSVM) is applied on multi-class problem to solve the unclassifiable region. Experimental results demonstrate the effectiveness of this approach.

国家自然科学基金(60472060);; 南京理工大学科研发展基金