DWT and Sub-pattern PCA for Face Recognition Based on Fuzzy Data Fusion

In realistic situation, the outlier could affect the face recognition rate severely. To overcome this problem, we propose a novel face recognition system to improve the recognition rate. The system can be divided into three aspects. Firstly, the 2D discrete wavelet transform (2D-DWT) is used for noise removal. Secondly, we use the principle component analysis (PCA) to extract features. In fact, the feature information from global face is not so robust that we intend to extract the local features, called the sub-pattern PCA (sp-PCA). Thirdly, we introduce an improved fuzzy fusion algorithm called adaptive membership grade to improve the ability of similar data separation. The experimental results show that the proposed system reveals better recognition rate.