Nonlinear fusion for face recognition using fuzzy integral

X CHEN,Z JING,G XIAO
DOI: https://doi.org/10.1016/j.cnsns.2005.07.005
IF: 4.186
2007-01-01
Communications in Nonlinear Science and Numerical Simulation
Abstract:Face recognition based only on the visual spectrum is not accurate or robust enough to be used in uncontrolled environments. Recently, infrared (IR) imagery of human face is considered as a promising alternative to visible imagery due to its relative insensitive to illumination changes. However, IR has its own limitations. In order to fuse information from the two modalities to achieve better result, we propose a new fusion recognition scheme based on nonlinear decision fusion, using fuzzy integral to fuse the objective evidence supplied by each modality. The scheme also employs independent component analysis (ICA) for feature extraction and support vector machines (SVMs) for classification evidence. Recognition rate is used to evaluate the proposed scheme. Experimental results show the scheme improves recognition performance substantially.
What problem does this paper attempt to address?